Research On Road Extraction Techniques In UWB SAR Imagery

Ultra wide band synthetic aperture radar (UWB-SAR) operating at VHF/UHF bandcan penetrate the foliage and the earth surface to produce high resolution images ofconcealed targets. For this reason, UWB-SAR has become important in civil andmilitary field. As road is an important target to be extracted, it has great meaning inSAR image processing area. This paper is organized by procedures in the study of roadextraction in UWB-SAR imagery.The main achievements and contributions of the dissertation are as follows:1. The characters of UWB-SAR imagery have some dissimilarity from theordinary SAR imagery. Four classical clutter statistical models are used to analyze thecorresponding clutter model of greyscale in image. Adaptability of conventional roaddetectors in UWB-SAR imagery is studied. A contrast enhancement method whichconbines linear transform in grayscale subsection and grayscale histogram equalizationis presented for road extraction in original UWB-SAR imagery.2. With the view of road feature selection, this paper analyzed and comparedseven classical road edge extraction techniques and then investigated their adaptabilityin UWB-SAR images. By comparing their advantages and disadvantages,a flow chartand an analytical thought on road extraction methods for UWB-SAR images is formed.To make use of road features,a road edge extraction method which conbines regionphase homogeneity measurement and edge detector based on morphologic template isproposed. UWB-SAR images are used to testify its effectiveness.3. The method to extract line segment from edge extraction result is studied. Bystudying the adaptation of classical thinning algorithms in UWB-SAR line segmentextraction from edge extraction result, a way to select edge extracion result of real roadis brought out to reduce the false alarms in edge extraction result and a thinning methodbased on geometric main axis and Hough transform are proposed which decreases thespurs in thinning processing and meets the needs of road connecting. UWB-SARimages are used to testify its effectiveness.4. The method of line segment organization is studied. Aiming at the problem ofconnecting scattered line segment with tiny structure, an improved method includesegment classification idea which mainly based on arrangement mark organizition isproposed. Taking the advantage of keeping real road segments and discarding segments of false alarm at the meantime, it realizes road segment connection. A road widthjudgement method based on cross section match is performed. Road width is the key torealize road identity. A GUI platform is designed to perform the UWB-SAR roadextraction. The main idea of GUI design includes friendly GUI, module functionextended and data transmittal optimization.This thesis is based on factual UWB SAR imagery. It performs theoretical analysisand experimental emulation in P wave-band SAR imagery and shows good applicationvalue.