Research On Infrared Weak And Small Target Detection Algorithm Based On Image Block Low-rank Reconstruction Theory

The rapid development of science and technology has greatly promoted the continuous update of modern military application strategies,and has put forward high requirements for the combat distance and detection accuracy of the infrared target detection and tracking system.The real infrared image has a low signal-to-noise ratio and serious background interference,while the target usually appears as a faint isolated light spot.Therefore,effectively highlighting the target and suppressing various types of complex background is the key to infrared small target detection technology.In recent years,the advancement of infrared small target detection algorithms has achieved more efficient detection results,but it still shows certain limitations in the face of stubborn residual background.Based on the study and analysis of previous research results,this paper proposes three more effective comprehensive detection models for the above problems.The main tasks are as follows:1.Aiming at the phenomenon that the infrared block image model(IPI)retains sparse and cluttered edges when processing complex backgrounds,a small infrared target detection method based on the effective rank of the image is proposed.The method includes two stages:preprocessing and target extraction.In the preprocessing stage,in order to strengthen the non-local autocorrelation between background pixels,the original infrared image is rearranged into an operation matrix using the local image block structure,and then the inexact augmented Lagrange multiplier(IALM)is used to separate the background and obtain the target foreground image containing edge residuals.In the target extraction stage,in order to further distinguish the target from the residual background,an effective-rank hierarchical model is established to process the target foreground image.First,perform singular value decomposition on the target foreground matrix;then apply the maximum attenuation of the singular value to determine the effective rank of the matrix to restore the hierarchies of interest;in the next step,use the minimum information entropy of the image to filter out the image hierarchy where the target is located;finally,use Threshold segmentation to optimize detection results.Experimental results show that this model can effectively distinguish real targets and clutter interference under different backgrounds.2.In order to solve the limitations of image effective-rank hierarchical model facing multi-target detection scenes,a small infrared target detection method based on local contrast and block low-rank reconstruction is proposed,which mainly includes the following steps:First,the local information of the image is used to construct the local contrast measurement sub-model of the inner and outer windows,and the saliency map of the original image is obtained by calculating the contrast between the test pixel and the surrounding background;Next,use the non-local information of the image to introduce a block low-rank reconstruction sub-model to reconstruct the original image,and then use the accelerated proximal gradient method(APG)to obtain the target foreground image.Finally,the image dot multiplication operation is used to synthesize the processing results of the sub-model,and the hierarchy of interest is restored through effective singular values to obtain the final experimental results.Experimental results show that this method expands the detection range of the image effective-rank hierarchical model from a single target to a multi-target scene,and shows a higher detection performance in the comprehensive evaluation with the contrast algorithm.3.Aiming at the missed detection phenomenon in the detection model of local contrast and block low-rank reconstruction,a dual-path infrared small target detection algorithm based on Hough line suppression is proposed based on the comprehensive consideration of the difference between the target and the background residual morphological characteristics.First,use the IPI model to obtain the target foreground map.Then,a Hough line suppression model is established to convert the global edge suppression problem into a local peak detection problem in the Hough space,so as to specifically suppress the residual background in the target foreground image.Then,a dual-path processing strategy is designed according to whether there is a Hough line in the foreground image of the target as a judgment basis to ensure the target detection performance under different complex backgrounds.Finally,the image effective-rank hierarchical model is used to distinguish the target from the residual pixel-level highlight noise to obtain the final detection result.A large number of experiments conducted in various complex scenes prove that this method is robust and has outstanding background suppression effect,and it is a detection method with superior performance..