This is a list of JARS reviewers from 2017.

This guest editorial introduces the JARS Special Section on Improved Intercalibration of Earth Observation Data.

Remote sensing of the atmospheric greenhouse gases, methane (CH4) and carbon dioxide (CO2), contributes to the understanding of global warming and climate change. A portable ground-based instrument consisting of a commercially available desktop optical spectrum analyzer and a small sun tracker has been applied to measure the column densities of atmospheric CH4 and CO2 at Yanting observation station in a mountainous paddy field of the Sichuan Basin from September to November 2013. The column-averaged dry-air molar mixing ratios, XCH4/XCO2, are compared with those retrieved by satellite observations in the Sichuan Basin and by ground-based network observations in the same latitude zone as the Yanting observation station.

Several sites from around the world are being used operationally and are suitable for vicarious calibration of space-borne imaging platforms. However, due to the proximity of these sites (e.g., Libya 4), a rigorous characterization of the landscape is not feasible, limiting their utility for sensor intercalibration efforts. Due to its accessibility and similarities to Libya 4, the Algodones Sand Dunes System in California, USA, was identified as a potentially attractive intercalibration site for space-borne, reflective instruments such as Landsat. In March 2015, a 4-day field campaign was conducted to develop an initial characterization of Algodones with a primary goal of assessing its intercalibration potential. Five organizations from the US and Canada collaborated to collect both active and passive airborne image data, spatial and temporal measurements of spectral bidirectional reflectance distribution function, and in-situ sand samples from several locations across the Algodones system. The collection activities conducted to support the campaign goal is summarized, including a summary of all instrumentation used, the data collected, and the experiments performed in an effort to characterize the Algodones site.

This paper describes the current ground-based calibration results of Landsat 7 Enhanced Thematic Mapper Plus (ETM+), Landsat 8 Operational Land Imager (OLI), Terra and Aqua Moderate Resolution Imaging Spectroradiometer (MODIS), Suomi National Polar orbiting Partnership Visible Infrared Imaging Radiometer Suite (VIIRS), and Sentinel-2A Multispectral Instrument (MSI), using an automated suite of instruments located at Railroad Valley, Nevada, USA. The period of this study is 2012 to 2016 for MODIS, VIIRS, and ETM+, 2013 to 2016 for OLI, and 2015 to 2016 for MSI. The current results show that all sensors agree with the Radiometric Calibration Test Site (RadCaTS) to within ±5% in the solar-reflective regime, except for one band on VIIRS that is within ±6%. In the case of ETM+ and OLI, the agreement is within ±3%, and, in the case of MODIS, the agreement is within ±3.5%. MSI agrees with RadCaTS to within ±4.5% in all applicable bands.

We compare field hyperspectral bidirectional reflectance distribution function (BRDF) measurements acquired by a hyperspectral goniometer system known as the goniometer of the Rochester Institute of Technology (GRIT) during an experiment in the Algodones Dunes system in March 2015 with NASA Goddard’s light detection and ranging, hyperspectral, and thermal imagery of the site acquired during the experiment. We augment our field spectral data collection with laboratory hyperspectral BRDF measurements of samples brought back from the Algodones Dunes site using GRIT and our second-generation goniometer GRIT-two (GRIT-T). In these laboratory experiments, we vary geophysical parameters such as sediment density and grain size distribution of the sediments that would typically impact observed BRDF with the goal of extending the range of applicability of our resulting BRDF spectral libraries. Geotechnical measurements on site confirm the variability of geophysical parameters such as density and grain size distributions within the dune system, and measurements with GRIT and GRIT-T demonstrate the impact on observed spectral variation. By augmenting field spectral libraries with laboratory BRDF, we show that a greater proportion of the dune system is more faithfully represented in the expanded spectral library. Beyond developing appropriate calibration data for airborne and satellite imagery of the Algodones Dunes, laboratory and field studies also support goals to develop reliable retrieval methods for geophysical quantities such as sediment density directly from spectral imagery. We consider approaches based on the Hapke model. Our approaches use the invariance of the observed functional forms of the single scattering phase function, which must be invariant to differences in the illumination geometry. Fill factor is retrieved and correlates with expected direct measurements of sediment density in a laboratory setting.

Attempts to use pseudoinvariant calibration sites (PICS) for establishing absolute radiometric calibration of Earth observation (EO) satellites requires high-quality information about the nature of the bidirectional reflectance distribution function (BRDF) of the surfaces used for these calibrations. Past studies have shown that the PICS method is useful for evaluating the trend of sensors over time or for the intercalibration of sensors. The PICS method was not considered until recently for deriving absolute radiometric calibration. This paper presents BRDF data collected by a high-performance portable goniometer system to develop a temporal BRDF model for the Algodones Dunes in California. By sampling the BRDF of the sand surface at similar solar zenith angles to those normally encountered by EO satellites, additional information on the changing nature of the surface can improve models used to provide absolute radiometric correction. The results demonstrated that the BRDF of a reasonably simple sand surface was complex with changes in anisotropy taking place in response to changing solar zenith angles. For the majority of observation and illumination angles, the spectral reflectance anisotropy observed varied between 1% and 5% in patterns that repeat around solar noon.

Since its launch in 2013, the Thermal Infrared Sensor (TIRS) onboard Landsat 8 has exhibited artifacts in its image data that can be attributed to stray-light. A 3-year effort was initiated to develop a stray-light correction algorithm to support TIRS calibration. A methodology was developed to predict the additional (stray-light) signal on each detector from an estimate of the stray-light source locations in the sensor’s out-of-field area. The initial version of the algorithm estimated the magnitude of out-of-field radiance sources through the use of geostationary wide-field thermal band imagers. However, this methodology necessitated a strong effort to cross calibrate the two sensors. Ultimately, a variation of the algorithm was implemented operationally into the United States Geological Survey ground system that utilizes image data from TIRS itself as an estimate of the out-of-field stray-light sources. This paper highlights the intercalibration techniques investigated while developing the stray-light correction algorithm. The impact of differing view-angles, spectral responses, and collection times on at-sensor radiance was considered to assess the feasibility of using data from Geostationary Operational Environmental Satellite geostationary instruments to estimate the out-of-field stray-light radiance incident on the TIRS detectors. Results of the studies presented here illustrate the complexities associated with intercalibration in the thermal and provide justification for the current form of the TIRS stray-light correction algorithm.

The ability of sensors to detect changes in the Earth’s environment is dependent on retrieving radiometrically consistent and calibrated measurements from its surface. Intercalibration provides consistency among satellite instruments and ensures fidelity of scientific information. Intercalibration is especially important for spaceborne satellites without any on-board calibration, as accuracy of instruments is significantly affected by changes that occur postlaunch. To better understand the key parameters that impact the intercalibration process, this paper describes a simulation environment that was developed to support the primary mission of the Algodones Dunes campaign. Specifically, measurements obtained from the campaign were utilized to create a synthetic landscape to assess the feasibility of using the Algodones Dunes system as an intercalibration site for spaceborne instruments. The impact of two key parameters (differing view-angles and temporal offsets between instruments) on the intercalibration process was assessed. Results of these studies indicate that although the accuracy of intercalibration is sensitive to these parameters, proper knowledge of their impact leads to situations that minimize their effect. This paper concludes with a case study that addresses the feasibility of performing intercalibration on the International Space Station’s platform to support NASA’s CLARREO, the climate absolute radiance and refractivity observatory, mission.

The monitoring of vegetation near high-voltage transmission power lines and poles is tedious. Blackouts present a huge challenge to power distribution companies and often occur due to tree growth in hilly and rural areas. There are numerous methods of monitoring hazardous overgrowth that are expensive and time-consuming. Accurate estimation of tree and vegetation heights near power poles can prevent the disruption of power transmission in vulnerable zones. This paper presents a cost-effective approach based on a convolutional neural network (CNN) algorithm to compute the height (depth maps) of objects proximal to power poles and transmission lines. The proposed CNN extracts and classifies features by employing convolutional pooling inputs to fully connected data layers that capture prominent features from stereo image patches. Unmanned aerial vehicle or satellite stereo image datasets can thus provide a feasible and cost-effective approach that identifies threat levels based on height and distance estimations of hazardous vegetation and other objects. Results were compared with extant disparity map estimation techniques, such as graph cut, dynamic programming, belief propagation, and area-based methods. The proposed method achieved an accuracy rate of 90%.

The intercomparison of the reflective solar bands (RSB) between the instruments onboard a geostationary orbit satellite and a low Earth orbit satellite is very helpful in assessing their calibration consistency. Himawari-8 was launched on October 7, 2014, and Geostationary Operational Environmental Satellite (GOES)-R was launched on November 19, 2016. Unlike previous GOES instruments, the Advanced Himawari Imager (AHI) on Himawari-8 and the Advanced Baseline Imager (ABI) on GOES-R have onboard calibrators for the RSB. Independent assessment of calibration is nonetheless important to enhance their product quality. MODIS and visible infrared imager radiometer suite (VIIRS) can provide good references for sensor calibration. The intercomparison between AHI and VIIRS is performed over a pseudoinvariant target. The use of stable and uniform calibration sites provides comparison with accurate adjustment for band spectral difference, reduction of impact from pixel mismatching, and consistency of BRDF and atmospheric correction. The site used is the Strzelecki Desert in Australia. Due to the difference in solar and view angles, two corrections must be applied to compare the measurements. The first is the atmospheric scattering correction applied to the top of atmosphere reflectance measurements. The second correction is applied to correct the BRDF effect. The atmospheric correction is performed using a vector version of the Second Simulation of a Satellite Signal in the Solar Spectrum (6SV) model and the BRDF correction is performed using a semiempirical model. Our results show that AHI band 1 (0.47  μm) has a good agreement with VIIRS band M3 within 0.15%. AHI band 5 (1.61  μm) shows the largest difference (5.09%) with VIIRS band M10, whereas AHI band 5 shows the least difference (1.87%) in comparison with VIIRS band I3. The methods developed in this work can also be directly applied to assess GOES-16/ABI calibration consistency, a topic we will address in the future.

