Object tracking is one of the important and developing issues in image processing and machine vision. In fact, Object tracking monitors the object position variations and tracks it during a video sequence for a particular purpose. Historically object tracking has been introduced and developed in military area, but after 1980’s much of attention have been paid to it in other fields. One of the problems in the tracking algorithm is occlusion. It is very important to present the tracking algorithms that are robust when occlusion problem is encountered. One of the methods for object tracking is a mesh-based scheme. Although this algorithm is more efficient than other methods but it fails to track the desired object in the presence of occlusion. In this thesis, we employ an adaptive mesh method to increase the robustness of algorithm against occlusion. In this algorithm, mesh points are selected according to amount of variations between two successive frames. Modifying the mesh structure in successive frames (and not using a fixed one for all the frames) helps the mesh to adapt itself with entrance and exit of different objects and also with occlusion through successive frames. The most important parts of the algorithm are implementation of the polygon approximation, which is an important approximation in the object tracking, modification the selection of mesh points, implementation of a method of triangulation and proposal of a simple, fast method for triangulation. We present all above-mentioned modifications and proposals to overcome limitations of mesh tracking algorithm.
