Traffic event detection method based on multi-modal edge fusion
By using multimodal edge fusion technology, a one-dimensional variational function model is constructed using visual and radar data. Kriging interpolation is then used to calculate the vehicle positions in congested queues. This solves the problem of perception blind spots caused by visual occlusion and radar multipath interference in dense vehicle queues, and achieves high-precision traffic event detection.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- ZHONGNAN TRANSPORT
- Filing Date
- 2026-06-16
- Publication Date
- 2026-07-17
AI Technical Summary
Existing traffic incident detection technologies are unable to accurately estimate the position of vehicles at the back of a congested queue due to visual obstruction and radar multipath interference in dense vehicle queues, and are prone to false alarms or missed alarms.
A multimodal edge fusion method is adopted to extract the accurate longitudinal distance value in front of the radar by visually identifying dense vehicle queue areas, constructing a one-dimensional variation function model, using Kriging interpolation to calculate the estimated position and variance of the vehicles at the rear of the queue, and setting a safety threshold to judge the reliability of the output results.
Under conditions of visual obstruction and radar multipath interference, this method accurately estimates the position of vehicles in congestion queues, reduces false alarm and false negative rates, and improves the reliability and accuracy of traffic incident detection.
Smart Images
Figure CN122416746A_ABST