Multi-modal space-time fusion target detection method and device and medium

Through the multimodal spatiotemporal fusion object detection method, combined with simulation and real data, a 3D and 2D object detection network with spatiotemporal feature fusion capability is established, which solves the problem of object detection error detection and follow-up loss in complex environments and improves the accuracy of detection and applicability reduces human resource consumption and avoids model overfitting.

CN119107638AActive Publication Date: 2024-12-10BEIJING UNIV OF POSTS & TELECOMM
View PDF 6 Cites 4 Cited by

Patent Information

Application Number
CN202411289279.X
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-09-14
Publication Date
2024-12-10
Estimated Expiration
2044-09-14

AI Technical Summary

Technical Problem

The existing target detection methods have high risk of error detection and follow-up loss in complex environments, and are difficult to achieve accurate and rapid identification and stable tracking under extreme weather conditions and high-speed dynamic target scenarios. It is difficult to collect data and consume a lot of human resources, so the model is easy Overfitting occurs.

Method used

Using multimodal spatiotemporal fusion object detection method, by constructing a data set containing simulated and real target data, a 3D point cloud target recognition network and 2D object detection model network with spatiotemporal feature fusion capability are established, and through weight transfer learning and multimodal The state decision-making layer information is integrated to build a target detection system suitable for complex environments.

Benefits of technology

It effectively reduces the risk of target misdetection and loss in complex environments, improves the applicability and accuracy of target detection, reduces human resource consumption, and avoids model overfitting.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119107638A_ABST
    Figure CN119107638A_ABST
Patent Text Reader

Abstract

The invention relates to a multi-modal space-time fusion target detection method and device and a medium, and the method comprises the steps: constructing a to-be-detected target data set which at least comprises simulation target point cloud data, real target point cloud data and real image data; based on the original input point cloud, obtaining a 3D point cloud target identification network with spatial-temporal feature fusion capability; based on the simulation target point cloud data and the real target point cloud data, carrying out weight transfer learning training on the 3D point cloud identification network; constructing a 2D target detection model network with a voting mechanism, and training the 2D target detection model network; and based on the 2D target detection model network and the 3D point cloud target identification network, constructing a target detection network of multi-modal decision-making layer information fusion. The invention provides a target detection scheme which is more suitable for a complex environment, and the accuracy of target detection and recognition in the complex environment is improved.
Need to check novelty before this filing date? Find Prior Art

Citation Information

Patent Citations

  • Three-dimensional target detection method and system based on multi-modal fusion

    CN115937819A

  • Three-dimensional target detection method based on multi-modal fusion and deformable attention

    CN117975436A

  • 3D target detection method and system based on multi-modal fusion

    CN118429954A

  • RepSurf feature and information interaction-based partially overlapped point cloud registration method

    CN118447060A

  • Transformer training method for paint surface AI defect identification

    CN121504856A

Cited By

  • Lightweight aviation multispectral target detection method

    CN120032196A

  • Unmanned aerial vehicle surveying and mapping information processing method and system of prefabricated model

    CN120352886A

  • Unmanned aerial vehicle detection method and system based on improved YOLOv11n

    CN120708100A

  • Target detection method and device, electronic equipment, storage medium and program product

    CN121190937A