We present a novel systematic method for change detection in dual polarimetric (Dual-pol) synthetic aperture radar (SAR) images based on swarm intelligence techniques and fractal geometry. As the two main algorithms of swarm intelligence, ant colony optimization (ACO) and particle swarm optimization (PSO) have great potential in change detection. Additionally, fractal geometry appears to be a highly effective means of characterizing textural features in Dual-pol SAR images. The proposed method exploits fractal images to form a new difference image. Fractal images are computed based on wavelet multiresolution analysis. Moreover, by minimizing an optimal function value in the iteration process, the changes are detected by applying ACO and PSO to the difference image. Experimental results of detecting changes in Dual-pol SAR images reveal that the proposed method is a highly effective and efficient means of change detection in Dual-pol SAR images.
We present a change detection method for terrain covers from multi-temporal SAR images based on a spatial chaotic model which is known to adequately characterize the coherent process of SAR imaging. The major problem of SAR change detection rises from both the presence of speckle noise and the pixel mis-registration that are commonly seen in the remote sensing image. By means of chaotic model, we first transform the images to fractal domain and then perform the CFAR detection. Simulated tests are conducted to quantitatively evaluate the impacts of these two major error sources on detection rate. Results from satellite SAR for landcover change detection clearly show that the proposed algorithm not only the speckle noise can be effectively suppressed without scarifying the spatial resolution; the excruciating mis-registration error was taken into account and removed.
A new synthetic aperture radar (SAR) image change detection algorithm was developed using a spatially chaotic model for coherent SAR images. This algorithm is tolerant of misregistration even when the signal-to-noise ratio is low. Despeckling is unnecessary, making the method especially attractive wherever the radiometric changes are subtle. To demonstrate the algorithm's performance, a varied set of multitemporal polarimetric SAR images was used for testing. As a reference, the simple image difference (DI) technique and the principal-component analysis (PCA) were used for comparison. The proposed method performs very well and can detect minute changes despite the presence of speckle, for which both DI and PCA fail.
The fractal dimension is used in conjunction with a neural network to quantify effects of the chaotic behavior of radar clutter on the geometric aspects of target detection by synthetic aperture radar. To demonstrate the effectiveness of the proposed method, results are compared with those of the conventional constant-false-alarm-rate algorithm and the neural network technique. It is shown that the use of the fractal dimension substantially improves the detection performance based on some figures of merit, including detection rate, false-detection rate, and loss detection rate.
The multiple-classifiers approach is utilized to fully take into account the complementary and supplementary information from different data sources for terrain cover classification. To combine the outputs of classifiers that may be conditionally dependent, a variance reduction technique was adopted for optimal voting and thus best information extraction. The effectiveness and efficiency of utilizing the variance-reduction technique was demonstrated using SAR and optical images. Results show that the classification accuracy is dramatically improved by the proposed method.
A recently developed dynamic learning neural network (DL) has been successfully applied to multispectral imagery classification and parameter inversion. For multispectral imagery classification, it is noises and edges such as streets in the urban area and ridges in the mountain area in an image that result in misclassification or unclassification which reduce the classificalion rate. At the image spectrum point of view, noises and edges are the high frequency components in an image. Therefore, edge detection and noise reduction can be done by extracting the high frequency parts from an image to improve the classification rale. Although both noises and edges are the high frequency components, edges represent some userul information while noises should be removed. Thus, edges and noiscs must be separated when the high frequency parts are extracted. The conventional edge detection or noise reduction melhods could not distinguish edges from noises. A new approach, Wavelet transform, is selected to fulfill this requirement. The edge detection and noise reduction pre-processing using Wavelet transform and image classification using dynamic learning neural network are presented in this paper. The experimental results indicate that it did improve the classification rate.1