The lack of updated maps on large scale representations has encouraged the use of remotely piloted aircraft systems (RPAS) to generate maps for a wide range of professionals. However, some questions arise: do the orthomosaics generated by these systems have the cartographic precision required to use them? Which problems can be identified in stitching orthophotos to generate orthomosaics? To answer these questions, an aerophotogrammetric survey was conducted in an environmental conservation unit in the city of Goiânia. The flight plan was set up using the E-motion software, provided by Sensefly—a Swiss manufacturer of the RPAS Swinglet CAM used in this work. The camera installed in the RPAS was the Canon IXUS 220 HS, with the number of pixels in the sensor array of 12.1 megapixel, complementary metal oxide semiconductor 1  ∶  2.3^? (4000  ×  3000  pixel), horizontal and vertical pixel sizes of 1.54  μm. Using the orthophotos, four orthomosaics were generated in the Pix4D mapper software. The first orthomosaic was generated without using the control points. The other three mosaics were generated using 4, 8, and 16 premarked ground control points. To check the precision and accuracy of the orthomosaics, 46 premarked targets were uniformly distributed in the block. The three-dimensional (3-D) coordinates of the premarked targets were read on the orthomosaic and compared with the coordinates obtained by the geodetic survey real-time kinematic positioning method using the global navigation satellite system receiver signals. The cartographic accuracy standard was evaluated by discrepancies between these coordinates. The bias was analyzed by the Student’s t test and the accuracy by the chi-square probability considering the orthomosaic on a scale of 1  ∶  250, in which 90% of the points tested must have a planimetric error of <0.13  m with a standard deviation of 0.08 m and altimetric errors of <0.30  m with a standard deviation of 0.20 m. It was observed that some buildings in the orthomosaics were not properly orthorectified. The orthomosaics generated with 8 or more points reached the scale of 1  ∶  250, whereas without control points the scale was 10-fold smaller (1  ∶  3000).

An algorithm based on time-frequency analysis is proposed to select an imaging time window for the inverse synthetic aperture radar imaging of ships. An appropriate range bin is selected to perform the time-frequency analysis after radial motion compensation. The selected range bin is that with the maximum mean amplitude among the range bins whose echoes are confirmed to be contributed by a dominant scatter. The criterion for judging whether the echoes of a range bin are contributed by a dominant scatter is key to the proposed algorithm and is therefore described in detail. When the first range bin that satisfies the judgment criterion is found, a sequence composed of the frequencies that have the largest amplitudes in every moment’s time-frequency spectrum corresponding to this range bin is employed to calculate the length and the center moment of the optimal imaging time window. Experiments performed with simulation data and real data show the effectiveness of the proposed algorithm, and comparisons between the proposed algorithm and the image contrast-based algorithm (ICBA) are provided. Similar image contrast and lower entropy are acquired using the proposed algorithm as compared with those values when using the ICBA.

Synthetic aperture radar (SAR) image segmentation is the key to SAR image automatic interpretation. However, speckle noise, intensity inhomogeneity, and irregular shaped objects with changing edge often make the SAR image segmentation very difficult, and existing algorithms have high computational complexity. We propose a region-based level set method using the local image intensity information. To represent the statistical characteristics of speckle noise, we first use a gamma statistical distribution to model every segmented SAR image. We then apply a modified region mean estimation formula to efficiently segment SAR images with inhomogeneity. Finally, Gaussian filtering is employed to regularize the level set function, which can avoid reinitialization. The experimental results on synthetic and real-world SAR images demonstrate that the proposed method has less computation cost, faster convergence rate, and more accurate segmentation results.

Hyperspectral (HS) pansharpening aims to generate a fused HS image with high spectral and spatial resolution through integrating an HS image with a panchromatic (PAN) image. A guided filter (GF) and principal component analysis (PCA) hybrid HS pansharpening method is proposed. First, the HS image is interpolated and the PCA transformation is performed on the interpolated HS image. The first principal component ( PC1) channel concentrates on the spatial information of the HS image. Different from the traditional PCA method, the proposed method sharpens the PAN image and utilizes the GF to obtain the spatial information difference between the HS image and the enhanced PAN image. Then, in order to reduce spectral and spatial distortion, an appropriate tradeoff parameter is defined and the spatial information difference is injected into the PC1 channel through multiplying by this tradeoff parameter. Once the new PC1 channel is obtained, the fused image is finally generated by the inverse PCA transformation. Experiments performed on both synthetic and real datasets show that the proposed method outperforms other several state-of-the-art HS pansharpening methods in both subjective and objective evaluations.

Synthetic aperture radar (SAR) target discrimination is usually performed in a supervised manner. However, supervised methods for SAR target discrimination may need lots of labeled training samples, whose acquirement is costly, time consuming, and sometimes impossible. This paper proposes an SAR target discrimination method based on semisupervised co-training, which utilizes a limited number of labeled samples and an abundant number of unlabeled samples. First, Lincoln features, widely used in SAR target discrimination, are extracted from the training samples and partitioned into two sets according to their physical meanings. Second, two support vector machine classifiers are iteratively co-trained with the extracted two feature sets based on the co-training algorithm. Finally, the trained classifiers are exploited to classify the test data. The experimental results on real SAR images data not only validate the effectiveness of the proposed method compared with the traditional supervised methods, but also demonstrate the superiority of co-training over self-training, which only uses one feature set.

Distinctive and robust local feature description is crucial for remote sensing application, such as image matching and image retrieval. A descriptor for multisource remote sensing image matching that is robust to significant geometric and illumination differences is presented. In the proposed method, a traditional scale-invariant feature transform algorithm is applied for local feature extraction and a feature descriptor, named robust center-symmetric local-ternary-pattern (CSLTP) based self-similarity descriptor, is constructed for each extracted feature point. The main idea of the proposed descriptor is a rotation invariance description strategy on local correlation surface. Unlike common distribution-based descriptors or geometric-based spatial pooling descriptors, the proposed descriptor uses rotation invariance statistically strategic for CSLTP description on a correlation surface, which is inherently rotation invariant and robust to complex intensity differences. Then, a bilateral matching strategy followed by a reliable outlier removal procedure in the geometric transformation model is implemented for feature matching and mismatch elimination. The proposed method is successfully applied for matching various multisource satellite images and the results demonstrate its robustness and discriminability compared to common local feature descriptors.

Phase unwrapping (PU) is one of the key processes in reconstructing the digital elevation model of a scene from its interferometric synthetic aperture radar (InSAR) data. It is known that two-dimensional (2-D) PU problems can be formulated as maximum a posteriori estimation of Markov random fields (MRFs). However, considering that the traditional MRF algorithm is usually defined on a rectangular grid, it fails easily if large parts of the wrapped data are dominated by noise caused by large low-coherence area or rapid-topography variation. A PU solution based on sparse MRF is presented to extend the traditional MRF algorithm to deal with sparse data, which allows the unwrapping of InSAR data dominated by high phase noise. To speed up the graph cuts algorithm for sparse MRF, we designed dual elementary graphs and merged them to obtain the Delaunay triangle graph, which is used to minimize the energy function efficiently. The experiments on simulated and real data, compared with other existing algorithms, both confirm the effectiveness of the proposed MRF approach, which suffers less from decorrelation effects caused by large low-coherence area or rapid-topography variation.

Notch filtering is a popular technique to remove radio-frequency interference (RFI), due to its ease of implementation. However, significant degradation in performance occurs when RFI exhibits characteristics of high nonstationarity. Motivated by this challenge, this paper proposes a method that can not only effectively remove nonstationary RFI but can also obtain a well-focused image. The time-frequency (TF) distribution is constructed by short-time Fourier transform (STFT) with a nonoverlapping window in each slow time. The points of TF distribution belonging to RFI can be detected through a threshold and removed. Having sorted the remaining TF points, only points with the highest value belonging to the target echoes are retained, whereas the other points are set to zero and cast as missing samples. The linear relationship between the STFT vector and the range profile in the fractional Fourier transform domain is then established. Using sparsity, the range profile can be reconstructed by solving a minimization problem without any distortion. Finally, a two-dimensional image is obtained by the fractional range-Doppler algorithm. The results of simulations and experiments showed the effectiveness of the proposed method.

Unmanned aerial vehicles are suited to various photogrammetry and remote sensing missions. Such platforms are equipped with various optoelectronic sensors imaging in the visible and infrared spectral ranges and also thermal sensors. Nowadays, near-infrared (NIR) images acquired from low altitudes are often used for producing orthophoto maps for precision agriculture among other things. One major problem results from the application of low-cost custom and compact NIR cameras with wide-angle lenses introducing vignetting. In numerous cases, such cameras acquire low radiometric quality images depending on the lighting conditions. The paper presents a method of radiometric quality assessment of low-altitude NIR imagery data from a custom sensor. The method utilizes statistical analysis of NIR images. The data used for the analyses were acquired from various altitudes in various weather and lighting conditions. An objective NIR imagery quality index was determined as a result of the research. The results obtained using this index enabled the classification of images into three categories: good, medium, and low radiometric quality. The classification makes it possible to determine the a priori error of the acquired images and assess whether a rerun of the photogrammetric flight is necessary.

