车载雷达关键场景的数据采集与标记方法  

Data Collection and Annotation Method for Radar on Some Key Scenarios

作　　者：黄楷博 邓伟文 王莹[2] 赵蕊 丁娟 Huang Kaibo;Deng Weiwen;Wang Ying;Zhao Rui;Ding Juan(School of Transportation Science and Engineering,Beihang University,Beijing 100191;College of Computer Science and Technology,Jilin University,Changchun 130015;Jiaxing Nanhu University,Jiaxing 314000)

机构地区：[1]北京航空航天大学交通科学与工程学院,北京100191 [2]吉林大学计算机科学与技术学院,长春130015 [3]嘉兴南湖学院,嘉兴314000

出　　处：《汽车工程》2024年第12期2257-2266,共10页Automotive Engineering

基　　金：浙江省“领雁”研发攻关计划项目(2023C01238);浙江省“尖兵”研发攻关计划项目(2023C01133)资助。

摘　　要：车载雷达的虚警和漏报是影响自动驾驶系统安全可靠的关键因素之一,因此需要大量带标签的测试数据开展针对性的研究。但虚警和漏报的发生概率较低,且雷达目标状态不稳定导致雷达目标难以标记。对此,本文首先根据雷达虚警和漏报的产生机制设计能高效获取雷达关键数据的测试方案。然后通过构建关联度函数,以量化雷达目标与场景目标之间的关联并使用遗传算法优化该函数,在此基础上建立雷达目标的自动标记方法。最后通过实采数据验证本文方法的有效性。实验结果表明所提出的方法能高效获取关键的虚警和漏报数据,本文的标记方法也能准确识别出场景目标对应的雷达目标,并区分出虚警目标和真实目标。False alarm and missed alarm of automotive radar are key factors affecting the safety and reliability of autonomous driving systems,thus requiring a large amount of labeled test data for targeted research.However,the occurrence probability of false alarm and missed alarm is low,and the unstable status of radar targets makes it difficult to label them.Therefore,in this paper,firstly efficient test schemes are designed to obtain key radar data based on the generation mechanism of radar false alarm and missed alarm.Then,by constructing a correlation function to quantify the correlation between radar targets and scene targets and using genetic algorithms to optimize this function,an automatic labeling method for radar targets is established.Finally,the effectiveness of the proposed method is verified through real data acquisition.The experimental results show that the proposed method can efficiently obtain crucial false alarm and missed alarm data.The labeling method in this paper can accurately identify radar targets corresponding to scene targets and distinguish between false alarm and real targets.

关 键 词：自动驾驶 毫米波雷达 虚警 漏报 目标标记 

分 类 号：U463.6[机械工程—车辆工程]