Human Motion Tracking Based on Adaptive Template Matching and GM(1,1)

To solve the defect of poor robustness and real- timeliness of traditional correlation matching algorithm, this paper proposes a novel method that combines normalized autocorrelation matching algorithm, template updating strategy and grey forecasting model GM(1,1) with priority of new information. The real-time updating of the template image size improves the deformation resistance capability of the autocorrelation matching algorithm, and assignment of great weight to new information improves the robustness of the grey model. The use of forecasting increases the speed of the adaptive template and improves the real-timeliness of the algorithm. Meanwhile, the results of template matching provide the GM(1,1) model with new data for forecasting the target position of the next frame of image. When human motion region is shaded, the forecast value is used to replace the real value to continue the motion and thus to improve the real-timeliness and robustness of the system.