High-speed laser remote sensing of defects inside a concrete specimen was demonstrated. In the proposed measurement setup, high-power laser pulses irradiated a concrete surface to generate vibration that can be detected by an optical interferometer, which was constructed using photorefractive crystal. The laser-based remote sensing system achieved inspection speeds of 25 Hz. The predominant frequency of a mock-up defect that was embedded in a concrete specimen was measured. The inspection result was identical to that obtained using a conventional hammering method.

The features extracted from deep convolutional neural networks (CNNs) have shown their promise as generic descriptors for land-use scene recognition. However, most of the work directly adopts the deep features for the classification of remote sensing images, and does not encode the deep features for improving their discriminative power, which can affect the performance of deep feature representations. To address this issue, we propose an effective framework, LASC-CNN, obtained by locality-constrained affine subspace coding (LASC) pooling of a CNN filter bank. LASC-CNN obtains more discriminative deep features than directly extracted from CNNs. Furthermore, LASC-CNN builds on the top convolutional layers of CNNs, which can incorporate multiscale information and regions of arbitrary resolution and sizes. Our experiments have been conducted using two widely used remote sensing image databases, and the results show that the proposed method significantly improves the performance when compared to other state-of-the-art methods.

A scheme is presented for ground moving target indication in multichannel synthetic aperture radar (SAR) systems. After the effect of different Doppler centroids is compensated, the spatial spectrum is symmetric around zero for a stationary scatterer but asymmetric around zero for a moving scatterer. Thus, moving targets can be detected through measuring the asymmetry of their spatial spectra. Moreover, since the asymmetry of the spatial spectrum can be measured directly in the spatial domain, the procedure of estimating the spatial spectrum is avoided and the computational efficiency is improved greatly. The effectiveness of this scheme is verified by both simulated and real SAR data.

In the remote sensing community, blur is a prevalent phenomenon especially for image using system parameter away from ideal truth. According to the relationship between dark channel and convolution, a modified and more applicable method is proposed here, which mainly contains blind kernel estimation and nonblind deconvolution. A reconstructed energy function, minimizing the sparsity and the value of dark channel, generates an accurate kernel; an effective module is introduced to preserve the texture and avoid artifacts; and finally a parallel framework is designed for large image. From the objective metrics on demo case, our approach is more effective to model and remove blurs than previous approaches, and furthermore we demonstrate its activity with experiments on real images.

Compressive sensing has been successfully applied to inverse synthetic aperture radar (ISAR) imaging of moving targets. By exploiting the block sparse structure of the target image, sparse solution for multiple measurement vectors (MMV) can be applied in ISAR imaging and a substantial performance improvement can be achieved. As an effective sparse recovery method, sparse Bayesian learning (SBL) for MMV involves a matrix inverse at each iteration. Its associated computational complexity grows significantly with the problem size. To address this problem, we develop a fast inverse-free (IF) SBL method for MMV. A relaxed evidence lower bound (ELBO), which is computationally more amiable than the traditional ELBO used by SBL, is obtained by invoking fundamental property for smooth functions. A variational expectation–maximization scheme is then employed to maximize the relaxed ELBO, and a computationally efficient IF-MSBL algorithm is proposed. Numerical results based on simulated and real data show that the proposed method can reconstruct row sparse signal accurately and obtain clear superresolution ISAR images. Moreover, the running time and computational complexity are reduced to a great extent compared with traditional SBL methods.

The around-view monitoring (AVM) system is one of the major applications of advanced driver assistance systems and intelligent transportation systems. We propose an on-line calibration method, which can compensate misalignments for AVM systems. Most AVM systems use fisheye undistortion, inverse perspective transformation, and geometrical registration methods. To perform these procedures, the parameters for each process must be known; the procedure by which the parameters are estimated is referred to as the initial calibration. However, when only using the initial calibration data, we cannot compensate misalignments, caused by changing equilibria of cars. Moreover, even small changes such as tire pressure levels, passenger weight, or road conditions can affect a car’s equilibrium. Therefore, to compensate for this misalignment, additional techniques are necessary, specifically an on-line calibration method. On-line calibration can recalculate homographies, which can correct any degree of misalignment using the unique features of ordinary parking lanes. To extract features from the parking lanes, this method uses corner detection and a pattern matching algorithm. From the extracted features, homographies are estimated using random sample consensus and parameter estimation. Finally, the misaligned epipolar geographies are compensated via the estimated homographies. Thus, the proposed method can render image planes parallel to the ground. This method does not require any designated patterns and can be used whenever cars are placed in a parking lot. The experimental results show the robustness and efficiency of the method.

A randomized subspace-based robust principal component analysis (RSRPCA) method for anomaly detection in hyperspectral imagery (HSI) is proposed. The RSRPCA combines advantages of randomized column subspace and robust principal component analysis (RPCA). It assumes that the background has low-rank properties, and the anomalies are sparse and do not lie in the column subspace of the background. First, RSRPCA implements random sampling to sketch the original HSI dataset from columns and to construct a randomized column subspace of the background. Structured random projections are also adopted to sketch the HSI dataset from rows. Sketching from columns and rows could greatly reduce the computational requirements of RSRPCA. Second, the RSRPCA adopts the columnwise RPCA (CWRPCA) to eliminate negative effects of sampled anomaly pixels and that purifies the previous randomized column subspace by removing sampled anomaly columns. The CWRPCA decomposes the submatrix of the HSI data into a low-rank matrix (i.e., background component), a noisy matrix (i.e., noise component), and a sparse anomaly matrix (i.e., anomaly component) with only a small proportion of nonzero columns. The algorithm of inexact augmented Lagrange multiplier is utilized to optimize the CWRPCA problem and estimate the sparse matrix. Nonzero columns of the sparse anomaly matrix point to sampled anomaly columns in the submatrix. Third, all the pixels are projected onto the complemental subspace of the purified randomized column subspace of the background and the anomaly pixels in the original HSI data are finally exactly located. Several experiments on three real hyperspectral images are carefully designed to investigate the detection performance of RSRPCA, and the results are compared with four state-of-the-art methods. Experimental results show that the proposed RSRPCA outperforms four comparison methods both in detection performance and in computational time.

Feature-based change detection technologies using multitemporal remote sensing images are widely applied to find newly increased built-up areas (NIBUA) during the period of observation. This paper proposes an automatic object-based NIBUA extraction method using high-resolution remote sensing images, which is based on the integration of spectrum feature, edge-derived line-density-based visual saliency (LDVS) feature, and texture-derived built-up presence index (PanTex) feature. In the proposed method, image segmentation is first employed to obtain objects as basic units of detection. Next, due to the complexity of built-up areas in high-resolution images, LDVS images and PanTex images are produced for each temporal image, respectively. Then, to highlight built-up areas in complex scenes, a comprehensive measure for each object is calculated by integrating the newly increased measures from spectrum, LDVS, and PanTex features via a manner of Dempster–Shafer evidence fusion. Finally, the object-based NIBUA can be extracted by conducting binarization on the newly increased fused measure image. Comparison studies and experimental results demonstrate that our method can achieve a robust extraction of NIBUA from high-resolution remote sensing images with a higher detection accuracy. We conclude that this automatic way can play a positive role in reducing the artificial workload of the interpreters and the cost of monitoring a large-region area. It is encouraged to employ this method in a variety of applications, such as illegal construction land monitoring, land use/cover map update, and city planning.

The uneven illumination phenomenon reduces the quality of remote sensing image and causes interference in the subsequent processing and applications. A variational method based on Retinex with double-norm hybrid constraints for uneven illumination correction is proposed. The L₁ norm and the L₂ norm are adopted to constrain the textures and details of reflectance image and the smoothness of the illumination image, respectively. The problem of separating the illumination image from the reflectance image is transformed into the optimal solution of the variational model. In order to accelerate the solution, the split Bregman method is used to decompose the variational model into three subproblems, which are calculated by alternate iteration. Two groups of experiments are implemented on two synthetic images and three real remote sensing images. Compared with the variational Retinex method with single-norm constraint and the Mask method, the proposed method performs better in both visual evaluation and quantitative measurements. The proposed method can effectively eliminate the uneven illumination while maintaining the textures and details of the remote sensing image. Moreover, the proposed method using split Bregman method is more than 10 times faster than the method with the steepest descent method.

General component substitution (CS) pansharpening methods establish a global model over the whole image plane and may lead to unexpected spectral distortion. This paper proposes a context adaptive CS pansharpening method, which features a totally local-based processing procedure. The method consists of two processing blocks. The first block is to simulate a low-resolution panchromatic band by a local linear regression model between panchromatic and multispectral bands. The second block extracts spatial details and adds details back to multispectral bands in locally varying ratios. By recasting the local linear regression model into the guided filtering framework and analyzing the implicit statistical assumptions underlying CS methods, the strengths of the local-based pansharpening algorithm are addressed. Experiments test 7 pairs of images acquired from different sensors, such as GF-2, Quickbird, and Worldview-2. Both quantitative and qualitative evaluations reveal that the presented method can better preserve the spectral information than some state-of-the-art methods.

We propose a moving target analysis algorithm using speeded-up robust features (SURF) and regular moment in inverse synthetic aperture radar (ISAR) image sequences. In our study, we first extract interest points from ISAR image sequences by SURF. Different from traditional feature point extraction methods, SURF-based feature points are invariant to scattering intensity, target rotation, and image size. Then, we employ a bilateral feature registering model to match these feature points. The feature registering scheme can not only search the isotropic feature points to link the image sequences but also reduce the error matching pairs. After that, the target centroid is detected by regular moment. Consequently, a cost function based on correlation coefficient is adopted to analyze the motion information. Experimental results based on simulated and real data validate the effectiveness and practicability of the proposed method.

