基于特征点匹配的排水管道声点云模型配准算法

A Drainage Pipe Acoustic Point Cloud Model Alignment AlgorithmBased on Feature Point Matching

摘要:

三维重建技术逐渐成为获取全面、完备、准确的排水管道信息的关键手段。而实际检测受到管道堵塞等工况与管道检测规程等因素限制,造成所获得的管道声呐点云模型会出现位姿不同、部分重叠或空缺等情况,需要通过配准获取完整管道模型。同时,传统ICP算法针对管道模型存在效率低、精度差的问题。因此,该文提出基于特征点匹配的粗配准与改进的ICP精细配准相结合的点云配准算法。首先,利用ISS特征点检测法检测出模型特征点,通过FPFH对特征点进行进一步的描述；其次,采用RANSAC算法筛选出正确特征匹配点集,利用四元数法解算出初始变换参数完成粗配准；最后,在粗配准基础上,通过改进最近对应点查询的ICP算法完成精细配准。实验结果表明了该文算法的可行性与优越性,能为后续排水管道缺陷检测提供高完备、全面、准确的点云模型。

关键词:声呐点云;点云配准;特征匹配;随机采样一致性;迭代最近点

Abstract:

3D reconstruction technology is becoming a key tool for obtaining comprehensive, complete and accurate information about drainage pipes. The actual inspection is limited by such factors as pipe blockage and pipe inspection protocols, resulting in different positions, partial overlaps or gaps in the obtained sonar point cloud model, and the need to obtain a complete pipe model through alignment. At the same time, the traditional ICP algorithm has the problems of low efficiency and poor accuracy for the pipeline model. Therefore, this paper proposes a point cloud alignment algorithm that combines coarse alignment based on feature point matching with improved ICP fine alignment. Firstly, the ISS feature point detection method is used to detect the model feature points, and the feature points are further described by the FPFH; secondly, the RANSAC algorithm is used to filter the correct feature matching point set, and the initial transformation parameters are solved by the quadratic method to complete the coarse alignment; finally, based on the coarse alignment, the ICP algorithm with improved nearest corresponding point query is used to complete the fine alignment. The experimental results demonstrate the feasibility and superiority of the algorithm, which can provide a high-precision point cloud data model for the subsequent detection of tunnel defects.

Key words:Sonar point cloud; point cloud alignment; feature matching; random sampling consistency; iterative nearest point