A traffic target detection method and system based on improved RT-DETR

By improving the structure of the RT-DETR model, embedding a dynamic perception enhancement module and a dual-scale collaborative attention module, and replacing them with a global contextual convolution module, the problem of multi-scale target detection in complex traffic scenarios of the RT-DETR model is solved, and the detection accuracy and robustness are improved.

CN122336263APending Publication Date: 2026-07-03NANCHANG UNIV
View PDF 0 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
NANCHANG UNIV
Filing Date
2026-06-01
Publication Date
2026-07-03

AI Technical Summary

Technical Problem

Existing RT-DETR models struggle to effectively detect multi-scale targets, especially small and occluded targets, in complex and dynamic traffic scenarios, and lack robustness under the influence of lighting changes and weather factors.

Method used

By embedding a dynamic perception enhancement module in the backbone network, replacing the adaptive feature interaction module with a dual-scale collaborative attention module, and replacing some standard convolutional blocks with global context convolutional modules, the model's multi-scale feature representation, robustness, and local semantic understanding capabilities are improved.

Benefits of technology

While maintaining real-time performance, it significantly improves the detection accuracy and robustness for multi-scale targets, especially the detection performance for small targets and under occlusion conditions.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122336263A_ABST
    Figure CN122336263A_ABST
Patent Text Reader

Abstract

This invention discloses a traffic target detection method, device, and medium based on an improved RT-DETR model, belonging to the field of autonomous driving and assisted driving technology. The method includes: acquiring initial driving video images and dividing them into a training image set and a test image set; structurally adjusting the initial RT-DETR model to construct an improved RT-DETR model; the structural adjustment includes embedding a Dynamic Perception Enhancement (DPB) module in the backbone network to enhance multi-scale extraction capabilities, replacing the adaptive feature interaction module with a dual-scale collaborative attention module (DSCAB) to achieve channel and spatial attention collaboration, and replacing some standard convolutional blocks with a global contextual convolutional module (GCC3) to achieve dynamic feature enhancement; iteratively optimizing and training the improved RT-DETR model using the training image set; and inputting the test image set into the model to output detection results. This invention solves the problem of low detection accuracy for multi-scale, small, and occluded targets in complex traffic scenarios, significantly improving the accuracy and robustness of target detection.
Need to check novelty before this filing date? Find Prior Art