面向大型机械装备逆向工程的多目视觉测量方法研究

The reverse engineering of mechanical equipment is the process to recover the 3D geometric model of a mechanical product by some measuring means from the real product when either its design drawing is only partially available or completely absent. This project is to investigate the theories and algorithms of multi-view based measurement for reverse engineering of large mechanical equipment. The main contents include: (1) Camera distortion modeling and parameter estimation; (2) Macro-feature based camera calibration; (3) Precision measurement by a fixed vision system composed of a large-view camera with low resolution and a small-view camera with high resolution; (4)Measurement error modeling and analysis. The main difficulties of the project lie in the low textured nature of mechanical products to be measured, their very large size, and the required high measurement accuracy. The project’s expected outcome could not only useful to the reverse engineering of large mechanical equipment, but it could also enrich visual metrology research and promote its further development.

机械装备逆向工程是指在没有产品图纸或者图纸不完整的情况下，按照现有产品模型的实物，利用相关技术获取产品外形的过程。本项目旨在研究探索大型机械装备逆向工程中的多目视觉测量理论和方法。主要研究内容包括：（1）摄像机畸变模型建模和参数估计；（2）基于宏特征的摄像机标定；（3）大视野低分辨率相机和小视野高分辨率相机的混合相机测量；（4）测量误差分析与建模。本项目的难点是对弱纹理大型物体的视觉高精度测量，项目成果不仅可以为大型机械装备逆向工程提供技术手段，而且可望丰富和发展基于视觉的大型物体测量理论。

本项目针对机械装备中没有产品图纸或图纸不完整的情况下，如何依据现有机械产品模型获取产品外形的问题进行研究。已圆满完成了研究内容，取得了良好的研究成果，达到了预期研究目标。项目主要研究成果包括：1)针对摄像机畸变参数，提出了一种基于单幅图像的摄像机畸变参数估计方法；2)针对摄像机外参数，研究了P3P问题，主要有P3P问题在光心附近的奇异点问题；3)采用四边形作为特征信息，研究了一种基于图像的四边形自动检测算法；4)针对视觉测量误差分析与建模，基于Matlab进行了大型物体视觉测量模拟和精度分析。该课题的研究不仅为机械装备逆向工程的测量问题，提出一种有效的途径和方法，而且也可以进一步为太原重型机械厂产品的设计与开发提供技术依据。该项目所取得的成果已发表在《Pattern Recognition Letters》、《中国科学》、《计算机辅助设计与图形学学报》等国内外学术刊物和学术会议上，发表与录用论文12篇。