The Research On Sea Target Detection Method In Airborne SAR Images

Synthetic aperture radar (SAR) is a kind of microwave imaging radar with high resolution, the possibility of day/night operation in all weather, so it is widely applied in military field. With the popularity of SAR equipment, SAR is gradually applied to civilian areas, such as target detection on the sea. Airborne SAR image can effectively maintain more features of the targets owing to the characteristic of high-resolution, while detection efficiency is affected by large amount of data caused by large number of bands in high-resolution images. The complexity of the ocean surface and the variable imaging parameters in addition to the differences of the targets which lead to the differences of the targets'scale all have a great impact on the precision of the target detection.Speckle noise derived from the SAR imaging will affect the accuracy of target detection, so several denoising methods are studied and compared in the paper. Since the target detection costs a lot of time in the original high-resolution image, multi-resolution processing is necessary in pre-processing. The experiment of pre-processing includes denoising and multi-processing. And the result shows that Gaussian pyramid processing and Frost filter can reduce the image resolution and maintain the features of the targets effectively.In the target detection process, at first, the constant false alarm rate (CFAR) algorithm is studied theretically. The experimental results demonstrate that K-distribution can fit the sea clutter of airborne SAR image well. Then, the characteristics of local CFAR and global CFAR are analyzed and it is concluded that both of them are based on the assumption that the targets are on the same scale, so they are not suitable for target detection when targets are in different scale and background is complex. In the end, a multi-resolution CFAR method for detecting target is proposed. In this method, target scale information is used to do multi-resolution processing. Local and global CFAR are used to detect different targets respectively and the final result are obtained by merging the detecting targets.5 airborne SAR images and 3 spaceborne (Radarsat-2)images are used to evaluate the performance of the proposed method. The results indicate that the FoM and the detection ratio of multi-resolution CFAR are 0.85 and 93.5% respectively. The experimental results demonstrate that multi-resolution CFAR could effectively detect target in images with complex background and different scale.