Weak and small target detection method based on probabilistic pipeline filtering

Abstract

The invention discloses a weak and small target detection method based on probabilistic pipeline filtering. The problem that an existing small target detection technology is low in precision and robustness is mainly solved. According to the implementation scheme, the method comprises the following steps: firstly, carrying out background suppression preprocessing on an input sequence image, and obtaining a binary image by using a maximum between-cluster variance method; then initializing pipeline parameters, and establishing a filtering pipeline by utilizing the existing N frames of binary images; and then probabilistic pipeline filtering is used to detect the preprocessed image, i.e., noisy points which are easy to distinguish are eliminated through spatial constraint, the probability thatcandidate targets which are difficult to distinguish belong to real targets is calculated by using gray scale distribution, regional characteristics and the like, and the candidate target with the maximum probability is output as the real target. Compared with traditional pipeline filtering, the method effectively solves the problem of interference of target-like noise on a real target, improvesthe precision and robustness of target detection, reduces the omission ratio and false alarm rate, and can be used for detecting and tracking a weak and small target under a complex background.

Description

Technical field [0001] The invention belongs to the technical field of pattern recognition, and particularly relates to a target detection method, which can be used for the detection and tracking of weak and small targets in a complex background. Background technique [0002] The detection of weak and small targets is one of the hot issues in the field of photoelectric detection and tracking. Due to the imaging principle and imaging distance, the energy of the target is much smaller than that of the background. At the same time, due to the relatively open field of view, a lot of noise interference will be introduced. Because the background is complex, the target is weak and small, and the target is often submerged in noise, this makes the image contrast low and the signal-to-noise ratio is high. At the same time, because the target pixel ratio is small, the target lacks texture and other pixel features, which can be used for target detection. less. [0003] At present, most of the...

AI Technical Summary

Problems solved by technology

Pipeline filtering is a commonly used multi-frame detection algorithm. The algorithm published by Liu Jin et al. in the Journal of Xidian University in 2007 "Infrared Weak and Small Target Detection Based on Mobile Weighted Pipeline Filtering" introduces weighted when modifying the coordinate position of the pipeline center Pipe center coordinate displacement, this method can reduce the interference of pipe edge noise on detection, but does not consider the correlation of multi-frame targets in terms of size, etc., and still keeps the pipe diameter unchanged, which often leads to detection failure in practical applications

Xu Zhiyong and others proposed a pipe diameter adaptive pipe filtering algorithm in the patent (patent number: CN106469313B) "A Weak Target Detection Method for Pipe Diameter Adaptive Time-Space Domain Filtering". This method modifies the pipe diameter according to the change of the target scale. However, due to the failure to make full use of the consistency constraints of multi-frame target features, the algorithm often cannot effectively overcome strong noise interference, which makes errors in the tracking of the pipe diameter, which in turn affects the robustness and accuracy of detection.

[0004] These methods only improve the method of updating the pipeline center and diameter from the perspective of pipeline design, and do not have sufficient correlation and continuity between multi-frame targets and their features The accuracy of target detection is low, and the robustness is poor, which cannot better meet the actual needs of current projects.

Images