Doppler beam sharpening (DBS) is a critical technology for airborne radar ground mapping in forward-squint region. In conventional DBS technology, the narrow-band Doppler filter groups formed by fast Fourier transform (FFT) method suffer from low spectral resolution and high side lobe levels. The iterative adaptive approach (IAA), based on the weighted least squares (WLS), is applied to the DBS imaging applications, forming narrower Doppler filter groups than the FFT with lower side lobe levels. Regrettably, the IAA is iterative, and requires matrix multiplication and inverse operation when forming the covariance matrix, its inverse and traversing the WLS estimate for each sampling point, resulting in a notably high computational complexity for cubic time. We propose a fast IAA (FIAA)-based super-resolution DBS imaging method, taking advantage of the rich matrix structures of the classical narrow-band filtering. First, we formulate the covariance matrix via the FFT instead of the conventional matrix multiplication operation, based on the typical Fourier structure of the steering matrix. Then, by exploiting the Gohberg–Semencul representation, the inverse of the Toeplitz covariance matrix is computed by the celebrated Levinson–Durbin (LD) and Toeplitz-vector algorithm. Finally, the FFT and fast Toeplitz-vector algorithm are further used to traverse the WLS estimates based on the data-dependent trigonometric polynomials. The method uses the Hermitian feature of the echo autocorrelation matrix R to achieve its fast solution and uses the Toeplitz structure of R to realize its fast inversion. The proposed method enjoys a lower computational complexity without performance loss compared with the conventional IAA-based super-resolution DBS imaging method. The results based on simulations and measured data verify the imaging performance and the operational efficiency.

A remote sensing image fusion method based on the matting model is presented. The matting model refers to each band of the hyperspectral (HS) image that can be decomposed into three components, i.e., alpha channel, spectral foreground, and spectral background. First, the panchromatic (PAN) is sharpened to enhance the spatial details and the intensity component of the HS image is obtained by the nonlinear synthesis method. Different from the traditional matting model-based method in which the PAN image serves as the alpha channel, we subsequently transform the sharpened PAN image and the intensity component to the principal components analysis domain to obtain the first principal components channel. The first principal components channel is selected as the alpha channel of the HS image. The selected alpha channel contains most of the spatial information of both the PAN and HS images. Finally, the HS foreground and HS background are estimated by the alpha channel, and the fused HS image can be reconstructed perfectly. Experimental results reveal that the proposed method is superior to the existing state-of-the-art methods.

A hyperspectral image (HSI) has high-spectral and low-spatial resolution. As a result, most targets exist as subpixels, which pose challenges during target detection. Moreover, limitations of target and background samples always hinder the detection performance. In this study, a hybrid method for subpixel target detection of an HSI is developed. The scores of matched filter (MF) and adaptive cosine estimator (ACE) are used to construct a hybrid detection space. The reference target spectrum and background covariance matrix are improved iteratively based on the distribution property of targets, using the hybrid detection space. As the iterative process proceeds, the reference target spectra get closer to the central line, which connects the centers of the target and the background, resulting in a noticeable improvement in target detection. One synthetic dataset and two real datasets are used in the experiments. The results are evaluated based on the mean detection rate (DR), receiver operating characteristic curve, and observations of the detection results. For the synthetic experiment, the hybrid method improves more than 10 times with regard to the average DR compared with that of the traditional MF and ACE algorithms, which use N-FINDR target extraction and Reed–Xiaoli detector for background estimation.

Scene classification is one of the most important issues in remote sensing (RS) image processing. We find that features from different channels (shape, spectral, texture, etc.), levels (low-level and middle-level), or perspectives (local and global) could provide various properties for RS images, and then propose a heterogeneous feature framework to extract and integrate heterogeneous features with different types for RS scene classification. The proposed method is composed of three modules (1) heterogeneous features extraction, where three heterogeneous feature types, called DS-SURF-LLC, mean-Std-LLC, and MS-CLBP, are calculated, (2) heterogeneous features fusion, where the multiple kernel learning (MKL) is utilized to integrate the heterogeneous features, and (3) an MKL support vector machine classifier for RS scene classification. The proposed method is extensively evaluated on three challenging benchmark datasets (a 6-class dataset, a 12-class dataset, and a 21-class dataset), and the experimental results show that the proposed method leads to good classification performance. It produces good informative features to describe the RS image scenes. Moreover, the integration of heterogeneous features outperforms some state-of-the-art features on RS scene classification tasks.

Land leveling is the initial step for increasing irrigation efficiencies in surface irrigation systems. The objective of this paper was to evaluate potential utilization of an unmanned aerial system (UAS) equipped with a digital camera to map ground elevations of a grower’s field and compare them with field measurements. A secondary objective was to use UAS data to obtain a digital terrain model before and after land leveling. UAS data were used to generate orthomosaic images and three-dimensional (3-D) point cloud data by applying the structure for motion algorithm to the images. Ground control points (GCPs) were established around the study area, and they were surveyed using a survey grade dual-frequency GPS unit for accurate georeferencing of the geospatial data products. A digital surface model (DSM) was then generated from the 3-D point cloud data before and after laser leveling to determine the topography before and after the leveling. The UAS-derived DSM was compared with terrain elevation measurements acquired from land surveying equipment for validation. Although 0.3% error or root mean square error of 0.11 m was observed between UAS derived and ground measured ground elevation data, the results indicated that UAS could be an efficient method for determining terrain elevation with an acceptable accuracy when there are no plants on the ground, and it can be used to assess the performance of a land leveling project.

Robust feature point matching is a critical procedure in feature-based remote sensing image registration. A point-matching algorithm is proposed, which uses the similarity of local neighborhood information of point features. First, we establish a set of initial correspondences. Then we focus on removing incorrect correspondences (outliers) by local neighborhood information and increasing the number of correct correspondences (inliers). Finally, a global structure constraint is constructed for each remaining correct correspondence to increase the number of inliers and raise the correct rate simultaneously. Experimental results compared with three state-of-the-art methods illustrate that the proposed method can find more correct matching points with high accuracy.

Metaclassification ensemble approach is known to improve the prediction performance of snow-covered area. The methodology adopted in this case is based on neural network along with four state-of-art machine learning algorithms: support vector machine, artificial neural networks, spectral angle mapper, K-mean clustering, and a snow index: normalized difference snow index. An AdaBoost ensemble algorithm related to decision tree for snow-cover mapping is also proposed. According to available literature, these methods have been rarely used for snow-cover mapping. Employing the above techniques, a study was conducted for Raktavarn and Chaturangi Bamak glaciers, Uttarakhand, Himalaya using multispectral Landsat 7 ETM+ (enhanced thematic mapper) image. The study also compares the results with those obtained from statistical combination methods (majority rule and belief functions) and accuracies of individual classifiers. Accuracy assessment is performed by computing the quantity and allocation disagreement, analyzing statistic measures (accuracy, precision, specificity, AUC, and sensitivity) and receiver operating characteristic curves. A total of 225 combinations of parameters for individual classifiers were trained and tested on the dataset and results were compared with the proposed approach. It was observed that the proposed methodology produced the highest classification accuracy (95.21%), close to (94.01%) that was produced by the proposed AdaBoost ensemble algorithm. From the sets of observations, it was concluded that the ensemble of classifiers produced better results compared to individual classifiers.

To obtain the geographical position of the ground target accurately, a geolocation algorithm based on the digital elevation model (DEM) is developed for an airborne wide-area reconnaissance system. According to the platform position and attitude information measured by the airborne position and orientation system and the gimbal angles information from the encoder, the line-of-sight pointing vector in the Earth-centered Earth-fixed coordinate frame is solved by the homogeneous coordinate transformation. The target longitude and latitude can be solved with the elliptical Earth model and the global DEM. The influences of the systematic error and measurement error on ground target geolocation calculation accuracy are analyzed by the Monte Carlo method. The simulation results show that this algorithm can improve the geolocation accuracy of ground target in rough terrain area obviously. The geolocation accuracy of moving ground target can be improved by moving average filtering (MAF). The validity of the geolocation algorithm is verified by the flight test in which the plane flies at a geodetic height of 15,000 m and the outer gimbal angle is <47°. The geolocation root mean square error of the target trajectory is <45 and <7  m after MAF.

Recently, hyperspectral image (HSI) classification has become a focus of research. However, the complex structure of an HSI makes feature extraction difficult to achieve. Most current methods build classifiers based on complex handcrafted features computed from the raw inputs. The design of an improved 3-D convolutional neural network (3D-CNN) model for HSI classification is described. This model extracts features from both the spectral and spatial dimensions through the application of 3-D convolutions, thereby capturing the important discrimination information encoded in multiple adjacent bands. The designed model views the HSI cube data altogether without relying on any pre- or postprocessing. In addition, the model is trained in an end-to-end fashion without any handcrafted features. The designed model was applied to three widely used HSI datasets. The experimental results demonstrate that the 3D-CNN-based method outperforms conventional methods even with limited labeled training samples.

