管路实物检测是保证管路制造质量和实现无应力自动化装配的关键环节。项目在已有的基于多目视觉的管路测量系统及研究成果的基础上，以提高管路测量的精度、效率和自动化、智能化程度为目标，开展极具发展前景的基于单幅图像的复杂多分支管路智能理解、三维重建及位姿求解所需的新理论、新方法研究，实现对复杂多分支管路的智能理解、位姿识别及精确三维重建，完成包括非圆形截面管路、 多弯曲半径及自由形状管路、多分支管路及其附件的高精度快速自动测量与智能识别。具体内容包含一维流形管路数学描述方法、非欧几何三维结构不变量在单幅及多幅图像三维重建中的重要意义、基于能量泛函的单幅图像管路三维重建方法、无纹理少特征管路智能感知与理解、高精度位姿感知与表达等关键实现技术。通过上述内容的研究，项目将解决基于单幅图像的多分支复杂管路智能理解与快速测量的瓶颈难题，加速机器视觉技术在多样化零件中的非接触、快速智能测量中的应用。

The tube measurement is the most effective method to ensure the tube's manufacturing quality and realize the automation stress-free assembly. The purpose of this project are to improve the accuracy, efficiency, automaticity and the intelligent level of the tube measurement. On the basis of our tube measurement system and research results, we investigate the most prospective theory and technique including the understanding, pose detection and accurate 3D reconstruction of complex tubes based on only one single image, to complete high precision and fast automatic measurement of complex tubes including non-circular-cross-section tubes, multi-bending radius tubes and free form tubes. The research contents includes one dimensional manifold tubes mathematical description, the significance of the three-dimensional geometry invariants in single and multiple image based 3D reconstruction, energy functional-based tube 3D reconstruction of a single image, intelligent perception and understanding of tubes with less texture characteristics, high precision position perception and expression and so on. Through the above studies, the key theory and technique on intelligent understanding and rapid measurement of complex multi-branch tubes will be solved, resulting in application promotion of the machine vision technology in the non-contact, rapid, intelligent measurement of various parts.