Research On Video Tracking Algorithm Based On Local Representation

In the field of computer vision and image processing, visual object tracking has been developed extremely in military affairs and civil Engineering. Currently, many visual object tracking algorithms have been proposed by scholars and experts, however there are lots of factors still affect actual real-time tracking targets. For example, the accuracy of the tracking will be seriously affected when the occlusion occurs. It is necessary to propose to a robust, accurate and real-time tracking algorithm in order to solve the problem of occlusion. Therefore, this paper mainly focuses on the occlusion problem in the object tracking, the main works and innovations are as follows:1. An accurate object tracking algorithm based on local representation. In this paper, the local representation, the compressive sensing theory and the construction of classifiers are studied, and an object tracking algorithm based on local representation is improved. This algorithm extract the local overlapping image blocks, and the feature information of unoccluded local block is used to distinguish the object when partial features are occluded. At the same time, overlapping image blocks can effectively avoid noise and error caused by training samples. In addition, the compressive sensing is applied to the positive and negative sample data in order to reduce the dimension and computational complexity. We qualitatively verify the proposed algorithm on 6 set of standard video test sequences, which proves that the proposed algorithm has good performance.2. Algorithm implementation using Open CV system. The object tracking algorithm based on local representation is performed using Open CV, and it is easy to operate. Under the open CV system, shooting video and video in standard library are simulated, respectively. The experimental results show that the proposed algorithm can deal with occlusion problem better, and the effectiveness of the algorithm is verified.