基于机器视觉的实体模型逆向重建系统  

Reverse Reconstruction System of Real Model Based on Monocular Vision

作　　者：张振凯 胡正发[1,2] 张伟 胡敏[2] 卢文豪 刘倩文 ZHANG Zhen-kai;HU Zheng-fa;ZHANG Wei;HU Min;LU Wen-hao;LIU Qian-wen(School of Physics and Optoelectronic Engineering,Guangdong University of Technology,Guangzhou 510000,China;Synergy Innovation Institute for Modern Industries of GDUT Dongyuan,Heyuan Guangdong 517000,China)

机构地区：[1]广东工业大学物理与光电工程学院,广州510000 [2]东源广工大现代产业协同创新研究院,广东河源517000

出　　处：《组合机床与自动化加工技术》2021年第1期57-60,65,共5页Modular Machine Tool & Automatic Manufacturing Technique

基　　金：广东省科技创新战略专项资金项目(2018B090902003);广东省重大科技专项项目(2011A080801015)。

摘　　要：为了解决点激光扫描重建过程中点云稀疏、颜色纹理特征不明显的问题提出了一种基于相机图像信息与激光点云融合的三维重建方法。该方法结合线激光三角测距、相机坐标变换关系获取物体深度信息和点云数据,然后利用随机抽样一致(RANSAC)原理进行图像配准,实现对单目标实体的逆向建模。实验结果表明,该系统具有较高的可靠性和良好的精度,当扫描角度为0.2°时,系统重建误差小于0.2 mm,重建精度高于0.35%。To solve the problems of sparse point clouds and inconspicuous color and texture features during point laser scanning reconstruction.A 3D reconstruction method based on the fusion of camera image information and laser point clouds is proposed.This method combines line laser triangulation and camera coordinate transformation to obtain object depth information and point cloud data,and then uses the random sampling consistency(RANSAC)principle to perform image registration to achieve reverse modeling of a single object.The results show that the system has high reliability and good accuracy.When the scanning angle is 0.2°,the reconstruction error of the system is less than 0.2 mm,and the reconstruction accuracy is higher than 0.35%.

关 键 词：三维重建 深度信息 激光点云 RANSAC 

分 类 号：TH16[机械工程—机械制造及自动化] TG506[金属学及工艺—金属切削加工及机床]