The Detection On Vehicle Target In High Resolution SAR Image

The daily development of the earth observation technology via radar and the constantly enhanced ability to collecting SAR images have gradually made the Automatic Target Recognition (ATR) technology for synthetic aperture radar (SAR) images a research hot in the field currently. Target detecting, the result of which directly affects the extracting of target features subsequently and will finally affect the precision of target recognizing and classifying, is the basis of the automatic target recognition for SAR images. Ground vehicle like tanks, missile launching vehicle and panzers are important targets in SAR battlefield surveillance and earh observation. It is highly practical to research the method of automatically detecting the vehicle target in SAR images. This paper focuses on researching the target detecting and speckle noise suppression on the basis of real SAR images which contain vehicle targets.Firstly, the paper covered the SAR imaging principle, the characters of the targets in the SAR images and the the statistical features. It is concluded from the analysis on the statistical features of the real SAR image which contain vehicle targets that the gray probability distribution of SAR image is basically subject to the Rayleigh distribution (when evaluatingσof the SAR image's mean square error) throughout the whole gray-scale distribution of the SAR image.Secondly, the paper expatiated on the Speckle noise generating mechanism and model. On the basis of traditional filtering method and the combination of Lee algorithm and median filtering, proposed an improved wavelet method for noise suppression. The evaluation of various methods for noise suppression shows that the improved wavelet method in suppressing noise and preserving image details are more superior.Finally, the paper described the CFAR detection techniques and three kinds of CFAR detectors and analyzed the conventional two-parameter CFAR detection method based on the constant false alarm rate (CFAR) detection theory. It is concluded that the deviation of the actual false alarm rate from the theory false alarm rate is too large due to the clear difference between statistical characteristics of SAR image and the Gaussian distribution. It provided the CFAR detecting method which is based on Rayleigh distribution according to the priori knowledge that the probability distribution of SAR images obey Rayleigh distribution.Considering that the edge of SAR image will not be detected because of the over sized CFAR decting template, we provided an improved CFAR deceting method on the combination of Morphological. The result of the test on SAR image using various detecting methods shows that CFAR detecting method is more superior on detecting performance.