基于相机与激光雷达融合多目标检测算法研究  

Research on Multi-target Detection Algorithm based on Camera and LiDAR Fusion

作　　者：刘志霞 王炜 仇焕龙[1,2] Liu Zhixia;Wang Wei;Qiu Huanlong(CATARC(Tianjin)Automotive Engineering Research Institute Co.,Ltd.,Tianjin 300300;China Automotive Technology and Research Center 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 [2]中国汽车技术研究中心有限公司,天津300300

出　　处：《中国汽车》2024年第4期36-42,共7页China Auto

基　　金：国家863项目(2011AA11A207)。

摘　　要：准确的多目标感知系统是自动驾驶技术的关键。本文提出了一种基于相机与激光雷达融合的多目标检测算法。针对相机传感器无法获得准确的目标距离等深度信息,激光雷达无法获得准确的目标类别信息的问题,首先采用嵌入自适应特征融合模块的YOLOv7网络处理相机数据,同时对激光雷达数据进行点云预处理以消除无用的噪声点;其次,利用坐标变换将激光点云数据和相机数据转换到像素坐标系中;最后,采用基于ROI感兴趣区域的方法对点云进行聚类处理,以参数加权的方式融合两种传感器的检测结果。实验结果表明,嵌入改进YOLOv7网络的融合算法能够检测出更加准确的目标信息。An accurate multi-target sensing system is the key to autonomous driving technology.A multi-target detection algorithm based on camera and LiDAR fusion is proposed.The problem that camera sensors cannot obtain accurate depth information such as target distance and LiDAR cannot obtain accurate target category information is addressed.Firstly,a YOLOv7 network embedded with an adaptive feature fusion module is used to process the camera data,while the LiDAR data is pre-processed with a point cloud to eliminate useless noise points;secondly,a coordinate transformation is used to convert the laser point cloud data and the camera data into a pixel coordinate system;finally,a ROI region of interest-based method is used to cluster the point cloud and fuse the two sensor detection results.The experimental results show that the fusion algorithm embedded in the improved YOLOv7 network is able to detect more accurate target information.

关 键 词：汽车工程 多目标检测 深度学习 自适应特征融合模块 YOLOv7 

分 类 号：TN9[电子电信—信息与通信工程]