Research on Mixing of Virtual and Actual Reality for Motor Augmented Assembly

Abstract

The effect of the mixing of virtual and actual reality will directly affect the realism and immersion of the augmented assembly system. The correct occlusion relationship and collision interaction between virtual and real objects are crucial to the fusion effect. The existing virtual-real occlusion handling methods cannot be applied to the augmented assembly system with complex contours and changeable scenes. Therefore, an occlusion handling method based on "virtual avatar" of motor equipment is proposed, which has strong robustness and real-time performance. Aiming at the problem that the number of wrong matching pairs of feature points is too large in the 3D registration process, we combine the method of Cosine similarity measurement on the basis of ORB algorithm to improve the accuracy of feature point matching. The current virtual and real object collision detection methods have large computation and low detection accuracy. We propose an octree hierarchical collision detection optimization algorithm based on model surface voxelization representation and volume difference, which improves the detection accuracy on the basis of ensuring computational efficiency. The experimental results show that the proposed methods have high real-time performance and accuracy, and greatly improve the realism and immersion of the augmented assembly system.