Multiple Kernels Precise Tracking

In this paper, a novel multiple kernel-based tracking approach with color and gradient features was proposed. Gradient and color histograms with multiple kernel-based tracking were integrated. The integration of color and gradients made tracking more robust to changes in luminance and avoided tracking algorithm drift under low illumination. Also, a new constraint derived from affine transformation was introduced. This addition enabled the tracker to provide the ability to process complex motions such as rotation, scale changes, and viewpoint changes etc. Experiments show that the proposed approach is effective and robust to complex motion and to cluttered environments.