Multi-sensor fusion method based on DS-GNN algorithm

Abstract

The invention discloses a multi-sensor fusion method based on a DS-GNN algorithm, and the method comprises the following steps that firstly the radar resolves multiple possible targets through its ownsignal processing to obtain their positions and speed information, the camera module provides the positions and speed information of the targets, joint calibration of the millimeter-wave radar and camera is required before the system detection to realize the fusion of multiple sensors in time and space, a global nearest neighbor (GNN) algorithm is introduced to match all the original target information and observed values one by one to confirm which observed values of the radar and camera are generated by the same vehicle, and finally a D-S evidence theory is used to effectively fuse the observed values, including target state update and existence probability update, wherein the state update process is basically consistent with standard Kalman filtering, and the existence probability is tracked simultaneously. The multi-sensor fusion method based on the DS-GNN algorithm provided by the invention can realize the accurate identification and tracking of target vehicles.

Description

technical field [0001] The invention relates to the technical field of vehicle intelligent driving and active safety, in particular to a multi-sensor fusion method based on a DS-GNN algorithm. Background technique [0002] For advanced driver assistance systems (ADAS) and autonomous driving, a single sensor is increasingly unable to meet people's safety requirements for increasingly complex traffic environments. However, both the radar and the camera have their own defects, and they will be affected by various adverse environmental factors, which will greatly weaken the ability of target recognition in specific scenes. In the vehicle collision avoidance method and system, due to the restriction of the perception ability of a single sensor, there are certain defects, and the vehicle collision cannot be effectively avoided in the actual full working condition scene. Therefore, the research on target tracking method based on multi-sensor fusion has great theoretical significan...

AI Technical Summary

Problems solved by technology

Although this type of algorithm can speed up the image recognition process, it cannot effectively take advantage of the sensor's recognition technology advantages, and the system robustness is not strong

Images