| As one of the key technologies of augmented reality,3D registration and tracking technology,its principles and implementation methods are important factors that affect the fusion of virtual reality and human-computer interaction experience.With the development of machine vision theory and the popularization of smart mobile devices,non-marking augmented reality applications running on smart mobile device platforms will become the mainstream in the future.Accordingly,in the complex environment,robustness and real-time put forward higher requirements for 3D registration tracking method.Aiming at the robustness and real-time problems mentioned above,an improved orb feature matching algorithm based on quadtree decomposition and RANSAC algorithm is proposed.The quadtree decomposition method based on the idea of region splitting and merging is applied to simplify the image retrieval domain of fast algorithm,which not only eliminates the regions whose gray changes are not obvious,but also introduces the prior knowledge of different fields.At the same time,the prior knowledge of different fields can be introduced.The image pyramid and gray-centroid method are added to enhance the scale invariance and rotation consistency of the algorithm itself.In the BRIEF description,256-bit binary characters are selected as feature descriptors based on the principle of least correlation,and Hamming distance is used for matching.Finally,the minimum block after the region splitting and merging is taken as the basic unit to eliminate the RANSAC mismatch and improve the matching accuracy.The experimental results show that the effective matching accuracy of the improved orb algorithm is 94.2%.On the premise of ensuring a good speed,it also improves the robustness and real-time performance of the unmarked 3D registration tracking method in complex environment.Because of no need to rely on special sensors,it has certain practicability in the field of augmented reality.Figure 44;Table 2;Reference 55... |
PDF Full Text Request