摘要
针对传统车位检测算法因难以捕捉距离跳变和确定车位姿态而导致检测准确性不足的问题,提出了一种基于拟合特征线段的改进车位检测算法。针对超声波雷达幅值衰减设计了能量增益补偿方案;根据车辆实时位姿和超声波距离数据拟合代表障碍物轮廓的特征线段,借助激光雷达对特征线段拟合精度进行评价,并进一步设计了融合车位初步检测、车位再判断、车位约束条件判断及车位姿态计算的检测流程。利用实车平台对所提算法进行了实验验证,实验结果表明,所提算法的平均检测准确率达92.5%,能为智能汽车泊车系统提供准确、可靠的目标车位信息。
Aiming at the problems of insufficient detection accuracy caused by the difficulty of capturing distance jumps and determining the attitude of parking spaces in traditional parking space detection algorithms,an improved parking space detection algorithm was proposed based on fitted feature line segments.First,an energy gain compensation scheme was designed for the ultrasonic radar amplitude attenuation;then,the characteristic line segment representing the outline of the obstacles was fitted based on the real-time vehicle poses and ultrasonic distance data,and the fitting accuracy of the characteristic line segments was evaluated with the help of lidar,and further detection processes of fusion parking space preliminary detection,parking space re-judgment,parking space constraint judgment and parking space attitude calculation were designed.Finally,the proposed algorithm was experimentally verified based on the real vehicle platforms.The experimental results show that the proposed algorithm may obtain an average detection accuracy of 92.5%,and provide accurate and reliable target parking space informations for the intelligent car parking systems.
作者
仇旭
朱浩
邓元望
QIU Xu;ZHU Hao;DENG Yuanwang(College of Mechanical and Vehicle Engineering,Hunan University,Changsha,410082)
出处
《中国机械工程》
EI
CAS
CSCD
北大核心
2020年第14期1747-1753,共7页
China Mechanical Engineering
基金
湖南省重点研发计划资助项目(2017GK2201)。
关键词
超声波雷达
车位检测
最小值滤波
特征线段
ultrasonic radar
parking space detection
minimum filtering
characteristic line segment