| With the development of the video analysis technology, video monitoring system has been gradually from the simple video image collection and storage system development to the intelligent video monitoring system. It has been applied in many industries. Because of different video scenes have different characteristics, there is very few video analysis algorithms can adapt to all of the scenes.This topic mainly studies the scene for indoor, indoor scenes with a light source complex, the target is easy to block the overlap of the shortcomings. Target detection and tracking algorithm in the traditional indoor scene under the prospect of the detection and tracking of the target frame mark is not stable, flicker, division, and many people lost marks for the same goal and effect of anti background occlusion is poor and other issues, so in the scene requires high tracking stability, its application has been great limit. In order to solve the above problem in practical application project, this paper mainly studies the stability detection and effective tracking algorithm of multiple targets in the indoor surveillance scene, focusing on the scene features of the optimization algorithm, to detect the target by using the gradient histogram, and then using TLD based tracking algorithm, aiming at the process of tracking errors reasonable correction, fusion SIFT corner feature and color feature of auxiliary to achieve stable tracking. In the process of multi- target tracking, a reasonable computer resource scheduling scheme is designed, which can track multiple targets simultaneously. The algorithm proposed in this paper with the city museum actual intelligent video analysis tracking system testing, has achieved good effect of detection and tracking, experimental tests show that the algorithm structure module proposed in this paper can be conveniently docking fusion with other video analysis algorithm, which laid the foundation for the realization of more complex intelligent video analysis system. |
PDF Full Text Request