关键词: 增强装配, 跟踪注册, 图像匹配, 遗传算法, 差分进化, 禁忌搜索

Abstract: To improve the robustness and applicability of tracking and registration method for textureless parts in augmented reality assembly guidance system,a tracking and registration method based on image matching was proposed.Edge orientation was used to weight the pixel value of distance image to improve the calculation accuracy of image matching degree.Accurate camera pose of initial frame was premise to ensure the tracking and registration accuracy.The differential evolution strategy and the tabu search strategy were used to modify the genetic algorithm,which improved the accuracy and search efficiency for tracking the camera pose.To ensure the accuracy and real-time of tracking and registration,tabu search algorithm was used to continuously track camera pose.Experimental results showed that the proposed method had better precision than ORB-based method in textureless parts tracking and registration,and had better robustness to messy environment,which could meet the real-time requirements of tracking and registration.

Key words: augmented assembly, tracking and registration, image matching, genetic algorithms, differential evolution, tabu search