基于多传感器信息融合的智能车辆局部环境识别

Local Environment Recognition Based on the Data Fusion of LMS and Vision

为了对自主车辆在实时行驶中的驾驶避障给予指示并进行路径规划,提出并验证了一种基于激光雷达深度信息以及图像纹理信息的局部环境判别算法;根据车辆的通过性指标对车辆进行动力性建模,通过改进的Dempster-Shafer判据对识别结果进行融合;并且输出固定监视区域内的通过性指数矩阵以指引自主车辆实时修正导向。

In order to guide the UGV deteetion, a new method of local environment recognition based on the data fusion of LMS and vision were given and tested. Firstly the structure of LMS system and vision system has been modeled separately, and then the algorithm was showed to describe the data processing procedure. The grid method has been used to classify obstacle points and evaluate their cost. Considering the motor vehicle trafficability, an upgraded Dempster-Shafer criterion was used to revise the decision-making of driving. The trafficability was evaluated by reliability matrix. At last, the experiment has been done to validate the whole system.