基于图像与高精定位融合数据的封闭场地自动驾驶汽车驾驶测试方法  

Driving test methods of autonomous vehicles in closed field based on image and high precision positioning fusion data

作　　者：戴力源 杨达[1] 李凯 彭鸿捷 孙峰 Dai Liyuan;Yang Da;Li Kai;Peng Hongjie;Sun Feng(School of Transportation&Logistics,Southwest Jiaotong University,Chengdu 610097,China;School of Computing&Artificial Intelligence,Southwest Jiaotong University,Chengdu 610097,China;Sichuan Intelligent Highway Technology Co.,Ltd.,Chengdu 610041,China)

机构地区：[1]西南交通大学交通运输与物流学院,成都610097 [2]西南交通大学计算机与人工智能学院,成都610097 [3]四川智慧高速科技有限公司,成都610041

出　　处：《计算机应用研究》2023年第11期3333-3340,3347,共9页Application Research of Computers

基　　金：国家自然科学基金资助项目(52172333);成都市重点研发计划资助项目(2022YF0500376SN)。

摘　　要：现有的自动驾驶汽车场景仿真测试结果真实性不足,基于实际测试数据的评价方法缺乏实时性。为结合真实数据对测试过程展开实时评价,提出了基于“原场景”库的自动驾驶测试场景生成方法,利用差分定位GPS以及视频图像处理等工具,配合测试场地的高精地图设计了自动驾驶汽车实时测试评价方法,在自动驾驶测试场对本方法进行了实地验证。结果表明本方法具有较高的可行性,能够全面实时地对车辆自动驾驶能力进行评价。The authenticity of the existing simulation test results of autonomous vehicle scenarios is insufficient,and the eva-luation method based on actual test data lacks real-time performance.In order to evaluate the test process in real time with real data,this paper proposed an automatic driving test scene generation method based on the original scene library.Using diffe-rential positioning GPS and video image processing tools,combined with the high-precision map of the test site,it designed a real-time test and evaluation method for autonomous vehicles.The method was verified by an autonomous driving test field.The results show that this method has high feasibility,and can comprehensively evaluate the automatic driving ability of vehicles in real-time.

关 键 词：智能交通 自动驾驶测试 封闭场地 差分GPS 综合评价 

分 类 号：U495[交通运输工程—交通运输规划与管理]