Infrared dim and small target detection based on spatiotemporal cascade low-rank sparse inversion

Abstract

In the field of small target detection in the space domain, the traditional tensor model has insufficient constraints on noise and target features in the strong clutter environment, which leads to the residual problem of edge noise in the detection results. Firstly, a low-rank sparse inversion model with multi-dimensional spatiotemporal cascade jumps was constructed, and the accurate target extraction and preliminary filtering of background clutter in the background of strong clutter were realized by optimizing the low-rank sparse representation ability of the image matrix. Secondly, in order to solve the problem of aliasing of signals and noise caused by the uncertainty of the target trajectory in complex dynamic scenes, an energy enhancement model based on multi-directional motion estimation was designed. By establishing a hypothetical framework of nine-direction motion and combining it with the bilateral filtering weighting mechanism, the directional focusing enhancement of the target energy and the further suppression of the noise component are realized in the spatiotemporal dimension, and the target detection output with high SNR is finally completed. In the experimental part, the experimental part is compared with 11 small target detection methods in 8 typical scenes, and the results show that the proposed algorithm can effectively cope with the weak and small target detection tasks in complex scenes.