The collocated normalized radar backscattering cross-section measurements from the Global Precipitation Measurement (GPM) Ku-band precipitation radar (KuPR) and the winds from the moored buoys are used to study the effect of different sea-surface slope probability density functions (PDFs), including the Gaussian PDF, the Gram–Charlier PDF, and the Liu PDF, on the geometrical optics (GO) model predictions of the radar backscatter at low incidence angles (0 deg to 18 deg) at different sea states. First, the peakedness coefficient in the Liu distribution is determined using the collocations at the normal incidence angle, and the results indicate that the peakedness coefficient is a nonlinear function of the wind speed. Then, the performance of the modified Liu distribution, i.e., Liu distribution using the obtained peakedness coefficient estimate; the Gaussian distribution; and the Gram–Charlier distribution is analyzed. The results show that the GO model predictions with the modified Liu distribution agree best with the KuPR measurements, followed by the predictions with the Gaussian distribution, while the predictions with the Gram–Charlier distribution have larger differences as the total or the slick filtered, not the radar filtered, probability density is included in the distribution. The best-performing distribution changes with incidence angle and changes with wind speed.

Many methods of hyperspectral image classification have been proposed recently, and the convolutional neural network (CNN) achieves outstanding performance. However, spectral–spatial classification of CNN requires an excessively large model, tremendous computations, and complex network, and CNN is generally unable to use the noisy bands caused by water-vapor absorption. A dimensionality-varied CNN (DV-CNN) is proposed to address these issues. There are four stages in DV-CNN and the dimensionalities of spectral–spatial feature maps vary with the stages. DV-CNN can reduce the computation and simplify the structure of the network. All feature maps are processed by more kernels in higher stages to extract more precise features. DV-CNN also improves the classification accuracy and enhances the robustness to water-vapor absorption bands. The experiments are performed on data sets of Indian Pines and Pavia University scene. The classification performance of DV-CNN is compared with state-of-the-art methods, which contain the variations of CNN, traditional, and other deep learning methods. The experiment of performance analysis about DV-CNN itself is also carried out. The experimental results demonstrate that DV-CNN outperforms state-of-the-art methods for spectral–spatial classification and it is also robust to water-vapor absorption bands. Moreover, reasonable parameters selection is effective to improve classification accuracy.

The objective of this research is to test Sentinel-1 SAR multitemporal data, supported by multispectral and SAR data at other wavelengths, for fine-scale mapping of above-ground biomass (AGB) at the provincial level in a Mediterranean forested landscape. The regression results indicate good accuracy of prediction ( R2=0.7) using integrated sensors when an upper bound of 400  Mg ha−1 is used in modeling. Multitemporal SAR information was relevant, allowing the selection of optimal Sentinel-1 data, as broadleaf forests showed a different response in backscatter throughout the year. Similar accuracy in predictions was obtained when using SAR multifrequency data or joint SAR and optical data. Predictions based on SAR data were more conservative, and in line with those from an independent sample from the National Forest Inventory, than those based on joint data types. The potential of S1 data in predicting AGB can possibly be improved if models are developed per specific groups (deciduous or evergreen species) or forest types and using a larger range of ground data. Overall, this research shows the usefulness of Sentinel-1 data to map biomass at very high resolution for local study and at considerable carbon density.

This paper focuses on cross calibrating the GaoFen (GF-1) satellite wide field of view (WFV) sensor using the Landsat 8 Operational Land Imager (OLI) and HuanJing-1A (HJ-1A) hyperspectral imager (HSI) as reference sensors. Two methods are proposed to calculate the spectral band adjustment factor (SBAF). One is based on the HJ-1A HSI image and the other is based on ground-measured reflectance. However, the HSI image and ground-measured reflectance were measured at different dates, as the WFV and OLI imagers passed overhead. Three groups of regions of interest (ROIs) were chosen for cross calibration, based on different selection criteria. Cross-calibration gains with nonzero and zero offsets were both calculated. The results confirmed that the gains with zero offset were better, as they were more consistent over different groups of ROIs and SBAF calculation methods. The uncertainty of this cross calibration was analyzed, and the influence of SBAF was calculated based on different HSI images and ground reflectance spectra. The results showed that the uncertainty of SBAF was <3% for bands 1 to 3. Two other large uncertainties in this cross calibration were variation of atmosphere and low ground reflectance.

With the development of earth observation programs, many multitemporal synthetic aperture radar (SAR) images over the same geographical area are available. It is demanding to develop automatic change detection techniques to take advantage of these images. Most existing techniques directly analyze the difference image (DI), and therefore, they are easily affected by the speckle noise. We proposed an SAR image change detection method based on frequency-domain analysis and random multigraphs. The proposed method follows a coarse-to-fine procedure: in the coarse changed regions localization stage, frequency-domain analysis is utilized to select distinctive and salient regions from the DI. Therefore, nonsalient regions are neglected, and noisy unchanged regions incurred by the speckle noise are suppressed. In the fine changed regions classification stage, random multigraphs are employed as the classification model. By selecting a subset of neighborhood features to create graphs, the proposed method can efficiently exploit the nonlinear relations between multitemporal SAR images. The experimental results on two real SAR datasets and one simulated dataset have demonstrated the effectiveness of the proposed method.

In recent compressed sensing (CS)-based pan-sharpening algorithms, pan-sharpening performance is affected by two key problems. One is that there are always errors between the high-resolution panchromatic (HRP) image and the linear weighted high-resolution multispectral (HRM) image, resulting in spatial and spectral information lost. The other is that the dictionary construction process depends on the nontruth training samples. These problems have limited applications to CS-based pan-sharpening algorithm. To solve these two problems, we propose a pan-sharpening algorithm via compressed superresolution reconstruction and multidictionary learning. Through a two-stage implementation, compressed superresolution reconstruction model reduces the error effectively between the HRP and the linear weighted HRM images. Meanwhile, the multidictionary with ridgelet and curvelet is learned for both the two stages in the superresolution reconstruction process. Since ridgelet and curvelet can better capture the structure and directional characteristics, a better reconstruction result can be obtained. Experiments are done on the QuickBird and IKONOS satellites images. The results indicate that the proposed algorithm is competitive compared with the recent CS-based pan-sharpening methods and other well-known methods.

Multifeature decision fusion is an effective way to promote the performance of target recognition of synthetic aperture radar (SAR) images. This paper proposes a joint multifeature decision fusion strategy for target recognition in SAR images based on multitask compressive sensing (MtCS). The proposed method can exploit the intercorrelations among different features by enforcing the constraint on the sparsity pattern. Furthermore, the time consumption for MtCS is almost the same with that of single feature-based compressive classification, such as sparse representation-based classification. Experiments on the moving and stationary target acquisition and recognition dataset and comparison with several state-of-the-art methods demonstrate the validity of the proposed method.

The rapid increase in urbanization due to population growth leads to the degradation of vegetation in major cities. This study investigated the spatial patterns of the ecoenvironmental conditions of inhabitants of two distinct Asian capital cities, Beijing of China and Islamabad of Pakistan, by utilizing Earth observation data products. The significance of urban vegetation for the cooling effect was studied in local climate zones, i.e., urban, suburban, and rural areas within 1-km2 quantiles. Landsat-8 (OLI) and Gaofen-1 satellite imagery were used to assess vegetation cover and land surface temperature, while population datasets were used to evaluate environmental impact. Comparatively, a higher cooling effect of vegetation presence was observed in rural and suburban zones of Beijing as compared to Islamabad, while the urban zone of Islamabad was found comparatively cooler than Beijing’s urban zone. The urban thermal field variance index calculated from satellite imagery was ranked into the ecological evaluation index. The worst ecoenvironmental conditions were found in urban zones of both cities where the fraction of vegetation is very low. Meanwhile, this condition is more serious in Beijing, as more than 90% of the total population is living under the worst ecoenvironment conditions, while only 7% of the population is enjoying comfortable conditions. Ecoenvironmental conditions of Islamabad are comparatively better than Beijing where ∼61% of the total population live under the worst ecoenvironmental conditions, and ∼24% are living under good conditions. Thus, Islamabad at this early growth stage can learn from Beijing’s ecoenvironmental conditions to improve the quality of living by controlling the associated factors in the future.

Along with the rapid development of the unmanned aerial vehicle technology, multiple vehicle tracking (MVT) in aerial video sequence has received widespread interest for providing the required traffic information. Due to the camera motion and complex background, MVT in aerial video sequence poses unique challenges. We propose an efficient MVT algorithm via driver behavior-based Kalman filter (DBKF) and an improved deterministic data association (IDDA) method. First, a hierarchical image registration method is put forward to compensate the camera motion. Afterward, to improve the accuracy of the state estimation, we propose the DBKF module by incorporating the driver behavior into the Kalman filter, where artificial potential field is introduced to reflect the driver behavior. Then, to implement the data association, a local optimization method is designed instead of global optimization. By introducing the adaptive operating strategy, the proposed IDDA method can also deal with the situation in which the vehicles suddenly appear or disappear. Finally, comprehensive experiments on the DARPA VIVID data set and KIT AIS data set demonstrate that the proposed algorithm can generate satisfactory and superior results.

