Scale Adaptive Target Tracking Method Based on Fast Compressive Tracking Algorithm

Abstract

The invention discloses a scale adaptive target tracking method based on a fast compression tracking algorithm. Firstly, the present invention adopts the context model to weight the Haar-like features, which enhances the robustness of the Haar-like features to illumination changes. Secondly, the step-by-step tracking is adopted, which not only enhances the anti-occlusion ability of the algorithm and the ability to deal with the change of target scale, but also maintains the real-time performance of the algorithm. Finally, a scale-adaptive method is proposed to achieve stable tracking of scale-varying targets. The present invention has good robustness to the change of the target scale, the change of the target appearance and the situation that the target is blocked, and can guarantee the frame rate at about 39 frames per second, which meets the requirement of real-time performance.

Description

technical field [0001] The invention belongs to the technical field of image processing, and in particular relates to video target tracking, face recognition, online learning and scale self-adaptation, etc., in particular to a scale-adaptive target tracking method based on a fast compression tracking algorithm. Background technique [0002] Object tracking is a research hotspot in the field of computer vision, and it has a wide range of applications in motion analysis, behavior recognition, intelligent monitoring, human-computer interaction and other fields [1,2]. The difficulty of target tracking is how to deal with the changes in the appearance of the target itself and the influence of factors such as illumination, occlusion, and background changes on the target [3,4]. In recent years, the target tracking algorithm based on online learning has received extensive attention. This algorithm regards the tracking problem as a special binary classification problem. The key is to...

AI Technical Summary

Problems solved by technology

But it mainly has two problems: First, the feature description is simple, and it is prone to tracking drift or target loss when the illumination changes or the appearance of the target changes greatly.

Second, the scale of the target window is fixed during the tracking process. When the target scale becomes larger or occluded, it is easy to cause tracking drift or target loss.

Literature [11] is the original author's improvement of the CT algorithm, but it only improves the processing speed of the algorithm, and does not solve the above problems well

Images