Research On Weak Target Detection In SAR Image Under Complex Scene

In recent years,the technology of weak target detection in synthetic aperture radar(SAR)has been paid more and more attention in more and more fields such as satellite remote sensing,missile early warning and guidance,search and tracking,astronomical observation,forest warning,medicine,biology,intelligent control and so on.For weak target detection has become an important research direction,the study of weak target detection technology has great application value.Aiming at the accurate extraction of vehicle targets in SAR images with low SNR,the target detection algorithm and the target discrimination algorithm are deeply studied in this paper.The main work of this paper includes the following:Firstly,the related techniques of weak target detection are summarized,including image speckle suppression algorithms,the Coherence-CFAR(CCFAR)algorithm and the target detection algorithm based on the Non-Subsampled Contourlet Transform(NSCT)domain.The causes of weak target are analyzed deeply.Due to the large radar incident angle,target self-scattering characteristics,ghosting by natural environment occlusion in the background,camouflage and so on,the Radar Cross Section(RCS)of the target is often low in the process of SAR imaging which results in the low gray value of the target area in SAR image and the small difference between the target and the background clutter.So the traditional target detection algorithm may not be detected accurately and effectively.Then,after analyzing the defects of the traditional visual significance models applying to the weak target detection,a weak target detection algorithm based on visual significance is proposed combined with the unique characteristics of SAR image.Primary feature saliency maps are extracted from the original scale of the SAR image.In order to suppress the clutter noise,the feature-significant maps are nonlinearly merged to generate a general saliency map.Finally,the regional growth technique and the mechanism of “winner take all”(WTA)are used to extract the target area from the significant map.Finally,through the comparison experiment,the algorithm has pretty characteristics of the detection rate and operation efficiency,which can be applied to the detection in complex scenes.In the aspect of weak target feature discrimination,this paper first segments the regions of Interest(ROI)accurately based on the change detection statistics and the KSW threshold segmentation.Then this paper calculates the geometric and contrast characteristics of the target according to the segmentation result.The target discrimination algorithm is verified experimentally.Finally,a corner-based target azimuth estimation algorithm is proposed.The algorithm first separates the target from ROI accurately,and extracts the corners based on Harris operator.Because of the obvious geometric distortion in the process of target imaging,based on the symmetry of the structure of the target itself,the algorithm uses the corner points and the target center of gravity symmetry points for the regression analysis to estimate the target azimuth.The experimental results show that the azimuth estimation of the algorithm has high accuracy.