Infrared small target detection method based on multi-angle prior analysis

Abstract

The invention discloses an infrared small target detection method based on multi-angle prior analysis, and the method comprises the specific steps: 1, carrying out the preprocessing of an original image, obtaining a gray image, and constructing a fragmentation tensor model of the original image; 2, filtering the grayscale image through a corresponding directional derivative model, and constructing a corresponding fragment tensor; step 3, adding a local structure weight to the grayscale image; 4, separating a target from a background; and step 5, picture reconstruction and target detection. According to the method, multi-direction prior analysis is provided, the target is enhanced in a re-weighting mode, and the target in the infrared image is accurately detected in a severe environment lacking texture information and space structure information, so that the false alarm rate is reduced. According to the method, in the small target detection process, the balance relation between the suppression background and the enhancement target is emphasized, that is, the information of the target edge is reserved while the background edge is obtained and suppressed.

Description

technical field [0001] The invention relates to the technical field of infrared small target detection, in particular to an infrared small target detection method based on multi-angle prior analysis. Background technique [0002] Due to its long imaging distance, excellent anti-interference performance, and the advantages of obtaining better images at any time of the day, infrared imaging has been used as an important means of photoelectric detection in space surveillance, remote sensing, and medical imaging. , target detection and tracking have a wide range of applications. But the performance of these applications depends heavily on robust object detection. However, because the infrared imaging system is far away from the target, the target is often presented as a point in the infrared image, and its distance information cannot be displayed. On the one hand, the target in the infrared image is very small, and cannot collect specific shapes and textures, and cannot utiliz...

AI Technical Summary

Problems solved by technology

In the IPT model for single-frame detection proposed by Dai et al., they use the two highest eigenvalues ​​in the structure tensor to describe the local geometry, while λ 1 -λ 2 When determining the edge, it is impossible to distinguish whether the edge belongs to the object or the background

This ambiguity directly leads to the distortion of object shape in small object detection results

In addition, only background-related priors are considered in the IPT model, and target-related priors are not considered

Images