大尺寸曲面零件的双目视觉测量网络规划研究

Study on binocular vision measurement network layout for large curved surface parts

大尺寸复杂曲面零件的视觉检测,需要多个视觉形成视觉网络才能实现完整三维零件重构。针对大尺寸复杂曲面零件的双目视觉测量网络规划问题,提出了一种基于改进的多目标并列选择遗传算法的视觉测量网络规划方法。首先建立双目视觉测量网络的几何模型,得出了双目视觉测量网络的解空间,为评价视觉测量网络规划的优劣,提出了视觉测量网络的覆盖率与分辨率的概念,通过遗传算法在解空间中优选地搜索出最优解,根据覆盖率与分辨率这两个评价指标,得出最优的规划方案。以大型螺旋桨叶片为例,进行双目视觉测量网格规划实验,验证了算法的可行性。

In the visual detection of large complex curved surface parts, multiple machine vision systems are needed to realize the three-dimension reconstruction of the complete parts. Aiming at the planning problem of binocular ster- eo vision measurement network layout of large complex curved surface parts, a planning method of visual measure- ment network based on improved multi-objective genetic algorithm with parallel selection is presented. Firstly, a geo- metric model of binocular stereo vision measurement network is established to get the solution-space of binocular ster- eo vision measurement network. Secondly, the concepts of coverage rate and resolution are presented to evaluate the merits of the vision measurement network layout. And then, the genetic algorithm is used to search the optimal solu- tions in solution-space, an optimum plan is determined based on the comprehensive comparison of these two evalua- ting indexes. Taking a large propeller blade as example, the experiment on binocular stereo vision measurement net- work planning was conducted, which verifies the feasibility of the algorithm.