Building Detection And Extraction From SAR Image Based On Multi-Features Association

As a type of active microwave imaging sensor that can work at any time day or night and under all weather conditions, Synthetic Aperture Radar(SAR) has a wide application prospect in agriculture, forestry, water, geological and natural disasters and other civilian neighborhood, in the military field also has its advantages richly endowed by nature. The development of high resolution, multi polarization, multi band, and have ability to interference are the main trend in the development of SAR Technology, It also makes the research on the detection and extraction of buildings based on SAR images have more comprehensive and more accurate data source. High resolution single polarization SAR images can be extracted more features for building detection, greatly improved the limitations of single feature in traditional methods. With the development of multi polarization and multi band technology, especially the polarimetric images compared to the single polarization image although the loss of resolution information they contain more abundant polarimetric scattering information, what's more, the typical object model and the color information that similar to multi spectral images, making building detection and extraction more intuitive, physical interpretation more reasonable. Considering the limitations of the reality of the situation, getting fully polarimetric imagery is often more difficult and high cost, Built-up areas extraction in intensity imagery based on the method of multiple feature weighted fusion and Built-up areas detection in fully polarimetric SAR imagery based on the method of feature constraints and preserving polarimetric scattering characteristics are proposed in this paper, and extracts contour of building area based on the result of detection in the intensity imagery. The main studies and contributions in this paper are as followings:(1) Analysis of the typical characteristics of the building in SAR imagery. Research and analysis of the typical imaging effect of the building, on this basis, with further analysis of projection of buildings, provide a theoretical basis for the subsequent research.(2) Based on single polarization imagery, In this paper, a novel method for built-up areas extraction using both statistical and structural features is proposed according to the built-up texture features. First of all, statistical texture features and structural features are respectively extracted by classical method of gray level co-occurrence matrix and method of variogram function, and the direction information is considered in this process. Next, feature weights are calculated innovatively according to the Bhattacharyya distance. Then, all features are weighted fusion. At last, the fused image is classified with K-means classification method and the built-up areas are extracted after post classification process. The proposed method has been tested by domestic airborne P band polarization SAR images. The results show that this method can effectively and accurately extract built-up areas in high resolution airborne SAR imagery.(3) Based on polarimetric scattering information and texture information, to detect parallel buildings and orientation building more comprehensively, put forward a new building detection method of feature constraint and preserving polarization scattering characteristics. These two characteristics are the mean of polarization azimuth and entropy feature based on the apan obtained through comparison of the characteristics. The imagery pixels are divided into three main scattering types based on Yamaguchi decomposition. Then the two features as the constraint index, the volume scattering of orientation building pixel re-divided into double-bounce scatter according to threshold, will be re divided the three scattering type as the initial inputs to the Wishart classifier classification characteristics of polarimetric classification, finally detected by merging categories out of the building,make the new three types as the initial input of Wishart classifier, obtain the result of classification, finally, extract the building through the combined category. The proposed method has been tested by alos-2 Spaceborne L band polarimetric data. The results show that this method can effectively and accurately detect built-up areas in full polarimetric SAR imagery.