Target tracking algorithm base on robot vision
                                                                           WANG Zhongfei, ZHANG Pengtao
                                   （School of mechanical engineering, Zhejiang University of Technology, Hangzhou 310000, China）

Abstract: Aiming at the low matching efficiency in the current target tracking algorithm of robot vision, robot vision theory, scaleinvariant feature transformation, feature point matching, and highdimensional space vector were researched, and a nearest neighbor search algorithm based on euclidean distance and vector angle was proposed. First, the euclidean distance from all vectors to the origin in high dimensional space was computerized and sorted, the angle between all vectors and stochastic selected reference vectors in high dimensional space was computerized and sorted. Then, the euclidean distance from the query vector to the origin was calculated, and a large number of nonnearest neighbors were eliminated the retrieval range was narrowed. Finally, the angle between the query vector and the reference vector was calculated. taking this angle as the center , the nearest neighbor of the query vector was retrieced. The results indicate that this method greatly reduces the matching time and improves the accuracy of feature matching.

Key words: robot vision; target tracking; scale invariant feature transformation（SIFT）; feature point matching