汽车工程 ›› 2025, Vol. 47 ›› Issue (10): 1923-1932.doi: 10.19562/j.chinasae.qcgc.2025.10.008

• • 上一篇    

共视场车载激光雷达下线标定设计与实现

付伟龙,王晶(),雷正潮,赵峻,谢欢   

  1. 陕西汽车控股集团有限公司,西安 710299
  • 收稿日期:2025-02-25 修回日期:2025-04-09 出版日期:2025-10-25 发布日期:2025-10-20
  • 通讯作者: 王晶 E-mail:10204547@sxqc.com
  • 基金资助:
    广东省重点领域研发计划项目(2020B090921001);陕西省2021年重点研发计划项目(2021LLRH-04-01-02)

Design and Implementation of Offline Calibration for Vehicle-Mounted LiDAR with Common Field of View

Weilong Fu,Jing Wang(),Zhengchao Lei,Jun Zhao,Huan Xie   

  1. Shaanxi Automobile Group Co. ,Ltd. ,Xi'an 710299
  • Received:2025-02-25 Revised:2025-04-09 Online:2025-10-25 Published:2025-10-20
  • Contact: Jing Wang E-mail:10204547@sxqc.com

摘要:

智能驾驶汽车量产过程中激光雷达的装配存在不可避免的误差,为了确保激光雷达能够提供准确可靠的感知信息,本文提出了一种针对智能驾驶汽车激光雷达下线标定方法及工位建设方案。构建标定工位,包含车辆停放区与具备空间特性标靶的标靶放置区;结合工位提取基准激光雷达地平面点云,与设计参考平面点云建立平面约束,同时利用基准激光雷达与标靶间的空间位置构建约束关系,完成基准激光雷达的标定;通过标靶检测建立基准激光雷达与左、右侧激光雷达间刚体变换标定模型,完成左、右侧激光雷达标定。建立联合仿真标定环境,设计开展仿真实验,验证了算法的有效性。通过实车实验对比分析,在稳定性与准确性方面,三轴旋转量、平移量的标定结果与手动标定相近。相比文献方案旋转量相对误差率明显降低,横滚量、俯仰量、偏航量较其分别降低约11.7%、3%、0.03%;标准差分别降低0.48、0.68、0.11。在标定效率方面,较手动标定提升约16倍,相比于文献方案提升约8倍。结果表明本文方案在精确性、稳定性和效率方面均展现出明显优势,为激光雷达下线标定提供了有力支撑与坚实保障。

关键词: 激光雷达, 下线标定, 共视场, 智能驾驶

Abstract:

During the mass production of intelligent driving vehicles, there are inevitable errors in the assembly of lidar. In order to ensure that the lidar can provide accurate and reliable perception information, in this paper a calibration method and workstation construction scheme for the off-line calibration of lidar in intelligent driving vehicles is proposed. A calibration workstation is constructed, which includes a vehicle parking area and a target placement area with a target having spatial characteristics. The ground plane point cloud of the reference lidar is extracted in combination with the workstation, and a planar constraint is established with the point cloud of the design reference plane. At the same time, the constraint relationship is constructed by using the spatial position between the reference lidar and the target to complete the calibration of the reference lidar. A rigid body transformation calibration model between the reference lidar and the left and right lidars is established through target detection to complete the calibration of the left and right lidars. oint simulation calibration environment is established, and simulation experiments are designed and carried out to verify the effectiveness of the algorithm. Through the comparative analysis of real vehicle tests, in terms of stability and accuracy, the calibration results of the three-axis rotation and translation are similar to those of manual calibration. Compared with the scheme in the literature, the relative error rate of the rotation is significantly reduced, with the roll, pitch, and yaw reduced by approximately 11.7%, 3%, and 0.03% respectively, and the standard deviation reduced by 0.48, 0.68, and 0.11 respectively. In terms of calibration efficiency, it is approximately 16 times higher than manual calibration and approximately 8 times higher than the scheme in the literature. The results show that the scheme proposed in this paper has obvious advantages in terms of accuracy, stability, and efficiency, providing strong support and a solid guarantee for the off-line calibration of lidar.

Key words: LiDAR, offline calibration, common field of view, intelligent driving