虚拟传感器与视觉传感器联合标定技术研究  

Research on Joint Calibration Technology of Virtual Sensor and Visual Sensor

作　　者：贾鹏飞 张蕾 刘全周 李占旗 JIA Pengfei;ZHANG Lei;LIU Quanzhou;LI Zhanqi(CATARC(Tianjin)Automotive Engineering Research Institute Co.,Ltd.,Tianjin 300300)

机构地区：[1]中汽研(天津)汽车工程研究院有限公司,天津300300

出　　处：《汽车实用技术》2021年第21期36-39,共4页Automobile Applied Technology

基　　金：天津市科技计划项目(编号:17YDLJGX00020)。

摘　　要：为了在虚拟场景下实现对ADAS控制器的测试,文章对相机的测距模型进行了研究,在虚拟场景下完成了相机的标定,对相机捕获的图像进行了畸变校正,设计了4层BP神经网络对虚拟传感器与相机进行联合标定,得到了虚拟传感器与视觉传感器的转换关系,并设计了实验进行验证,通过测试最终的数据误差为0.37%。In order to test the ADAS controller in the virtual scene,the ranging model of the camera was studied in this paper.The camera calibration was completed in the virtual scene and the distortion of the image captured by the camera was corrected.A 4-layer BP neural network was designed to calibrate the virtual sensor and the camera.The conversion relationship between the virtual sensor and the visual sensor was obtained.The experiment is designed,and the data error is 0.37%.

关 键 词：ADAS BP神经网络 联合标定 

分 类 号：U495[交通运输工程—交通运输规划与管理]