节点文献

多组双目激光融合的三维扫描系统研究

Three-dimensional Scanning System with Multiple Sets of Binocular Image and Structured Light

【作者】 周建华

【导师】 朱明程; 徐渊;

【作者基本信息】 深圳大学 , 集成电路工程(专业学位), 2016, 硕士

【摘要】 近年来,随着逆向工程、医学成像和工业检测等行业的迅速发展,三维扫描作为其中核心技术之一,受到了普遍的关注和研究。现阶段的三维扫描设备多是采用竖式扫描结构,往往无法与三维打印机同步运行;此外多数设备专用于扫描点云数据,后期点云数据处理在电脑端进行。所以总的来说,目前三维扫描设备存在着功能单一、效率低下、集成度低等特点。针对这些问题,本课题设计出一套四个轴向环绕形状的多组双目激光扫描系统,此系统通过四组双目激光测量模组同时向前投射结构光,利用视差原理获得对物体全方位的扫描点云数据,并通过在Zedboard平台上软件实现,以达到扫描处理一体化的目的,提高系统的集成度,此外系统采用模向扫描的方式,实时的三维信息输出可以适配到三维打印机,方便对接,为其功能多样化提供了基础。本文首先阐述了立体视觉、结构光、光带质心提取、点云数据拟合平滑、统一坐标系建立、点云数据的融合等关键点的相关原理,其中一些模块本文针对性地提出了改进型的算法,如光带质心提取过程里加入邻域滤波、点云融合采用加权距离确定融合点等。文章着重地介绍了系统设计,包括硬件与软件两大部分,硬件上创新性地设计出扫描支架平台、立体视觉模组模具,并将其与升降平台、激光器、图像传感器无缝对接起来,组成整套完整的三维扫描设备。软件上为充分利用已有硬件的特性,采用了如异步通知、多进程等编程模型以达到提高系统扫描效率的目的。经过这样的设计,系统的扫描从双目激光图像采集、光带提取、点云生成拟合到点云的融合都可以流水型的层次进行,保证了三维信息的持续准确输出。最后通过实验结果及数据测量表明,本文所设计的多组双目激光的三维扫描系统精度合适,真实感强且总体效果令人满意,能满足于当前物体三维扫描的需求,亦为将来的功能拓展提供了基础。

【Abstract】 In recent years, with the reverse engineering, medical imaging and industrial testing and other relevant industries, the rapid development of 3d scanning, as one of the core technology, has received the widespread attention and research. The current 3d scanning equipment mostly is adopt vertical scanning structure, which often cannot run to sync with 3d printers. In addition the majority of equipment dedicated to the scanning point cloud data, the late point cloud data processing on the computer side. In general, the 3d scanning is single function, inefficiency, low level of integration. To solve these problems, this topic design a set of four axial round shape more groups of binocular laser scanning system. Four groups of binocular laser measurement module throw to object onwards with structure light at the same time, which obtain scanning point cloud data using the parallax principle of objects comprehensive. And that is realizing software development on Zedboard platform, to achieve the purpose of the scanning process integration, improve the system integration. This system scan object horizontally to obtain three-dimensional information, which can be adapted to 3d printers and provides the basis for its function diversification.This article first expounds the centroid extraction, stereo vision, structured light, light fitting point cloud data smoothing, establish unified coordinate system, the point cloud data fusion and the related principle of some modules, this paper puts forward an improved algorithm, such as the image centroid extraction process to join neighborhood filtering, point cloud fusion weighted distance is adopted to define the fusion point, etc.The paper emphatically introduces the system design, including hardware and software. The design includes scanning platform, the stereo vision module, the lifting platform, laser, image sensor, which is a set of the complete 3d scanning device. To make full use of existing hardware, the design of software adopted such as asynchronous notification, multi-process programming model to enhance the efficiency of the scanning system. Through this design, the system of scanning from binocular image acquisition, the extracting line laser, the smoothness of cloud point, establishing coordinate system,the fusion of point cloud can be pile line operation, ensure the output of 3d information being accurately.Finally, experimental results and measurements show that the 3d scanning system with multiple sets of binocular image and structured light is strong sense of reality, appropriate Precision and satisfying effect, which can meet the demand of the 3d scanning industry and provides a basis for future expansion.

  • 【网络出版投稿人】 深圳大学
  • 【网络出版年期】2017年 05期
  • 【分类号】TN24
  • 【下载频次】229
节点文献中: 

本文链接的文献网络图示:

本文的引文网络