高速场景下一种多传感器多目标检测跟踪系统的评价方法  

An Evaluation Method for Multi-Sensor Multi-Target Detection and Tracking System Based on Expressway Condition

作　　者：任玉科 Ren Yu-ke(Tongji University,School of Automotive Studies,Shanghai 201804,China)

机构地区：[1]同济大学汽车学院,上海201804

出　　处：《内燃机与配件》2023年第10期15-19,共5页Internal Combustion Engine & Parts

摘　　要：多源传感器融合感知结果的有效评估是自动驾驶的重要组成部分,对系统性能的优化、迭代和升级具有至关重要的作用。本文系统阐述了典型的多传感器多目标检测和跟踪系统的算法框架,并提出了一种以人工标注的激光雷达数据作为真实值的评估方法。通过该评价方法,使用在高速公路路况下采集的数据,实际评价了一个基于相机和毫米波雷达的检测和跟踪系统,获得了详实的测试数据,证明了该评价方法对于系统的改进具有指导意义。Effective evaluation of fusion perception results is an important part of autonomous driving and plays a crucial role in the optimization,iteration,and upgrade of system performance.In this paper,the algorithm framework of a typical multi-sensor multi-target detection and tracking system is described systematically,and an evaluation method based on marked lidar data as ground real value is proposed.By using this evaluation method,a detection and tracking system based on camera and millimeter wave radar is actually evaluated by using the data collected under expressway conditions,and detailed test data is obtained,which proves that the evaluation method has guiding significance for the improvement of the system.

关 键 词：自动驾驶 评价方法 多传感器多目标检测和跟踪系统 

分 类 号：U461.91[机械工程—车辆工程]