摘要
针对中国大学生无人驾驶方程式大赛(formula student autonomous China,FSAC)中的单一传感器检测环境适应性差等问题,提出一种基于激光雷达和相机融合的障碍物检测算法。对激光点云进行滤波、去地面和条件欧式聚类处理,确定锥桶位置;采用YOLOv7算法对图像进行检测并获取颜色信息;将传感器进行时空对齐,采用二次最近邻算法进行匹配,获取锥桶障碍物的位置和颜色信息。采用FSAC赛车作为实验平台,在动态测试中,该算法比取交集融合算法的准确率提升了5.52%,误差降低了29.47%,速度提升了21.66%,可以很好地满足检测的准确性和实时性,较好地实现了无人赛车的感知任务,同时为无人驾驶车辆的融合感知提供了一定参考依据。
In this paper,an obstacle detection algorithm based on the fusion of lidar and camera is proposed to address the poor adaptability of single sensor detection environment in the Formula Student Autonomous China(FSAC).Firstly,the laser point cloud is filtered,de-ground and conditional Euclidean clustering to determine the position of the cone barrel;secondly,the YOLOv7 algorithm is employed to detect the image and obtain the color information;finally,the sensor is spatiotemporal aligned,and the second nearest neighbor algorithm is used for matching to obtain the position and color information of cone barrel obstacles.Using FSAC car as the experimental platform,in the dynamic test,compared with the intersection fusion algorithm,the accuracy of the algorithm is improved by 5.52%,the error reduced by 29.47%,and the speed increased by 21.66%.The improved data well meet the accuracy and real-time requirements of detection,better realize the perception task of unmanned cars,and provide a reference for the fusion perception of unmanned vehicles.
作者
兰建平
郭文韬
杨亚会
张友兵
叶子恒
董冯雷
LAN Jianping;GUO Wentao;YANG Yahui;ZHANG Youbing;YE Ziheng;DONG Fenglei(Institute of Automotive Engineers,Hubei University of Automotive Technology,Shiyan 442002,China)
出处
《重庆理工大学学报（自然科学）》
北大核心
2023年第10期107-116,共10页
Journal of Chongqing University of Technology：Natural Science
基金
湖北省教育厅科学研究计划指导性项目(B2022397)
湖北省重点研发计划项目(2021BED004)
湖北汽车工业学院教学研究与改革项目创新创业教育专项(SCJY202314)。
作者简介
兰建平,男,硕士,讲师,主要从事汽车主动安全与智能驾驶研究,E-mail:20120006@huat.edu.cn;通信作者:郭文韬,男,硕士研究生,主要从事智能网联汽车多传感器方面研究,E-mail:2250663394@qq.com。
