自动驾驶多目标追踪运动IMU局部信息补偿优化  

Automatic Driving Multi-target Tracking Motion IMU Local Information Compensation Optimization

作　　者：李睿敏 贾雄伟 任小文 王海峰[2] LI Ruimin;JIA Xiongwei;REN Xiaowen;WANG Haifeng(School of Mechanical and Electrical Engineering,Xi’an Traffic Engineering Institute,Xi’an 710300,China;College of Mechanical and Electrical Engineering,Qingdao University,Qingdao 266071,China)

机构地区：[1]西安交通工程学院机械与电气工程学院,西安710300 [2]青岛大学机电工程学院,青岛266071

出　　处：《自动化与仪表》2024年第9期33-36,共4页Automation & Instrumentation

基　　金：校级科学研究重点项目(2024KY-13);校级中青年基金项目(2023KY-08);陕西省教育厅科研计划项目(20JK0751)。

摘　　要：多目标追踪是自动驾驶系统中的关键模块之一,其结果的优劣主要取决于追踪模块中数据关联过程的准确度。通过引入观测点自身的惯性测量单元(IMU)或全球定位系统(GPS)数据,在一帧数据到达之后计算当前帧局部坐标系与上一帧局部坐标系之间的旋转和平移关系,并对已追踪的物体状态按得到的坐标变换关系进行运动补偿,使其抵消因观测点自身运动造成的偏移量;这种运动补偿增强了追踪模块的数据关联环节,提高追踪时三维包围框的关联成功率,降低误关联数量,改善多目标追踪的精度;在相关追踪框架及KITTI数据集上的原型验证表明,所提的运动补偿优化方法实现了1%左右的精度提升。Multi-target tracking is one of the key modules in the automatic driving system,and its result mainly depends on the accuracy of the data association process in the tracking module.By introducing the inertial measurement unit(IMU)or global positioning system(GPS)data of the observation point itself,the rotation and displacement relations between the local coordinate system of the current frame and the local coordinate system of the previous frame are calculated after the arrival of a frame of data,and the motion compensation is carried out for the state of the tracked object according to the obtained coordinate transformation relationship.Make it offset the offset caused by the motion of the observation point itself.The motion compensation enhances the data association link of the tracking module,improves the success rate of the association of the 3D enclosing frame,reduces the number of false associations,and improves the accuracy of the multi-target tracking.The prototype validation on the relevant tracking framework and KITTI dataset shows that the proposed motion compensation optimization method achieves about 1%accuracy improvement.

关 键 词：自动驾驶 运动补偿 多目标追踪 运动特征 全球定位系统 惯性测量单元 

分 类 号：TP273[自动化与计算机技术—检测技术与自动化装置]