Research On Airplane Extraction From SAR Imagery

Airplane is an important military target. It is a significant task that extracts the airplanes quickly and effectively from SAR images. With the data of high resolution SAR images increasing,in order to be able to accurately extract the airplane,the paper adopts the hierarchical method to deal with it. First the paper analyzes the shape and scattering characteristics of the airplanes in the SAR images, and then rapidly extracts the region of the airport,detects the airplane and finally extracts the features of the airplanes. The work of this paper includes the following aspects:(1) The precondition of airplane detection in SAR images is the runway and the airport extraction. First,this paper gives a detailed analysis of the airports performance and characteristics in the SAR images,and then summarizes gray and structural features of the airport; In order to increase the speed and validity, a new method including the line extraction by improved LSD algorithm,the short line segments linking and the search of parallel lines is designed to extract the airport; Finally,based on the extraction result of the airport runway,use the region growing to get the whole airport;(2) Study the methods of the detection of airplanes in the region of the airport. First a detailed analysis of the geometry of the airplanes is given,and then the shape and scattering characteristics of the airplane are summarized in SAR images. Followed the target detection by the improved CFAR detection and the improved extension fractal detection are studied. Finally in order to get a whole target,a new method based on marked point process is proposed to detect the airplane,and the experiment shows the result is effective.(3) Extract the features of the airplanes. Before the feature extraction of the targets,the target slices are first segmented by the CFAR algorithm,the KSW algorithm and the MRF algorithm. The three segmentation methods are analyzed and compared through experimental results. Then,extract the traditional geometric features,peak features and texture features of the airplanes based on the results of the segmentation. The airplanes have their special geometric features. We improve the method of the azimuth estimation and the calculation of the length of the wingspan and the airframe; the new method of measure of the angle of sweepback that gets the projection length by angle search is proposed; the airplanes have the bilateral symmetry. An improved approach for the degree of bilateral symmetry based on the Hausdorff distance measure and the ratio of statistical symmetry points is proposed.