Multi-view 3D target detection method and system for long-distance target

The multi-view image input backbone network and the remote detection denoising module generate false queries, which solves the inaccuracy and instability of long-distance object detection and achieves more efficient long-distance object detection.

CN120472443APending Publication Date: 2025-08-12YUNNAN UNIV
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202510379412.9
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2025-03-28
Publication Date
2025-08-12

AI Technical Summary

Technical Problem

The existing multi-view 3D object detection methods have inaccuracy and instability in long-distance object detection, especially in complex environments, where long-distance objects cannot be effectively detected, and high computing resource requirements or insufficient ability to capture dynamic environmental changes.

Method used

The multi-view image input backbone network is used to extract image features, and the features are enhanced by depth band convolution and attention weight, the context information between long-distance pixels is captured, and false queries are generated through the remote detection denoising module to improve the accuracy of long-distance object detection.

Benefits of technology

It improves the accuracy and stability of long-distance object detection, enhances the ability to capture long-distance targets, reduces the computational complexity, and adapts to complex environment changes.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN120472443A_ABST
    Figure CN120472443A_ABST
Patent Text Reader

Abstract

The invention discloses a multi-view 3D target detection method and system for a long-distance target, and belongs to the field of computer vision and image data processing. The method comprises the following steps: inputting a multi-view image into a backbone network to extract image features; enhancing the features of the central region according to the image features, and capturing context information between long-distance pixels; generating a false query of a long-distance target through a remote detection denoising module; and decoding the false query of the long-distance target and the characteristics of the central area, and carrying out loss calculation through a detection head to complete detection. The system comprises a feature extraction module, a long-distance feature module, a remote detection denoising module and a decoder module. According to the method, the accuracy of long-distance object detection is improved by generating three-dimensional query by utilizing false depth information, the stability is better in a complex real world environment, and the method has better target detection accuracy.
Need to check novelty before this filing date? Find Prior Art