Target Tracking Algorithm Intelligent Mobile Robot Mean Shift

Moving target tracking is one of the hot spots in the intelligent mobile robot research domain, which has a widespread application prospect. This paper deeply studies, improves and simulations experiments the target tracking algorithm based on Mean-Shift theory. In addition, taking the NWU-RR-II intelligent mobile robot as platform, it verifies the correctness and effectiveness of the modified algorithm. The primary research and work are summarized as follows:Firstly, through in-depth learning principle of the traditional Mean-Shift tracking algorithm, this paper carries out a series of experiments in the software of MATLAB7.0. The results show that the algorithm has many advantages such as simple in principle, very good real-time, however, in cluttered environment it has a few weaknesses, which are fixed kernel-bandwidth, the lack of prediction mechanism, unable to solve occlusion and so on.Secondly, to the object scale change, a modified method is proposed that is multi-scale space theory combined with particle filter. The information in the tracking window, which is calculated by the multi-scale space theory, is predicted and modified by the particle filter. The correction is introduced to get the proportion of the object area and then the tracking window size is adjusted dynamically. The experimental results show that the improved algorithm could track the target more precisely; Aiming at the occlusion problem, the tracking process is divided into normal condition, partial and full occlusion according to the change of similarity coefficient. The target position is determined through the mechanism which is the combination of Mean-Shift algorithm with Kalman filter and the strategy of finding object. Experiments have suggested that the improved algorithm has better robust for occlusion.Finally, this paper focuses on how to control the mobile robot to track the target in many different contexts. The system program is written with LabVIEW software. The results indicate that the improved algorithm can satisfy the robust and real-time requirement of the robot moving target tracking.