Research on the Tracking Registration Method of Markerless Augmented Reality for Product Assembly

Abstract Background There are many problems in the assembly of complex products in the industrial production process, such as the complex assembly process, the variety of parts involved, and the assembly process mainly depends on the familiarity of assembly personnel with the assembly process. The emergence of augmented reality technology can well solve the above problems. Using the characteristics of virtual real fusion of augmented reality technology can enable the assembly personnel to receive the virtual assembly guidance information corresponding to the operation in real time during the assembly operation. Methods In order to improve the applicability and robustness of the augmented reality tracking registration method in the product assembly scene, combined with the weak texture of the product assembly operation object and the complex scene, a real-time and robust unmarked tracking registration method is proposed. Firstly, the improved linear parallel multi-modal LineMod template matching method is used to detect the weakly textured assembly target base, obtain the initial pose of the camera and solve the problem of re-initialization of tracking registration interruption. Then, the ICP registration algorithm is improved to improve the registration speed while accurately positioning the camera pose, so as to determine the benchmark for virtual-real fusion. Finally, according to the pose relationship between the assembly and the assembly guidance information established in the offline stage, the visual guidance of assembly process is realized. Results Firstly, the performance of the tracking registration method proposed in this paper is verified on the public weak texture data set. Then take the motor as the assembly target to realize the visual guidance of the assembly process. Conclusions The method proposed in this paper has good real-time performance and robustness in the application scene of product assembly, which has chaotic environment and the lack of texture of operating objects