基于RGB图像和点云数据融合的汽车零配件配准  

Automotive parts registration based on RGB image and point cloud data fusion

作　　者：库宗帆 陈灯[1] 郑朝晖 KU Zong-fan;CHEN Deng;ZHENG Zhao-hui(Hubei Province Key Laboratory of Intelligent Robot,Wuhan Institute of Technology,Wuhan 430079,China;School of Mathematics and Physics,Wuhan Institute of Technology,Wuhan 430079,China)

机构地区：[1]武汉工程大学智能机器人湖北重点实验室,湖北武汉430079

出　　处：《计算机工程与设计》2025年第4期1182-1189,共8页Computer Engineering and Design

基　　金：湖北省教育厅科技计划研究基金项目(B2022056);武汉工程大学研究生教育创新基金项目(CX2022364)。

摘　　要：针对工业场景下经典迭代最近点(iterative closest point,ICP)算法在点云位姿估计中初始位姿敏感度高、迭代时间长的问题,提出一种基于RGB图像的快速点云配准方法。分别采集RGB图像和点云数据,使用ORB(oriented FAST and rotated BRIEF)算法提取RGB图像特征点,利用Brute-Force算法进行初始匹配,采用随机采样一致性算法优化匹配,得到单应矩阵和旋转平移矩阵,求解汽车零配件初始位姿。进一步采用主成分分析法和双向KD树近邻搜索算法对预处理后的点云数据进行精确配准。实验结果表明,所提算法相较ICP算法,在配准速度和精度上分别提高了87.2%和5.0%,相对于FR-ICP(fast and robust iterative closest point)算法,在配准精度相当的情况下,配准速度提高了55%。To solve the problems of high initial pose sensitivity and long iteration time of the classical iterative closest point(ICP)algorithm in point cloud pose estimation in industrial scenarios,a fast point cloud registration method based on RGB images was proposed.MethodsRGB image and point cloud data were collected respectively,the oriented FAST and rotated BRIEF(ORB)algorithm was used to extract the RGB image feature points,the Brute-Force algorithm was used for initial matching,and the random sampling consistency algorithm was used to optimize the matching,and the single response matrix and rotation translation matrix were obtained to solve the initial pose of auto parts.Furthermore,principal component analysis(PCA)method and two-way KD tree nearest neighbor search algorithm were used to accurately register the preprocessed point cloud data.Experimental results show that the proposed algorithm improves the registration speed and the accuracy by 87.2%and 5.0%respectively compared with the ICP algorithm.Compared with the fast and robust iterative closest point(FR-ICP)algorithm,the regi-stration speed is increased by 55%with comparable registration accuracy.

关 键 词：图像处理 点云配准 迭代最近点算法 特征提取 特征匹配、随机采样一致性 主成分分析法 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]