Object Tracking Using Regional Mutual Information And Edge Correlation-based Tracking Algorithm

In order to issues the problem of moving object tracking in aerial video sequences for surveillance application,This paper proposed a new target tracking algorithm after analysis of the structure of edge correlation-based(EC) tracking algorithm and regional mutual information-based(RMI) tracking algorithm. In this structure, using a novel defined crowded criterion, we are able to distinguish a crowded background and a simple background. The proposed crowded criterion determines the crowded and non-crowded backgrounds is made up of two defined intensity variation and relative power measures that are calculated from four rectangular areas around the object. Because the EC tracking algorithm cannot track the object well in the crowded frames, this article will use the RMI tracking algorithm to track the object in the crowded frames of video sequences. On the contrary, due to sensitivity of the RMI tracking algorithm to scale variations of the object and the characteristics of less computational complexity of the EC tracking algorithm, this paper selects the EC tracking algorithm for the object tracking in the non-crowded frames of the video sequences. However, although the RMI tracking algorithm is suitable for the frames with the crowded backgrounds, background brightness changes, noisy conditions, and existence of clutter, etc,it is very sensitive to target deformation. In order to solve the problem of the algorithm is slow,this paper use Powell-Golden parameter optimization method to optimize the RMI tracking algorithm.The experimental results show the advantages of the proposed tracking structure in comparison with other conventional algorithms.