A new multi-vision-based reconstruction algorithm for tube inspection

Complex tubes are used widely in aerospace vehicles today, and their accurate assembly can determine the equipment’s performance and longevity. The machining precision of the tube determines its assembly reliability. As most tubes are metallic, springback is a major factor preventing them from realizing accuracy requirements in primary processing. Thus, it is important to inspect the processed tubes and then fix any geometric errors to satisfy the assembly requirements. However, the widely adopted tube inspection method in the literature is time-consuming and inconvenient because it is greatly dependent on human operation. In fact, there is no effective inspection method for tubes with complex shapes and large dimensions. To address this, an automatic tube reconstruction algorithm based on multi-vision is proposed in this paper. The algorithm discretizes the tube into many small cylinders, referred to as primitives. Multi-vision technology and the tube edges are then used to reconstruct the primitives to form the initial model, from which a three-dimensional model can be constructed within 2 min. Our algorithm dramatically improves the reconstruction speed because it concentrates only on the reconstruction of finite cylinders rather than point clouds on the tube surface. And the reconstruction accuracy is 0.17 mm, allowing arc recognition of bending angles ranging from 1° to 180°. Also, the restrictions due to reflection on the surface and the lack of necessary texture for matching are solved at the same time. A comparison of reconstructed and computer-aided design (CAD) models resolves geometric error and springback for machining parameter optimization, providing improved accuracy particularly for tube bending, which is of great significance for the realization of automated tube production.