基于双目视觉的冷凝器管口识别与重建研究

为改善现有大型冷凝器清洗装置对冷凝器管口三维定位的不足，便于操作人员远程监控清洗效果，设计冷凝器清洗机器人视觉系统。采用平行双目视觉结构模型，通过冷凝器管口轮廓识别和基于全多重网格算法的立体匹配，实现管口三维定位和表面重建。实验结果表明：该系统能够有效计算冷凝器各管口三维坐标；与传统方法相比，管板表面三维重建效果良好，能够满足冷凝器可视化要求。

Research on Recognition and Reconstruction of Condenser Nozzle Based on Binocular Vision

In order to improve the lack of three-dimensional positioning of the condenser nozzle in the existing large-scale condenser cleaning device, and facilitate the operator to remotely monitor the cleaning effect, a condenser cleaning robot vision system was designed. Using parallel binocular vision structure model, through the condenser nozzle contour recognition and stereo matching based on the full multi-grid algorithm, the nozzle three-dimensional positioning and surface reconstruction were realized. The experimental results show that by using the system, the three-dimensional coordinates of each condenser nozzle can be effectively calculated; compared with the traditional method, the three-dimensional reconstruction effect of the tube sheet surface is good, which can meet the visualization requirements of the condenser.