Multi-target tracking method and system based on BYTE data association algorithm and medium

By adopting a multi-objective tracking method based on BYTE data association algorithm in the autonomous driving perception system, the frame drop problem and insufficient generalization ability caused by occlusion in the prior art are solved, and a more stable and safe obstacle tracking effect is achieved.

CN119722748APending Publication Date: 2025-03-28东风悦享科技有限公司 +1
0 Cites 0 Cited by

Patent Information

Application Number
CN202411786296.4
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-12-06
Publication Date
2025-03-28

AI Technical Summary

Technical Problem

The existing autonomous driving perception multi-target obstacle tracking algorithm can easily lead to frame loss problems under occlusion and other conditions, and has weak generalization capabilities, making it difficult to effectively track obstacles in different scenarios.

Method used

A multi-objective tracking method based on BYTE data association algorithm is adopted to collect obstacle detection data in real time, and use a dynamic threshold algorithm of gold sine factor to divide high-score and low-score obstacles, combine with the improved Kalman filtering algorithm to perform state prediction, and build a feature extraction model and Hungarian algorithm for optimal matching, improving the stability and generalization ability of the tracking algorithm.

Benefits of technology

It effectively solves the problem of frame loss caused by occlusion, improves the stable tracking ability of autonomous driving vehicles to obstacles, improves driving safety and smoothness, and enhances the generalization ability of the algorithm.

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

Abstract

The invention relates to a multi-target tracking method and system based on a BYTE data association algorithm and a medium, and the method comprises the steps: U1, collecting the detection data information of an obstacle in real time, dividing the detection data of the obstacle into a high-score obstacle and a low-score obstacle through a dynamic threshold algorithm based on a golden sine factor, state data information of the high-resolution obstacle is obtained; and U2, based on the state data information of the high-resolution obstacle, predicting the state of the obstacle by adopting an improved Kalman filtering algorithm to obtain the predicted state data information of the obstacle. According to the method, the stability of perceiving the output obstacle is improved, the driving safety and smoothness of the automatic driving vehicle are further improved, and the generalization ability of the tracking algorithm is improved based on the obstacle convex hull through the continuous frame matching strategy.
Need to check novelty before this filing date? Find Prior Art