大型复杂曲面三维测量的立体视觉规划研究

发布时间：2018-01-11 03:00

本文关键词：大型复杂曲面三维测量的立体视觉规划研究　出处：《哈尔滨理工大学》2016年硕士论文　论文类型：学位论文

【摘要】：现在大型装备制造业对产品精度与生产效率的要求越来越高,三维外形检测技术作为质量控制的有效手段,正逐渐在生产制造过程中起到越来越重要的作用。立体视觉测量对大型零部件三维外形检测具有测量精度高、灵活性好的特点,获得广泛的认可。但是,在大型复杂曲面零部件测量过程中,传统手持式双目立体视觉测量装置往往存在局部漏检、精度低以及工作效率低等问题,因此本文设计了一套基于双目立体视觉的全自动三维测量装置,并深入研究了三维测量中的立体视觉规划问题,建立了相应的规划算法。首先,根据对双目立体视觉测量过程的研究,确定了视觉测量装置的结构方案和硬件配置,并开发了测量装置的上位机和下位机系统,上位机系统主要包含通信参数设置、图像的显示与采集、路径及姿态的输入与回显等模块,下位机系统主要完成数据采集及对测量机构的运动控制。在给出了视觉测量网络的测量覆盖率与特征分辨率的规划目标的基础上,分析确定了视觉测量网络规划中的决策变量和约束条件,设计了视觉测量网络规划的多目标规划数学模型,为求解最优规划Pareto解集奠定基础。运用多目标遗传算法对视觉测量网络规划进行应用研究。具体设计了算法中的编码方式、种群处理方式、约束条件处理方法、个体适应度函数、遗传算法及算法的控制参数。以大型螺旋桨叶片为例,对其进行了测量网络规划,得到了一组满足规划目标及约束条件的全局Pareto解。并通过实验验证了规划结果的可行性。本文最后提出了一种可满足区域约束条件的最短路径规划方法,以解决立体视觉测量中的路径规划问题。算法借助区域内寻路与区域间寻路的两层搜索实现了最短路径规划,通过与穷举法的对比,验证了算法的正确性。本文不但设计了一种视觉测量装置,而且为大型复杂曲面三维测量提供了立体视觉规划方法,对提高大型零件的测量精度具有重要意义。
[Abstract]:Nowadays, the requirement of product precision and production efficiency for large-scale equipment manufacturing industry is more and more high, 3D shape detection technology as an effective means of quality control. Stereo vision measurement has the characteristics of high measurement precision, good flexibility, and has been widely recognized. In the process of measuring large and complex curved surface parts, the traditional hand-held binocular stereo vision measurement device often has some problems, such as local missing detection, low precision and low working efficiency. Therefore, this paper designs a set of automatic 3D measurement device based on binocular stereo vision, and deeply studies the stereo vision planning problem in 3D measurement, and establishes the corresponding planning algorithm. According to the research of binocular stereo vision measurement process, the structure and hardware configuration of the vision measuring device are determined, and the upper and lower computer systems of the measuring device are developed. The upper computer system mainly includes communication parameter setting, image display and acquisition, path and attitude input and echo module. The lower computer system mainly completes the data acquisition and the motion control of the measuring mechanism. The planning target of the measurement coverage and feature resolution of the vision measurement network is given. The decision variables and constraints in vision measurement network planning are analyzed and the multi-objective programming mathematical model of vision measurement network planning is designed. In order to solve the optimal programming Pareto solution set, the application of multi-objective genetic algorithm to the vision measurement network planning is studied. The coding method and population processing method of the algorithm are designed in detail. The control parameters of constraint condition processing method, individual fitness function, genetic algorithm and algorithm. Taking the large propeller blade as an example, the measurement network planning is carried out. A set of global Pareto solutions satisfying the planning objectives and constraints are obtained. The feasibility of the planning results is verified by experiments. At last, a shortest path planning method is proposed to satisfy the regional constraints. Law. In order to solve the problem of path planning in stereo vision measurement, the algorithm realizes the shortest path planning with the help of the two-layer search between the region and the region. The algorithm is compared with the exhaustive method. In this paper, not only a vision measuring device is designed, but also a stereo vision planning method is provided for the 3D measurement of large complex surfaces, which is of great significance to improve the measuring accuracy of large parts.
【学位授予单位】：哈尔滨理工大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TG806

【参考文献】