一种基于雷达和机器视觉信息融合的车辆识别方法.pdfVIP

汽 车 工 程 2015年(第37卷)第6期 Automotive Engineering 2015(Vol.37)No.6 2015119 一种基于雷达和机器视觉信息融合的车辆识别方法* 王宝锋,齐志权,马国成,陈思忠 (北京理工大学机械与车辆学院,北京 100081) [摘要] 为提高先进驾驶员辅助系统对车辆前方环境识别的准确性,提出一种基于雷达和视觉传感器信息融 合的车辆识别方法。 系统工作前预先对毫米波雷达和摄像头进行联合标定,并确定雷达坐标系和摄像头坐标系的 变换关系。 车辆识别过程中,首先根据雷达信息确定图像坐标系中的车辆识别感兴趣区域;然后对感兴趣区域进行 对称性分析获得车辆对称中心,并对车辆底部阴影特征进行分析处理完成车辆边缘检测;最后根据逆透视变换得到 车辆识别宽度,根据识别宽度对识别结果进行验证。 结果表明该算法具有较强的环境适应性和准确率,弥补了单一 传感器在车辆识别中的不足。 关键词:车辆识别;雷达;机器视觉;信息融合 Vehicle Detection Based on Information Fusion of Radar and Machine Vision Wang Baofeng,Qi Zhiquan,Ma Guocheng Chen Sizhong School of Mechanical Engineering,Beijing Institute of Technology,Beijing 100081 [Abstract] For increasing the detection accuracy of advanced driver assisted system of thefrontal surround- ingsofvehicle,avehicledetectionmethodbasedontheinformationfusionofradarandvisionsensorisputforward. Before thesystemstartsoperation,aco-calibrationofmillimeter-waveradarandcameraisconducted,andthetrans- formation relationship between radar and camera coordinatesisobtained. Theprocessof vehicle detection starswith the determination of the region of interest (ROI)for vehicle detection in image coordinatesbased on radar informa- tion. Then asymmetry analysisisperformedonROItogetthesymmetricalcenterofvehicle,andthefeaturesofve- hicle bottom shadow are analyzed and processed with vehicle edge detection completed. Finally,through inverse perspective mappingthewidthofvehicleisobtained,basedonwhichtheresultsof detectionareverified. Itisdem- onstrated that the algorithm proposed has strong environment adaptability and good accuracy,remedying the defects of vehicle detection with single sensor. Keywords:vehicle detect