We developed an algorithm (named GMI_XCO2) to retrieve the global column-averaged dry air mole fraction of atmospheric carbon dioxide (XCO2) for greenhouse-gases monitor instrument (GMI) and directional polarized camera (DPC) on the GF-5 satellite. This algorithm is designed to work in cloudless atmospheric conditions with aerosol optical thickness (AOT)<0.3. To quantify the uncertainty level of the retrieved XCO2 when the aerosols and cirrus clouds occurred in retrieving XCO2 with the GMI short wave infrared (SWIR) data, we analyzed the errors rate caused by the six types of aerosols and cirrus clouds. The results indicated that in AOT range of 0.05 to 0.3 (550 nm), the uncertainties of aerosols could lead to errors of −0.27% to 0.59%, −0.32% to 1.43%, −0.10% to 0.49%, −0.12% to 1.17%, −0.35% to 0.49%, and −0.02% to −0.24% for rural, dust, clean continental, maritime, urban, and soot aerosols, respectively. The retrieval results presented a large error due to cirrus clouds. In the cirrus optical thickness range of 0.05 to 0.8 (500 nm), the most underestimation is up to 26.25% when the surface albedo is 0.05. The most overestimation is 8.1% when the surface albedo is 0.65. The retrieval results of GMI simulation data demonstrated that the accuracy of our algorithm is within 4 ppm (∼1%) using the simultaneous measurement of aerosols and clouds from DPC. Moreover, the speed of our algorithm is faster than full-physics (FP) methods. We verified our algorithm with Greenhouse-gases Observing Satellite (GOSAT) data in Beijing area during 2016. The retrieval errors of most observations are within 4 ppm except for summer. Compared with the results of GOSAT, the correlation coefficient is 0.55 for the whole year data, increasing to 0.62 after excluding the summer data.

Existing vegetation radiative transfer models seldom consider the effects of rice panicles on canopy spectrum. An ear-leaf model was established by adding a rice ear module to 4SAIL-RowCrop model. The geometrical optical model of the rice panicle was developed by assuming that the rice panicle is an upright cylinder. The rice panicle component was considered to be part of the canopy, and seven rice canopy scene components were defined including canopy without ear, the first and second kind of sunlit soil, the first and second kind of shaded soil, sunlit ear, and shaded ear. Thus, the directional radiation of the spike layer was described using the geometrical optical model. Reflectance and transmittance of the rice canopy were calculated using the 4SAIL model. The reflection and radiation characteristics of the soil layer were modeled using 4SAIL-RowCrop. Then, the row-sowed rice canopy scene reflectance was simulated by iterating reflectivity of the seven components in proportion. The ear-leaf model has good simulation accuracy for forward and backward reflectance with the mean R2, relative root mean square error, and RE values between the simulated and measured spectra of typical bands of 0.91, 19%, and 10%, respectively. Thus, the ear-leaf model provides a more technical support for the simulation of canopy reflectance and retrieval of growth parameters in rice.

Oil production in a steppe region disturbs the landscape and damages the steppe ecosystem. The objective of this research was to detect areas damaged by oil production in an oil field within the Russian Volga–Ural steppe region using winter Landsat imagery. We developed a practicable and effective approach using winter snow season multispectral Landsat satellite imagery. To this end, we applied seven algorithms of spectral or texture-based transformation: K-means, maximum likelihood estimation, topsoil grain size index, soil brightness, normalized differential snow index, tasselled cap, and co-occurrence measures. The co-occurrence texture measure variance shows the optimal result of identifying damaged areas. The unique feature of our method is that it can differentiate damaged areas from the bare soil of cropland within a cold steppe region where the area damaged by oil production is mixed with bare (fallow) croplands that have a polygonal shape similar to well pads. Such similarities can lead to confusion in object-based classification. Using the co-occurrence measures, we found that from 1988 to 2015, damaged area is nearly three times as big in the peak period of the oil field development (2001 and 2009) as in 1988. Landscape fragmentation also peaked in 2001 and 2009. Our approach for this project is useful and cost effective regular monitoring of damages from oil production for both the Volga–Ural steppe region and other cold steppe regions.

Five years of CloudSat observations during the Indian summer monsoon (ISM) season are used to investigate the vertical structure of stratocumulus (Sc) clouds and associated dynamics over the western part of Arabian Sea. The amount of Sc clouds formed over these regions are quantified in terms of their frequency of occurrence (FOC) and the same is used to study their vertical structure, which is very important from their radiative impact standpoint. The analysis revealed that Sc clouds predominantly form in the 0.5- to 3-km altitude region with mean thickness of 2.5 km over the study region. However, the maximum FOC of Sc clouds is found at ∼1.2-km altitude. The prevailing meteorological conditions over the study region are also investigated using ERA-Interim reanalysis datasets, to figure out why the Sc clouds persistently form over this region during ISM period. The analysis revealed that the presence of strong thermal inversion at 850-hPa level formed due to large-scale subsidence, presence of a low-level jet at 850-hPa level, and the advection of dry air at 650-hPa level from the Arabian Desert lead to the inhibition of vertical development of Sc clouds over the Arabian Sea. The significance of the present study lies in characterizing the vertical structure of Sc clouds formed persistently over the Arabian Sea during ISM and divulging the prevailing dynamics and thermal structure responsible for the formation of the Sc clouds over the study region.

For fast and more effective implementation of tracking multiple targets in a cluttered environment, we propose a multiple targets tracking (MTT) algorithm called maximum entropy fuzzy c-means clustering joint probabilistic data association that combines fuzzy c-means clustering and the joint probabilistic data association (PDA) algorithm. The algorithm uses the membership value to express the probability of the target originating from measurement. The membership value is obtained through fuzzy c-means clustering objective function optimized by the maximum entropy principle. When considering the effect of the public measurement, we use a correction factor to adjust the association probability matrix to estimate the state of the target. As this algorithm avoids confirmation matrix splitting, it can solve the high computational load problem of the joint PDA algorithm. The results of simulations and analysis conducted for tracking neighbor parallel targets and cross targets in a different density cluttered environment show that the proposed algorithm can realize MTT quickly and efficiently in a cluttered environment. Further, the performance of the proposed algorithm remains constant with increasing process noise variance. The proposed algorithm has the advantages of efficiency and low computational load, which can ensure optimum performance when tracking multiple targets in a dense cluttered environment.

Road extraction in imagery acquired by low altitude remote sensing (LARS) carried out using an unmanned aerial vehicle (UAV) is presented. LARS is carried out using a fixed wing UAV with a high spatial resolution vision spectrum (RGB) camera as the payload. Deep learning techniques, particularly fully convolutional network (FCN), are adopted to extract roads by dense semantic segmentation. The proposed model, UFCN (U-shaped FCN) is an FCN architecture, which is comprised of a stack of convolutions followed by corresponding stack of mirrored deconvolutions with the usage of skip connections in between for preserving the local information. The limited dataset (76 images and their ground truths) is subjected to real-time data augmentation during training phase to increase the size effectively. Classification performance is evaluated using precision, recall, accuracy, F1 score, and brier score parameters. The performance is compared with support vector machine (SVM) classifier, a one-dimensional convolutional neural network (1D-CNN) model, and a standard two-dimensional CNN (2D-CNN). The UFCN model outperforms the SVM, 1D-CNN, and 2D-CNN models across all the performance parameters. Further, the prediction time of the proposed UFCN model is comparable with SVM, 1D-CNN, and 2D-CNN models.

This paper discusses airborne light detection and ranging (LiDAR) point cloud filtering (a binary classification problem) from the machine learning point of view. We compared three supervised classifiers for point cloud filtering, namely, Adaptive Boosting, support vector machine, and random forest (RF). Nineteen features were generated from raw LiDAR point cloud based on height and other geometric information within a given neighborhood. The test datasets issued by the International Society for Photogrammetry and Remote Sensing (ISPRS) were used to evaluate the performance of the three filtering algorithms; RF showed the best results with an average total error of 5.50%. The paper also makes tentative exploration in the application of transfer learning theory to point cloud filtering, which has not been introduced into the LiDAR field to the authors’ knowledge. We performed filtering of three datasets from real projects carried out in China with RF models constructed by learning from the 15 ISPRS datasets and then transferred with little to no change of the parameters. Reliable results were achieved, especially in rural area (overall accuracy achieved 95.64%), indicating the feasibility of model transfer in the context of point cloud filtering for both easy automation and acceptable accuracy.

Soil erosion is an important phenomenon that contributes to the degradation of agricultural land. Even though it is a natural process, human activities can significantly increase its impact on land degradation and present serious limitation on sustainable agricultural land use. Nowadays, the risk of soil erosion is assessed either qualitatively by expert assessment or quantitatively using model-based approach. One of the primary factors affecting the soil erosion assessment is a cover-management factor, C-factor. In the Czech Republic, several models are used to assess the C-factor on a long-term basis based on data collected using traditional tabular methods. This paper presents work to investigate the estimation of both long-term and short-term cover-management factors using remote sensing data. The results demonstrate a successful development of C-factor maps for each month of 2014, growing season average, and annual average for the Czech Republic. C-factor values calculated from remote sensing data confirmed expected trend in their temporal variability for selected crops. The results presented in this paper can be used for enhancing existing methods for estimating C-factor, planning future agricultural activities, and designing technical remediations and improvement activities of land use in the Czech Republic, which are also financially supported by the European Union funds.

The objective of research is to evaluate the relationship between vegetation indices obtained in different phenological phases by hyperspectral sensors and crambe grain yield. The study was conducted in the winter crop season in 2015 in an experimental agricultural area in the city of Cascavel, Paraná. Spectral information about 12 random points were acquired during all phenological phases of the crambe crop. The data collected were separated into different spectral bands. Then, the vegetation indices normalized difference vegetation index and normalized difference moisture index (NDMI) were generated. Remote sensing data were correlated with grain yield, and linear regression models were elaborated and evaluated. At 66 days after sowing (DAS), in the beginning of the flowering phase, the NDMI presented a negative correlation with grain yield (Radj2: 0.49; RMSE: 134.80  kg ha−1). The red-light region (648 to 672 nm) at 66 DAS presented a positive linear correlation with grain yield (Radj2: 0.36; RMSE: 151.38  kg ha−1). The flowers in the canopy of the plant during the flowering phase contributed to the increase of the reflectance in the red-light and midinfrared regions. Excessive precipitation and winds during flowering, granulation, and maturation led to grain damage and yield variability, reducing the explanatory capacity of the production with the models.

