论文总字数:30007字
摘 要
工业机器人领域中常用的两种机器人编程方式为人工示教与离线编程,人工示教操作简单但是过程繁琐、效率低,离线编程安全、精度高但是难以事先获取机器人作业场景的三维模型,对于小批量、个性化定制的工件生产任务来说这一问题显得尤为突出。针对这一问题,本文对基于点云扫描的机器人作业场景快速重构方法进行了研究,将YASKAWA六自由度机械臂与三维线结构光传感器相结合,对工件上表面进行扫描以获取点云数据,对点云数据进行处理以生成三维STL模型,最终对模型进行孔洞修补及可视化显示。完成的主要工作包括:
点云的获取与处理。本文利用三维线结构光传感器获取作业场景的点云数据,对点云数据进行坐标系转换、工作平面去除、离群点去噪等预处理,接着对点云数据进行聚类,分割为多个子工件点云,采用体素栅格下采样对点云数据进行精简。
三维重建与模型修复。首先对点云数据建立基于KD树的索引,通过最小二乘法拟合平面,计算各点的法向量信息,接着基于贪婪投影三角化算法对点云进行建模,生成工件的三维STL模型;其次建立半边数据结构存储STL模型中的三角片信息,利用半边结构查找孔洞,基于最小角优先的思想对孔洞进行修补。
最终本文在各部分研究的基础上,设计了良好的人机交互界面,实现了文件读写、模型显示、数据处理等功能,并对不同的模型进行了测试,获得了不错的效果。
关键词:工业机器人,点云处理,三维重建,STL模型,孔洞修补
Abstract
There’re two kinds of robot programming methods commonly used in industry, which are manual teaching and off-line programming. Manual teaching is easy to learn but its process is cumbersome and inefficient. Off-line programming is safe and accurate, but it is difficult to obtain the three-dimensional model of the workpiece in advance. This problem is particularly prominent for small batch and customized workpiece production. In order to solve this problem, researches on the method of rapid reconstruction of working scene based on point cloud scanning have been conducted. Three-dimensional linear structured light sensor is combined with the six DOF manipulator to scans the upper surface of workpieces. After the collection of point cloud data, several processes will be carried out to generate the three-dimensional STL model. Then the STL model will be repaired and visualized. The main tasks completed include:
Acquisition and processing of point cloud data. First use three-dimensional linear structured light sensor to obtain the point cloud data of the working scene, then preprocesses such as coordinate transformation, working plane removal, outlier filtering will be carried out. The point cloud data will be clustered into several sub clouds, and simplified by voxel sampling.
Three-dimensional reconstruction and model repair. Firstly, the point cloud data is indexed based on KD tree, normal information of each point is calculated by using the least square method to fit the plane. Then the point cloud data is modeled based on the greedy projection triangulation algorithm to generate the three-dimensional STL model of the workpiece. Secondly, the half-edge data structure is established to store the triangle information in the STL model, then the holes on the model will be found and repaired based on the idea of minimum angle priority.
Based on the research of each part, a nice human-computer interaction interface is designed, including file reading and writing module, model display module, data processing module and other modules. Tests on different models are presented at the end of the paper which obtain good results.
KEY WORDS: Industrial robot, Point cloud processing, three-dimensional reconstruction, STL model, Mesh repair
目 录
摘 要 I
Abstract II
第一章 绪论 1
1.1 课题的研究背景 1
1.2 国内外研究现状 2
1.2.1 点云获取技术的研究现状 2
1.2.2 点云处理技术的研究现状 4
1.2.3 三维重建技术的研究现状 6
1.2.4 模型修复技术的研究现状 6
1.3 本论文的主要工作 8
1.4 本论文的章节安排 9
第二章 点云数据的获取与处理 10
2.1 传感器的选型 10
2.2 点云预处理 12
2.2.1 机器人坐标系变换 12
2.2.2 基于直通滤波的平面去除 13
2.2.3 基于统计滤波的离群点去噪 14
2.3 基于欧氏聚类的点云分割 15
2.4 基于体素采样的点云精简 17
2.4.1 体素采样的原理 17
2.4.2 体素大小的选择 18
2.5 本章小结 19
第三章 三维重建与模型修复 21
3.1 建立点云索引 21
3.1.1 基于KD树的点云空间索引 21
3.1.2 基于KD树的最近邻查找算法 22
3.2 基于最小二乘法拟合平面计算法向量 23
3.3 贪婪投影三角化 24
3.4 半边数据结构 25
3.5 模型孔洞修补 27
3.6 本章小结 29
第四章 软件系统的开发 31
4.1 三维重建软件的总体框架设计 31
4.2 总体结果展示 31
4.3 本章小结 34
第五章 总结与展望 36
5.1 总结 36
5.2 展望 36
参考文献 38
致 谢 40
绪论
课题的研究背景
1959年,有着“机器人之父”美誉的约瑟夫·恩格尔伯格先生发明了世界上第一台工业机器人(如图1-1),自此之后,工业机器人逐渐走进人们的视野,成为人们的得力助手。机器人并不是简单地代替工人进行劳动,而是综合了计算机、控制学、结构学、传感技术、人工智能等高新技术,能够完成复杂的、需要快速分析和判断的工作,并且机器人具有可长时间持续工作、工作精度高、能在恶劣环境中运行的优点。因此,机器人的可应用领域日益广泛,机器人技术的发展也愈加快速,机器人的应用情况,可以体现一个国家工业自动化发展的水平。
剩余内容已隐藏,请支付后下载全文,论文总字数:30007字
该课题毕业论文、开题报告、外文翻译、程序设计、图纸设计等资料可联系客服协助查找;