چكيده به لاتين
Target tracking is one of the most important issues in the field of image processing and machine vision. Today, various applications of industrial, commercial, military, etc are widely used in various techniques of video tracking. Despite of many advances made in recent decades in this area, there are still many challenges in this field. One of the major challenges in the field of video tracking is the need for accurate modeling of target object behavior. Various methods have been developed to analyze the behavior of the objectives. Among these approaches, probabilistic approaches can well model the behavior of a system model.
One of the most popular algorithms available in using probabilistic approach is the particle filter algorithm. This algorithm is very efficient in dealing with nonlinear and non-Gaussian systems. Despite of superiority of the particle filter algorithm in comparison with the other statistical algorithms, the need for high processing volumes in this algorithm has made it unusable in some applications. Accordingly, in this thesis, by presenting a suitable strategy and making changes to the conventional algorithm, the required computational burden has been reduced, so that the proposed algorithm can be implemented in low cost and low power processors. The result of the implementation show that the proposed method improved at least 33% and up to 44% compared with the basic method, and at least 5.08% up to 10.67% compared with the recent approach.