基于点阵激光的弱纹理对象三维建模研究  

Research on 3D Modeling of Weak Texture Objects Based on Lattice Laser

作　　者：郝金鹏 坎标[1] HAO Jinpeng;KAN Biao(School of Mechanical Engineering and Rail Transit,Changzhou University)

机构地区：[1]常州大学机械与轨道交通学院

出　　处：《仪表技术与传感器》2024年第10期100-104,共5页Instrument Technique and Sensor

基　　金：国家自然科学基金项目(51105046)。

摘　　要：针对弱纹理对象表面信息要素少,机器视觉三维建模困难的问题,提出一种使用双目视觉技术配合点阵激光重建目标对象的方法。采用点阵激光赋予目标对象表面纹理,提出相应的特征点识别及匹配算法。根据相邻视角中相机间的空间位姿关系拼接多视角的点云三维图。实验结果表明,重建的三维模型直径平均误差为2.05%,三维模型边缘和实物边缘的重合度为87.4%,点云坐标与实际点的三维坐标偏差较小,能满足针对弱纹理对象的三维建模需求。To address the challenges in reconstructing objects with weak textures,which have limited surface information and make 3D modeling difficult using machine vision,this paper proposed a method that utilized binocular vision technology in conjunction with point lattice laser to reconstruct target objects.The point lattice laser was used to texture the surface of the target object,and a corresponding feature point recognition and matching algorithm was introduced.The spatial pose relationship between adjacent cameras in multiple perspectives was used to stitch together the point cloud 3D maps.The experimental results show that the average diameter size error of the reconstructed 3D model is 2.05%,the coincidence degree of the edges of the 3D model and the edges of the physical object is 87.4%,and the deviation of the point cloud coordinates from the actual 3D coordinates of the points is small,which can meet the 3D modeling requirements for weakly textured objects.

关 键 词：双目视觉 点阵激光 弱纹理 三维建模 特征点识别及匹配 空间位姿 多视角 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]