Research On Ship Detection And Feature Extraction Based On SAR Images

In order to obtain higher azimuth resolution radar images, Synthetic Aperture Radar(SAR) equivalently turns real aperture antenna of smaller size into larger size with dataprocessing method through detecting the relative motion between the radar platform and theprobe scene. It is of important application value to use SAR to detect ship targets in the vastocean, and the research on how to detect SAR ship targets can greatly promote china’s SARsystem into practical use. This article concentrates on speckle reduction of SAR image data,ship target detection algorithm and ship target feature extraction, etc.There are a large number of speckles in SAR images due to the imaging mechanism,which will cause great difficulties in processing and interpreting SAR images.Considering the different characteristics between the unipolar and multi-polarization SARimage, research and analysis has been done on median filtering, the MMSE estimationfiltering algorithm, Gamma MAP filtering algorithm which are mainly used in unipolarSAR image and polarimetric whitening filtering algorithm, optimal weight filtering,statistics estimated filtering algorithm based on best texture which are mainly used inmulti-polar SAR image respectively.This article focuses on the ship targets detection algorithm SAR images. Research hasbeen done on CFAR algorithm and KSW algorithm for unipolar SAR image data as well asSPAN algorithm, optimal polarization detection algorithm and polarization whitening filterdetection algorithm for multi-polarization SAR image data. Given the fact that traditionalalgorithm will blur the edge of ship target, we bring up ship target detection algorithm basedon decomposition theory in advantage of traditional self-adaptability of polarimetricwhitening filter detection algorithm based on the rectangular sliding window. Targetscattering mechanism taken into account, with Freeman decomposition results, thisalgorithm re-distinguishes uniformity area if there is target edge structural region in SARimage so that we can estimate parameters accurately and eliminate problem with fuzzytarget edge. Measured data from Qingdao Sea proves that the algorithm is better able toretain the target edge information, which provides a more accurate test results for further feature extraction, classification and identification.Finally, we extract the main parameters of the ship targets detected. Consideringcharacteristics of the SAR image data, we derive formulas and methods for specificcharacteristic parameters; with which we perform parameter estimation on features such asgeometry, location, movement, etc., with measured data in Qingdao.