Real-time Tracking Based on Compression Sensing of Multiple Features

Traditional compression sensing tracking algorithms use the gray feature of images to describe the tracking target, which not only cause a major fluctuations of the classification result of the classifier, but also give rise to the accumulation of classification error, so that the tracking target may be lost. In order to improve the robustness of traditional compression sensing tracking algorithm, this paper introduces the differential feature information of images, and describes tracking target with multiple features. This method improves the positioning precision of the target, makes up the instability and inaccuracy caused by single feature tracking method, constructs the feature weighting classifier, improves the tracking stability and accuracy, meets the real-time requirements and has higher practicability.