The Research And Application Of PCNN On Automatic Target Recognition Based On SAR

Synthetic Aperture Radar (SAR) is widely used in environmental monitoring, earth-resource detecting, and military supervising. SAR images contain more information than optical image for it's penetrability. We can get knowledge of relative areas which covered by SAR by SAR images interpretation. SAR target recognition is an important process in SAR images interpretation and analysis. The process of recognition contains detection, discrimination, and recognition. Pulse Coupled Neural Network (PCNN) is a self organized neural network which don't need supervised. It has features of scale-invariant and rotation invariant. PCNN is easily realized with hardware and it's real time features is better than other artificial neural networks.Research on target recognition based on SAR platform is very hot in domestic and international. Local statistics filter method, structural filter method, and wavelet filter method is used in speckle suppression. Constant False Alarm Rate (CFAR) method is mainly used in SAR target detection. Target and non-target is classified in SAR discrimination mainly by genetic algorithms. At classification and recognition phase, Markov random field(MRF)segmentation, Gabor wavelet transform feature extraction, and wavelet method which base on edge detection is used.This paper research on SAR de-noises and target recognition with PCNN.Results were reached in this paper about adoption of PCNN in the de-noise and recognition of SAR image through tests. In the experiments,there are lots of datasets is needed, so a simulation of SAR image is developed. The SAR simulator is base on calculation and raster ability of GPU to generate image of a given scene. SAR image of various stance of target is produced by this simulator is used in all experiments.