Method and Implementation of Target Tracking Based on Correlation Filter

摘要

This paper focuses on robust target tracking under complex background. Base on correlation filter, dual filter is built by adding scale estimation to Minimum Output Sum of Square Error (MOSSE) where target transition and scale change can be estimated simultaneously. Meanwhile, Histogram of Gradient is used as feature descriptor to achieve stable tracking. Finally, the improved algorithm is transported and optimized to TI DSP multi-core platform. In the end, both quantitative and qualitative experiments are conducted to demonstrate the robustness and real-time performance of the method.