A Novel Motion Recognition and Tracking Algorithm based on Multi Example Learning and Adaptive Neural Network

By consulting a large number of domestic and foreign references, conducting in-depth research on the relevant knowledge of target tracking, understanding the working principle and derivation process of particle filtering. On the basis of studying traditional machine learning algorithms (including boosting algorithm, online boosting algorithm, Adaboost algorithm, MIL algorithm, etc.) and researching the application of sparse knowledge in the field of target tracking, I have gained an understanding of the basic principles and derivation process of MIL algorithm and sparse representation successfully applied in the direction of target tracking. Therefore, this paper proposes two improved schemes based on the MIL algorithm, namely PFWMIL and COMIL. To demonstrate the effectiveness of the proposed algorithm, experimental results of multiple image sequences were presented, and the experimental results of different tracking algorithms were compared. The results demonstrate that the target tracking algorithm proposed in this paper is feasible and robust.