The spectral and spatial information fusion can significantly improve the performance of hyperspectral image classification. A spectral–spatial feature extraction method, which is called morphological-based feature space discriminant analysis (MBFSDA), is proposed. In contrast to conventional feature extraction methods such as linear discriminant analysis and its different versions, which maximize the class discrimination in the spectral feature space, the proposed MBFSDA method maximizes the class discrimination in the spatial–spectral feature space. MBFSDA uses the spatial information (shape and size) of morphological profiles (MPs) in the FSDA. In the MBFSDA method, after applying the principal components analysis transformation on the hyperspectral image, an MP is created from each principal component. Each MP is considered as a multiband image that contains the spatial information instead of the spectral features. Then, in each MP, the between spatial bands scatters are maximized. As a result, different spatial features, which contain the minimum redundant information, are extracted. After that, the between-class and within-class scatter matrices are estimated using the extracted spatial features. Then, the Fisher criterion is used to maximize the class discrimination and to extract useful features for classification. The features extracted from each MP are given to a classifier. This process is repeated for each MP and eventually, the majority voting rule is used to provide the final classification map. The experimental results on three real hyperspectral images show the superiority of MBFSDA compared with some spectral–spatial classification methods in terms of classification accuracy.

Ordinary unsupervised change detection methods are mostly applied to the magnitude of spectral change vector (MSCV) of bitemporal images. However, the information contained in the different bands of the difference image is not fully exploited. We propose an unsupervised change detection method based on undecimated wavelet transform (UWT) in combination with chi-squared transform (CT). First, the UWT is applied to the difference image, and multiresolution images of each band of the difference image are obtained. Second, spatial constraint CT (SCT) is applied to the same resolution levels of each band of the difference image. Then, the results of SCT at all resolution levels are fused by a majority voting rule. After that, the fusion result as an initial solution of change detection is further improved by a Markov random field (MRF) model. In the implementation of the MRF model, the nonnegative matrix factorization is applied to fuse the MSCV and spectral angle mapper of bitemporal images to obtain the difference image. The experiments on three sets of multitemporal and multispectral images indicate that the proposed method performs well in comprehensive indices compared with other methods.

Ships on synthetic aperture radar (SAR) images will be severely defocused and their energy will disperse into numerous resolution cells under long SAR integration time. Therefore, the image intensity of ships is weak and sometimes even overwhelmed by sea clutter on SAR image. Consequently, it is hard to detect the ships from SAR intensity images. A ship detection method based on local region power spectrum of SAR complex image is proposed. Although the energies of the ships are dispersed on SAR intensity images, their spectral energies are rather concentrated or will cause the power spectra of local areas of SAR images to deviate from that of sea surface background. Therefore, the key idea of the proposed method is to detect ships via the power spectra distortion of local areas of SAR images. The local region power spectrum of a moving target on SAR image is analyzed and the way to obtain the detection threshold through the probability density function (pdf) of the power spectrum is illustrated. Numerical P- and L-band airborne SAR ocean data are utilized and the detection results are also illustrated. Results show that the proposed method can well detect the unfocused ships, with a detection rate of 93.6% and a false-alarm rate of 8.6%. Moreover, by comparing with some other algorithms, it indicates that the proposed method performs better under long SAR integration time. Finally, the applicability of the proposed method and the way of parameters selection are also discussed.

A fast implementation of the linear-prediction-based band selection (LPBS) method is proposed, and the method is denoted as fast-LPBS. The original LPBS method is an effective unsupervised band selection approach and can achieve satisfying classification performance in practice. It performs linear projections to measure the dissimilarity between spectral channels and adopts sequential forward search to avoid exhaustive search. However, LPBS is time-consuming because it involves a random initialization and much high-complexity calculation, such as high-order matrix multiplication, which limits its applications in many cases. The proposed fast-LPBS method derives the recursive formulae of linear prediction errors and achieves incremental calculation, which reduces the computational cost significantly. Moreover, a simplified initialization strategy is applied, which further reduces the computational complexity of fast-LPBS. Although the recursive formula and the simplified initialization strategy are used, these processes do not deteriorate the proposed method’s classification performance, in other words, fast-LPBS can yield almost the same results as LPBS in a much shorter time. Experimental results on simulated and real datasets verify that the proposed method can run fast and maintain the high classification performance as the original one.

Mapping vegetation dynamic types in mining areas is significant for revealing the mechanisms of environmental damage and for guiding ecological construction. Dynamic types of vegetation can be identified by applying interannual normalized difference vegetation index (NDVI) time series. However, phase differences and time shifts in interannual time series decrease mapping accuracy in mining regions. To overcome these problems and to increase the accuracy of mapping vegetation dynamics, an interannual Landsat time series for optimum vegetation growing status was constructed first by using the enhanced spatial and temporal adaptive reflectance fusion model algorithm. We then proposed a Markov random field optimized semisupervised Gaussian dynamic time warping kernel-based fuzzy c-means (FCM) cluster algorithm for interannual NDVI time series to map dynamic vegetation types in mining regions. The proposed algorithm has been tested in the Shengli mining region and Shendong mining region, which are typical representatives of China’s open-pit and underground mining regions, respectively. Experiments show that the proposed algorithm can solve the problems of phase differences and time shifts to achieve better performance when mapping vegetation dynamic types. The overall accuracies for the Shengli and Shendong mining regions were 93.32% and 89.60%, respectively, with improvements of 7.32% and 25.84% when compared with the original semisupervised FCM algorithm.

Three-dimensional (3-D) imaging and the 3-D micromotion feature extraction technique of space targets provide significant information for target classification and recognition. The idea of multiantenna interferometric processing from the interferometric inverse synthetic aperture radar (InISAR) is introduced into the 3-D imaging of micromotional space targets. After the modeling analysis of the micromotion characteristic induced by the target on the range–slow-time plane, a 3-D interferometric imaging method for space precession targets in wideband radar is proposed. The real 3-D images of the target in both the normal mode and the squint mode are obtained by the proposed method. Meanwhile, some 3-D precession features and structure parameters of the target are extracted accurately. Finally, simulation results validate the effectiveness and robustness of the proposed method. Compared with the existed approaches, this study can achieve real 3-D imaging and extract precession features and structure parameters for space precession target only by a monostatic radar with multi-antenna.

The relationship between surface infrared (IR) emissivity and soil moisture content has been investigated based on satellite measurements. Surface soil moisture content can be estimated by IR remote sensing, namely using the surface parameters of IR emissivity, temperature, vegetation coverage, and soil texture. It is possible to separate IR emissivity from other parameters affecting surface soil moisture estimation. The main objective of this paper is to examine the correlation between land surface IR emissivity and soil moisture. To this end, we have developed a simple yet effective scheme to estimate volumetric soil moisture (VSM) using IR land surface emissivity retrieved from satellite IR spectral radiance measurements, assuming those other parameters impacting the radiative transfer (e.g., temperature, vegetation coverage, and surface roughness) are known for an acceptable time and space reference location. This scheme is applied to a decade of global IR emissivity data retrieved from MetOp-A infrared atmospheric sounding interferometer measurements. The VSM estimated from these IR emissivity data (denoted as IR-VSM) is used to demonstrate its measurement-to-measurement variations. Representative 0.25-deg spatially-gridded monthly-mean IR-VSM global datasets are then assembled to compare with those routinely provided from satellite microwave (MW) multisensor measurements (denoted as MW-VSM), demonstrating VSM spatial variations as well as seasonal-cycles and interannual variability. Initial positive agreement is shown to exist between IR- and MW-VSM (i.e., R²  =  0.85). IR land surface emissivity contains surface water content information. So, when IR measurements are used to estimate soil moisture, this correlation produces results that correspond with those customarily achievable from MW measurements. A decade-long monthly-gridded emissivity atlas is used to estimate IR-VSM, to demonstrate its seasonal-cycle and interannual variation, which is spatially coherent and consistent with that from MW measurements, and, moreover, to achieve our objective of investigating the relationship between land surface IR emissivity and soil moisture.

Based on the study of earth infrared radiation and further requirement of anticloud interference ability for a spinning projectile’s infrared attitude measurement, a compensation method of cloud infrared radiation interference is proposed. First, the theoretical model of infrared radiation interference is established by analyzing the generation mechanism and interference characteristics of cloud infrared radiation. Then, the influence of cloud infrared radiation on attitude angle is calculated in the following two situations. The first situation is the projectile in cloud, and the maximum of roll angle error can reach   ±  20  deg. The second situation is the projectile outside of cloud, and it results in the inability to measure the projectile’s attitude angle. Finally, a multisensor weighted fusion algorithm is proposed based on trust function method to reduce the influence of cloud infrared radiation. The results of semiphysical experiments show that the error of roll angle with a weighted fusion algorithm can be kept within   ±  0.5  deg in the presence of cloud infrared radiation interference. This proposed method improves the accuracy of roll angle by nearly four times in attitude measurement and also solves the problem of low accuracy of infrared radiation attitude measurement in navigation and guidance field.

