Infrared weak and small target detection method based on multi-frame motion feature accumulation learning

By employing a multi-frame motion feature accumulation learning method, utilizing the Scharr operator to enhance edge information, and combining motion learning and accumulation learning modules to extract target motion and spatiotemporal information, the adaptability and noise interference issues of infrared weak target detection in complex backgrounds are solved, achieving efficient feature extraction and fusion and improving detection accuracy.

CN121033095APending Publication Date: 2025-11-28CHENGDU CHONGHU INFORMATION TECH CO LTD
0 Cites 0 Cited by

Patent Information

Application Number
CN202511011188.4
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2025-07-22
Publication Date
2025-11-28

AI Technical Summary

Technical Problem

Existing infrared weak target detection methods have poor adaptability in complex backgrounds, making it difficult to effectively capture the features of targets with irregular shapes and size changes. Furthermore, single-frame detection is easily affected by noise, and the temporal features are not fully utilized in multi-frame detection.

Method used

A method based on multi-frame motion feature accumulation learning is adopted. The image edge information is enhanced by the Scharr operator. The motion learning module and the accumulation learning module are used to extract the target motion information and spatiotemporal information respectively. The feature extraction and fusion are combined with the final fusion module, and the model is optimized using the Soft-IoU loss function.

Benefits of technology

It improves the accuracy and robustness of infrared weak target detection, effectively detects targets with irregular shapes and size changes in complex scenes, reduces noise interference, and enhances the effect of multi-frame detection.

✦ Generated by Eureka AI based on patent content.
Patent Text Reader

Abstract

The invention relates to the technical field of infrared dim and small target detection, and provides an infrared dim and small target detection method based on multi-frame motion feature accumulation learning, which comprises the following steps of: 1, performing feature enhancement on an original image, namely obtaining edge information of the original image by using a Scharr operator, inputting the original image sequence and the corresponding edge image sequence into the model; 2, the model detects motion information in the image sequence through a motion learning module; 3, the model learns accumulated time sequence information and space information of a multi-frame sequence through an accumulated learning module, and a memory bank is established; and 4, inputting the output of the motion learning module, the output of the cumulative learning module and the original image of the key frame into a final fusion module for final feature extraction and fusion, then carrying out channel compression on an output feature map to obtain a prediction feature map, and carrying out final prediction. According to the invention, multi-frame infrared weak and small target detection can be well carried out.
Need to check novelty before this filing date? Find Prior Art