Automatic extraction of buildings from remote sensing data is an attractive research topic, useful for several applications, such as cadastre and urban planning. This is mainly due to the inherent artifacts of the used data and the differences in viewpoint, surrounding environment, and complex shape and size of the buildings. This paper introduces an efficient deep learning framework based on convolutional neural networks (CNNs) toward building extraction from orthoimages. In contrast to conventional deep approaches in which the raw image data are fed as input to the deep neural network, in this paper the height information is exploited as an additional feature being derived from the application of a dense image matching algorithm. As test sites, several complex urban regions of various types of buildings, pixel resolutions and types of data are used, located in Vaihingen in Germany and in Perissa in Greece. Our method is evaluated using the rates of completeness, correctness, and quality and compared with conventional and other “shallow” learning paradigms such as support vector machines. Experimental results indicate that a combination of raw image data with height information, feeding as input to a deep CNN model, provides potentials in building detection in terms of robustness, flexibility, and efficiency.
This study aims to highlight the differences, in terms of robustness and efficiency, of the use of LIDAR point clouds compared to dense image matching (DIM) point clouds at urban areas that contain buildings with complex structure. The application is conducted over an area in the Greek island of Milos using two different types of data: (a) a dense point cloud which extracted by DIM using a variation of the stereo-method semi-global matching (SGM) at RGB digital aerial images, and (b) a georeferenced LIDAR point cloud. For the case of the DIM point cloud, the following steps were applied: aerial triangulation, rectification of the original images to epipolar images, extraction of disparity maps and application of a 3D similarity transformation. The evaluations that were executed included urban and rural areas. At first step, a direct cloud-to-cloud comparison between the georeferenced DIM and LIDAR point clouds was carried out. Then, the corresponding orthoimages generated by the DIM and LIDAR point clouds undergo a quality control. Although the results show that the LIDAR point clouds respond better at such complex scenes compared to DIM point clouds, the latter gave promising results. In this context, the Quality Assurance issue is also discussed so as to be more efficient towards the challenge of the increasingly greater demands for accurate and cost effective applications.
Airborne LiDAR monitoring integrated with field data is employed to assess the fundamental period and the seismic loading of structures composing an urban area under prescribed earthquake scenarios. Α piecewise work-flow is adopted by combining geometrical data of the building stock derived from a LiDAR-based 3D city model, structural data from in-situ inspections on representative city blocks and results of soil response analyses. The procedure is implemented in the residential area of Kalochori, (west of Thessaloniki in Northern Greece). Special attention is paid to the in-situ inspection of the building stock in order to discriminate recordings between actual buildings and man-made constructions that do not conform to seismic design codes and to acquire additional building stock data on structural materials, typologies and number of stories which is not feasible by the LiDAR process. The processed LiDAR and field data are employed to compute the fundamental period of each building by means of code-defined formulas. Knowledge of soil conditions in the Kalochoti area allows for soil response analyses to obtain free-field at ground surface under earthquake scenarios with varying return period. Upon combining the computed vibrational characteristics of the structures with the free-field response spectra, the seismic loading imposed on the structures of the urban area under investigation is derived for each one of the prescribed seismic motions. Results are presented in GIS environment in the form of spatially distributed spectral accelerations with direct implications in seismic vulnerability studies of an urban area.
The constant technological evolution in Computer Vision enabled the development of new techniques which in conjunction with the use of Unmanned Aerial Vehicles (UAVs) may extract high quality photogrammetric products for several applications. Dense Image Matching (DIM) is a Computer Vision technique that can generate a dense 3D point cloud of an area or object. The use of UAV systems and DIM techniques is not only a flexible and attractive solution to produce accurate and high qualitative photogrammetric results but also is a major contribution to cost effectiveness. In this context, this study aims to highlight the benefits of the use of the UAVs in critical infrastructure monitoring applying DIM. A Multi-View Stereo (MVS) approach using multiple images (RGB digital aerial and oblique images), to fully cover the area of interest, is implemented. The application area is an Olympic venue in Attica, Greece, at an area of 400 acres. The results of our study indicate that the UAV+DIM approach respond very well to the increasingly greater demands for accurate and cost effective applications when provided with, a 3D point cloud and orthomosaic.
This paper studies the use of high resolution satellite optical and SAR images for 1:5,000 mapping production, which is essential for public work and environmental impact assessment studies. The images were used for the extraction of DEMs and their “fit for purpose” use was investigated, through the examination of parameters like accuracy, reliability and performance of morphological features. Orthoimages from satellite optical images using the produced DEMs with or without breaklines were produced. An application was developed on Antiparos island, a Greek island with irregular terrain. The data includes: (a) a triplet of Pleiades (1A, tri-stereo) satellite images, with a resolution of 0.5m, (b) a TanDEM-X Intermediate DEM, a preliminary version of the forthcoming TanDEM-X global DEM, and (c) an accurate DEM produced from the Greek National Cadastre & Mapping Agency S.A. was used as the reference DEM. The georeferencing of the optical images was computed using GCPs which were measured with GNSS. DEMs were extracted using all the possible combinations of the images triplet using automated image matching without any filtering or editing and were evaluated using the reference DEM. The combination of images which yielded the best DEM was then used to manually editing 3D points and collecting breaklines in order to produce a better DEM, which was also evaluated using various statistical measures and geo-morphological features. Orthoimages were created and evaluated using DEMs from optical and SAR data. A discussion about the use of the computed mapping products for the various stages of the public work studies is included.