Small Target Detection Method Based on Random Sampling and Sparse Matrix Restoration in Infrared Scenes

Abstract

The present invention proposes a small target detection method based on random sampling and sparse matrix recovery in infrared scenes. Randomly sample each pixel position in a single frame of infrared image to obtain an infrared image with random characteristics; perform patch transformation on the randomly sampled infrared image, and divide the randomly sampled infrared image into multiple small images without overlapping areas , and perform one-dimensional vectorization processing to obtain a two-dimensional matrix after patch transformation; perform principal component analysis on the two-dimensional matrix after patch transformation to obtain a sparse matrix and a low-rank matrix; use the patch inverse transformation method for the step sparse matrix Image recovery, respectively obtain the corresponding infrared small target image and infrared image background; use a low-rank matrix to determine the segmentation threshold in infrared small target detection, perform image segmentation on the infrared small target image according to the segmentation threshold, and detect the infrared small target. The method of the invention is simple and has short calculation time.

Description

Technical field [0001] The invention belongs to the technical field of image detection and processing, and specifically relates to a small target detection method based on random sampling and sparse matrix restoration in an infrared scene. Background technique [0002] Small target detection technologies in infrared scenes are mainly divided into two categories. One is to analyze the imaging characteristics of small infrared targets in a single frame image, and the other is to use a set of infrared image sequences according to the movement of small infrared targets. Analyze with the way of track correlation. At present, the detection technology of small targets in single-frame infrared images mainly focuses on the study of the imaging characteristics of the target, and marks the targets in the infrared image that meet the characteristics according to the imaging characteristics, and finally obtains the infrared target. There is less analysis of low-rank characteristics combined ...

AI Technical Summary

Problems solved by technology

There is less analysis of the low-rank characteristics combined with the infrared background, especially in the previous sparse matrix recovery method, the interception of small images requires overlapping areas to ensure strong correlation in the algorithm, but it will increase during the target detection process The amount of calculation of the algorithm, and the influence of noise interference cannot be well overcome in the detection, the work of preprocessing the infrared image is required, and there is a large time consumption. In addition, for small infrared targets using the track correlation method In the detection method, the main idea is single-frame detection, and the single-frame target is screened through the method of track correlation between frames, so as to obtain the detection result of small targets, which can improve the accuracy of the algorithm to a certain extent, but there are algorithms The calculation process takes a long time and the system complexity is high

Images