Research on Target Tracking Algorithm Based on Kernel Correlation

With the development of sensor and image processing technology, computer vision plays an increasingly significant role in the chemical engineering because of its characteristics such as low cost, high resolution, and non-contact measurement. In this paper, the motion probability map can be obtained by sparse optical flow based on Harris corner point. Then the coarse contour of silicon dioxide particles which is the input of kernelized correlation filtering (KCF) algorithm can be generalized. KCF algorithm can easily complete tracking task under the influence of disturbance including light change, video shaking and so forth. A contour refining and tracking method are proposed. The geometric active contour (GAC) algorithm can use function as implicit expression of contour and can design the different energy functional to control contour evolution. By minimizing of energy functional, the refining contour is evolved. Then the target tracking is realized according to the refined contour.