基于BP神经网络的雷达与视觉前车信息融合研究

摘要：针对智能汽车单一传感器对前方车辆运动信息和方位识别误差较大的问题，在 Prescan

车辆在多种运动模式下数据，使用 BP 神经网络实现视觉和雷达对被跟踪车辆的运动信息的

融合，准确识别前方车辆的车速、车距、相对角度信息。研究结果表明：视觉和雷达单独识10

别的前方车辆车距、车速、相对角度信息与真值相比存在较大误差，而经过 BP 神经网络融

关键词：车辆工程；BP 神经网络；雷达；视觉；信息融合；

中图分类号：U461.91 文献标识码:A 15

contain large error information. In this paper,intellegent vehicle with radar and camera is

established in Prescan software,and the motion and position information ofahead vechile in

different motion condition are captured .The information fusion of radar and vision based on BP

neural network is uesd to obtain speed, distance and angle with higher accuracy of ahead 25

vechile.The result shows that the speed, distance and angle obtained by single radar or vision have

relative large difference compared with ture information value,and the fusion information output

by BP nerual network have relative small difference compared with ture information value.The

information fusion of radar and vision can effectively improve the motion and position recognition

accuracy of ahead vehicle. 30

Key words: automobile engineering ; BP neural network; radra;vision;information fusion;

交通事故的发生并提高道路的通行能力，因此，已经得到了各国政府的重视并逐渐成为汽车40