Double layer local contrast measure and multi-directional gradient comparison for small infrared target detection
Long Ren,Zhibin Pan,Yue Ni
DOI: https://doi.org/10.1016/j.ijleo.2022.168891
IF: 3.1
2022-05-01
Optik
Abstract:Infrared small target detection is one of the key technologies in the search and track (IRST) based on infrared imaging equipment. At present, the performance of small target detection based on single frame infrared image is directly related to the accuracy of subsequent target tracking, so it has been studied a lot. However, the existing small target detection algorithms have certain limitations in detection accuracy and real-time performance, especially when the contrast between the target and the background area is not high or the background is complex, especially in the complex sea or sky background, due to the influence of a large amount of noise and clutter in the background, the existing infrared small target detection algorithms have a high false alarm rate. To solve the above problems, this paper proposes a small target detection algorithm based on weighted double layer local contrast and multi-directional gradient map, which realizes the accurate detection of small targets from two aspects of targets’ local contrast and gradient. Firstly, we design an improved two layer local contrast measurement architecture, and use the weighted mean method to better represent the gray value of the local window; Secondly, a local contrast comparison method based on target and background is proposed to enhance the intensity of small targets and suppress some background clutter; Then, the multi-directional gradient map is used to further suppress the noise so as to improve the contrast between the target and the background. At the same time, singular value decomposition (SVD) method is used to extract the main features including small targets, which can effectively suppress the small texture interference around the targets in the background without losing the target intensity; Finally, an adaptive threshold method is used to separate small targets from their background. Experimental results show that compared with the existing algorithms, the proposed detection algorithm can effectively reduce the false alarm rate in different complex scenes, and the computational efficiency is improved compared with some multi-scale small target detection methods. At the same time, the signal to clutter ratio (SCR), background suppression factor (BSF) and receiver operating characteristic (ROC) curve are also better than these existing state of the art algorithms, which can display good robustness.
optics