Clouds play a large role in the Earth’s global energy budget, but the impact of cirrus clouds is still widely questioned and researched. Cirrus clouds reside high in the atmosphere and due to cold temperatures are comprised of ice crystals. Gaining a better understanding of ice cloud optical properties and the distribution of cirrus clouds provides an explanation for the contribution of cirrus clouds to the global energy budget. Using radiative transfer models (RTMs), accurate simulations of cirrus clouds can enhance the understanding of the global energy budget as well as improve the use of global climate models. A newer, faster RTM such as the visible infrared imaging radiometer suite (VIIRS) fast radiative transfer model (VFRTM) is compared to a rigorous RTM such as the line-by-line radiative transfer model plus the discrete ordinates radiative transfer program. By comparing brightness temperature (BT) simulations from both models, the accuracy of the VFRTM can be obtained. This study shows root-mean-square error &lt;0.2  K for BT difference using reanalysis data for atmospheric profiles and updated ice particle habit information from the moderate-resolution imaging spectroradiometer collection 6. At a higher resolution, the simulated results of the VFRTM are compared to the observations of VIIRS resulting in a &lt;1.5  %   error from the VFRTM for all cases. The VFRTM is validated and is an appropriate RTM to use for global cloud retrievals.

A method is proposed to perform target deception jamming against spaceborne synthetic aperture radar. Compared with the traditional jamming methods using deception templates to cover the target or region of interest, the proposed method aims to generate a verisimilar deceptive target in various attitude with high fidelity using the electromagnetic (EM) scattering. Based on the geometrical model for target deception jamming, the EM scattering data from the deceptive target was first simulated by applying an EM calculation software. Then, the proposed jamming frequency response (JFR) is calculated offline by further processing. Finally, the deception jamming is achieved in real time by a multiplication between the proposed JFR and the spectrum of intercepted radar signals. The practical implementation is presented. The simulation results prove the validity of the proposed method.

A channel estimation (CE) procedure based on compressed sensing is proposed to estimate the multiple-input multiple-output sparse channel for traffic data transmission from drones to ground stations. The proposed procedure consists of an offline phase and a real-time phase. In the offline phase, a pilot arrangement method, which considers the interblock and block mutual coherence simultaneously, is proposed. The real-time phase contains three steps. At the first step, it obtains the priori estimate of the channel by block orthogonal matching pursuit; afterward, it utilizes that estimated channel to calculate the linear minimum mean square error of the received pilots. Finally, the block compressive sampling matching pursuit utilizes the enhanced received pilots to estimate the channel more accurately. The performance of the CE procedure is evaluated by simulating the transmission of traffic data through the communication channel and evaluating its fidelity for car detection after demodulation. Simulation results indicate that the proposed CE technique enhances the performance of the car detection in a traffic image considerably.

It is of great significance to investigate the changes in vegetation and its response to climate change in Tibet due to the sensitivity and vulnerability of the area to climate change. The spatiotemporal pattern of the normalized difference vegetation index (NDVI) and its trends between 2001 and 2015 were depicted using NDVI from the moderate resolution imaging spectroradiometer (MODIS). The responses of vegetation to climatic variables were analyzed through linear regression and correlation analysis with tropical rainfall measuring mission precipitation data and MODIS land surface temperature (LST) data. The results showed that (1) the average annual NDVI gradually decreased from the southeast to the northwest in accordance with the variations in LST and precipitation, (2) the annual NDVI increased from 2001 to 2015 at a rate of 0.3  ×  10^−  3  per year. The LST exhibited an average annual increase of 0.05°C while precipitation remained relatively stable, (3) the correlation between NDVI and precipitation was positive in the central region, whereas it became negative in the southeast and northeast. The correlation between NDVI and LST was opposite of that between NDVI and precipitation, and (4) the increases in NDVI in the tropical monsoon rain forest and rain forest, subalpine coniferous forest, and alpine meadow vegetation types in the southeast depended more on LST than precipitation. In contrast, the increases in NDVI responded strongly to precipitation in the alpine bush and meadow, alpine grassland, alpine desert, and alpine desert steppe vegetation types in the northwest.

The growing use of optimization for geographic object-based image analysis and the possibility to derive a wide range of information about the image in textual form makes machine learning (data mining) a versatile tool for information extraction from multiple data sources. This paper presents application of data mining for land-cover classification by fusing SPOT-6, RADARSAT-2, and derived dataset. First, the images and other derived indices (normalized difference vegetation index, normalized difference water index, and soil adjusted vegetation index) were combined and subjected to segmentation process with optimal segmentation parameters obtained using combination of spatial and Taguchi statistical optimization. The image objects, which carry all the attributes of the input datasets, were extracted and related to the target land-cover classes through data mining algorithms (decision tree) for classification. To evaluate the performance, the result was compared with two nonparametric classifiers: support vector machine (SVM) and random forest (RF). Furthermore, the decision tree classification result was evaluated against six unoptimized trials segmented using arbitrary parameter combinations. The result shows that the optimized process produces better land-use land-cover classification with overall classification accuracy of 91.79%, 87.25%, and 88.69% for SVM and RF, respectively, while the results of the six unoptimized classifications yield overall accuracy between 84.44% and 88.08%. Higher accuracy of the optimized data mining classification approach compared to the unoptimized results indicates that the optimization process has significant impact on the classification quality.

A hyperspectral imager was used to differentiate herbicide-resistant versus herbicide-susceptible biotypes of the agronomic weed kochia, in different crops in the field at the Southern Agricultural Research Center in Huntley, Montana. Controlled greenhouse experiments showed that enough information was captured by the imager to classify plants as either a crop, herbicide-susceptible or herbicide-resistant kochia. The current analysis is developing an algorithm that will work in more uncontrolled outdoor situations. In overcast conditions, the algorithm correctly identified dicamba-resistant kochia, glyphosate-resistant kochia, and glyphosate- and dicamba-susceptible kochia with 67%, 76%, and 80% success rates, respectively.

Multisensor data fusion is one of the most common and popular remote sensing data classification topics by considering a robust and complete description about the objects of interest. Furthermore, deep feature extraction has recently attracted significant interest and has become a hot research topic in the geoscience and remote sensing research community. A deep learning decision fusion approach is presented to perform multisensor urban remote sensing data classification. After deep features are extracted by utilizing joint spectral–spatial information, a soft-decision made classifier is applied to train high-level feature representations and to fine-tune the deep learning framework. Next, a decision-level fusion classifies objects of interest by the joint use of sensors. Finally, a context-aware object-based postprocessing is used to enhance the classification results. A series of comparative experiments are conducted on the widely used dataset of 2014 IEEE GRSS data fusion contest. The obtained results illustrate the considerable advantages of the proposed deep learning decision fusion over the traditional classifiers.

Spatially correlated errors are typically ignored in data assimilation, thus degenerating the observation error covariance R to a diagonal matrix. We argue that a nondiagonal R carries more observation information making assimilation results more accurate. A method, denoted TC_Cov, was proposed for soil moisture data assimilation to estimate spatially correlated observation error covariance based on triple collocation (TC). Assimilation experiments were carried out to test the performance of TC_Cov. AMSR-E soil moisture was assimilated with a diagonal R matrix computed using the TC and assimilated using a nondiagonal R matrix, as estimated by proposed TC_Cov. The ensemble Kalman filter was considered as the assimilation method. Our assimilation results were validated against climate change initiative data and ground-based soil moisture measurements using the Pearson correlation coefficient and unbiased root mean square difference metrics. These experiments confirmed that deterioration of diagonal R assimilation results occurred when model simulation is more accurate than observation data. Furthermore, nondiagonal R achieved higher correlation coefficient and lower ubRMSD values over diagonal R in experiments and demonstrated the effectiveness of TC_Cov to estimate richly structuralized R in data assimilation. In sum, compared with diagonal R, nondiagonal R may relieve the detrimental effects of assimilation when simulated model results outperform observation data.

The emergence of high resolution (HR) and ultra high resolution (UHR) airborne remote sensing imagery is enabling humans to move beyond traditional land cover analysis applications to the detailed characterization of surface objects. A residential roof condition assessment method using techniques from deep learning is presented. The proposed method operates on individual roofs and divides the task into two stages: (1) roof segmentation, followed by (2) condition classification of the segmented roof regions. As the first step in this process, a self-tuning method is proposed to segment the images into small homogeneous areas. The segmentation is initialized with simple linear iterative clustering followed by deep learned feature extraction and region merging, with the optimal result selected by an unsupervised index, Q. After the segmentation, a pretrained residual network is fine-tuned on the augmented roof segments using a proposed k-pixel extension technique for classification. The effectiveness of the proposed algorithm was demonstrated on both HR and UHR imagery collected by EagleView over different study sites. The proposed algorithm has yielded promising results and has outperformed traditional machine learning methods using hand-crafted features.

The wind turbine is a device that converts the wind’s kinetic energy into electrical power. Accurate and automatic extraction of wind turbine is instructive for government departments to plan wind power plant projects. A hybrid and practical framework based on saliency detection for wind turbine extraction, using Google Earth image at spatial resolution of 1 m, is proposed. It can be viewed as a two-phase procedure: coarsely detection and fine extraction. In the first stage, we introduced a frequency-tuned saliency detection approach for initially detecting the area of interest of the wind turbines. This method exploited features of color and luminance, was simple to implement, and was computationally efficient. Taking into account the complexity of remote sensing images, in the second stage, we proposed a fast method for fine-tuning results in frequency domain and then extracted wind turbines from these salient objects by removing the irrelevant salient areas according to the special properties of the wind turbines. Experiments demonstrated that our approach consistently obtains higher precision and better recall rates. Our method was also compared with other techniques from the literature and proves that it is more applicable and robust.

Erratum to reflect a change of the corresponding author of the paper.

Erratum to correct an error in the paper “Integrative image segmentation optimization andmachine learning approach for high quality land-useand land-cover mapping using multisource remotesensing data.”

Erratum corrects an author affiliation.