摘要
针对无人驾驶方程式赛车中点云目标检测精度不高、存在误检漏检的现象,提出一种基于空间距离特征约束的点云三维目标检测算法。该算法通过引入空间距离特征约束自适应调整聚类阈值,改善聚类效果,同时增加目标判定准则对检测结果进一步判别,提升目标检测精度。通过无人驾驶方程式赛车实车测试实验,该算法在直线赛道场景和弯道场景下的检测精度优于现存同类算法,与欧式聚类目标检测算法相比,在同等运行时间条件下,召回率提高32.7%,适用于中国大学生无人驾驶方程式大赛和其他自动驾驶领域。
In the autonomous racing car,the low accuracy of point cloud object detection results in missed detection and detection failure.Aiming at the problems,this paper proposed a point cloud 3D object detection algorithm based on the spatial distance feature constraints.The algorithm adaptively adjusted the clustering threshold through spatial distance feature constraints to improve the clustering effect.In addition,the object determination criterion was added to further determine the detection results,which raised the accuracy of object detection.After the autonomous formula-racing test,the experiment results for straight track scene and curve track scene showed that our algorithm was superior to similar existing methods in terms of detection accuracy.Compared to the Euclidean clustering object detection algorithm,the recall increased by 32.7%in the same running time,which was suitable for Formula Student Autonomous China(FSAC)and autonomous driving field.
作者
李云鹏
毛琳
杨大伟
刘长宏
LI Yun-peng;MAO Lin;YANG Da-wei;LIU Chang-hong(School of Electromechanical Engineering, Dalian Minzu University, Dalian Liaoning 116605, China)
出处
《大连民族大学学报》
2022年第1期30-34,68,共6页
Journal of Dalian Minzu University
基金
国家自然科学基金项目(61673084)
辽宁省自然科学基金资助项目(20170540192,20180550866)。
关键词
点云目标检测
空间距离特征
无人驾驶方程式大赛
point cloud object detection
spatial distance feature
Formula Student Autonomous China(FSAC)