采用传感器融合的车辆前方碰撞预警系统  

Vehicle front collision warning system based on sensor fusion

作　　者：胡远志[1] 董泰宏 罗毅 蒲浩 HU Yuanzhi;DONG Taihong;LUO Yi;PU Hao(Key Laboratory of Advanced Manufacture Technology for Automobile Parts,Ministry of Education,Chongqing 400054,China)

机构地区：[1]重庆理工大学汽车零部件先进制造技术教育部重点实验室,重庆400054

出　　处：《重庆理工大学学报（自然科学）》2022年第7期46-53,共8页Journal of Chongqing University of Technology：Natural Science

基　　金：重庆理工大学研究生创新项目(clgycx 20202015)。

摘　　要：前方碰撞预警系统(FCW)是汽车自动安全系统之一,设计良好的FCW需要对车间碰撞时间(TTC)特征参数进行深入研究。针对激光雷达识别静止物体不足、视觉传感器检测车间TTC精度低稳定性差的问题,设计了一种基于激光雷达与摄像头融合的车间碰撞时间估计算法。该算法先进行摄像头与激光雷达的数据配准与同步,利用摄像头YOLOv3算法检测当前车道前方的车辆目标;再提取图像中的特征点并基于特征点匹配进行车辆目标跟踪;最后针对激光雷达点云数据、摄像头图像数据分别计算车间碰撞时间,使用卡尔曼滤波方法融合2种结果。测试结果表明:与传统单一传感器的碰撞时间估计相比,该算法表现更加稳定,异常值少,估计误差更小。Front collision warning system(FCW)is one of the automatic safety systems of vehicles.A well-designed FCW needs to deeply study the characteristic parameters of workshop collision time(TTC).Aiming at the problems of insufficient recognition of stationary objects by lidar,low accuracy and poor stability of visual sensor in detecting workshop TTC,a workshop collision time estimation algorithm based on the fusion of lidar and camera is designed.The algorithm first carries out the data registration and synchronization between the camera and the lidar,and uses the camera YOLOv3 algorithm to detect the vehicle target in front of the current lane.Then the feature points in the image are extracted and the vehicle target is tracked based on feature point matching.Finally,for the LIDAR point cloud data and camera image data,the workshop collision time is calculated respectively,and the Kalman filter method is used to fuse the two results.The test results show that compared with the traditional single sensor collision time estimation,the algorithm is more stable,less outliers and less estimation error.

关 键 词：碰撞时间 点云聚类 YOLOv3 特征匹配 传感器融合 卡尔曼滤波 

分 类 号：U463.6[机械工程—车辆工程]