Research On Target Algorithm Based On SAR Image

Synthetic Aperture Radar(SAR)is a high-resolution radar using microwave imaging.With its all-day and all-weather capability,SAR is widely used in national defense military and civilian fields.Among them,SAR image target detection technology occupies a very important position in the field of modern military target attack,combat environment reconnaissance and surveillance,and becomes the focus of scholars in various countries.Because the SAR image is obtained by the speckle imaging mechanism,the speckle noise exists in the SAR image.The strength of the noise will affect the research of the SAR target detection technology to different degrees.Therefore,how to effectively suppress the speckle noise is an indispensable part of the research field of the SAR image.In addition,in the SAR image target detection technology,how to effectively reduce interference,obtain interesting targets from complex SAR images,has great research value.This paper mainly focuses on three aspects of SAR image speckle suppression,target detection and algorithm implementation,and analyzes and studies existing problems.The research results of this paper are as follows:Firstly,aiming at the problem that the existing speckle suppression algorithm loses edge detail information while denoising,this paper proposes an anisotropic diffusion algorithm based on weighted Manhattan distance based on the introduction of Euclidean distance anisotropy diffusion,using weighted Manhattan distance.Effectively distinguishes between the image's homogenous and heterogeneous regions,Using the improved diffusion function to control the diffusion speed,the performance of noise suppression and image edge detail protection is improved.Secondly,the target detection can not adapt to the complex background clutter in the global CFAR,and the local dual parameter CFAR detection algorithm uses local information to lead to the error of the detection results.On the basis of the non-local mean dual-parameter CFAR algorithm,an improvement was made,Using target pre-detection to reduce the interference of target to background clutter statistics,and two sliding windows were used to speed up the algorithm processing efficiency.At the same time,the ratio of distance and slice distance are used to measure the degree of difference in SAR image slices,and the background clutter is reorganized to make the estimated value of the background parameter more accurate.The experimental results show that the improved algorithm has a higher detection rate and a lower false alarm rate.Thirdly,when there is speckle noise and target shadow interference in the SAR original image,the significant region obtained by the saliency detection algorithm is inaccurate,and an improved LC-based saliency detection algorithm is proposed to calculate the saliency value.When the effect of the original pixel values is introduced,the target shadow interference is removed,and the influence of the noise is weakened by using the local similarity method.Finally,the performance of the proposed algorithm is verified by experiments.Fourthly,aiming at the problem of poor real-time performance in actual engineering application of SAR targets detection,a set of radar target rapid recognition software based on dual-chip DSP is designed.The algorithm of combining coarse matching and fine matching is proposed to accelerate the processing speed of the algorithm.At the same time,a set of dual-chip DSP multi-core parallel processing framework is designed,which has strong application value and leading significance.Finally,the real-time performance of the software was verified through experiments.