Research Onadaptive Multi-cue Integrition Method Of Particle Filter Based Target Tracking

The visual object tracking is a key issue in many vision-based applications, such as visual surveillance, human-computer interaction, visual navigation of robots, military guidance and so on. With the rapid growth of the information techniques in the last tens of years, the object tracking has attracted many researchers'attentions and has become a very popular research topic. Although many effective visual object tracking methods have been proposed, there are still a lot of difficulties in designing a robust tracking algorithm due to the challenging complex scenarios such as significant illumination changes in environment, pose variations of the object and non-linear deformations of shapes, and noise and dense clutters in complex background, etc. Therefore, robust design of video object tracking algorithm is still a challenging task.In this paper, we used the framework of particle filter, focused on the target appearance modeling, and proposed an adaptive multi-cue integration target observation model representation. We proposed a weight adaptively updating algorithm, and a particle filter (PF) is designed for object tracking based on multiple cues with adaptive parameters. The main contributions of this dissertation are summarized as follows:To adaptively adjust the weight of each feature, by analyzing particles'distribution, this paper proposed a frame-by-frame weight adaptively updating algorithm. The method based on particle distribution can weigh the reliability of each feature and update the weights. To a large extent, the method is able to adapt to the complexity of the tracking environment, and ensures the accuracy of tracking. But it is easily affected by errors.Basing on the consideration of weight's time-series consistency, this paper proposed a weight tracking method. By using particle filter to track the feature weight and combining with the frame by frame adjustment algorithm, we proposed double-particle filter based multi-cue integration tracking method. This method not only achieves that the features weights can adjust frame by frame according to the actual situation, but also avoids the weight mutation caused by the tracking error, which ensures the tracking results stable and reliable. The proposed weight updating algorithm and tracking method provide a stable and reliable basis for the accuracy and robustness of tracking algorithm. Experimental results on videos with various tracking conditions show the significant improvements of the proposed methods, comparing with the existing integration algorithms.