CAM Shift Object Tracking Algorithm Based on Color and Edge Feature

As well known, the continuously adaptive mean shift(CAM Shift) tracking algorithm is only adapt to the track of special object, and is apt to the effect to light and background color. In order to overcome this shortcoming, an improved CAM Shift tracking algorithm is developed in this paper, in which the target model is established via multidimensional histogram in color space and the target weight is added via edge feature. Firstly, the multidimensional colors are uniformly quantized to attain the eigenvalue, and the target model is established based on the estimate on probability density of kernel function. And then, Sobel operator is utilized to detect the feature of target edge. By combining with color feature, the target probability distribution image is determined via giving their different weights. Finally, the target is searched via Mean Shift algorithm, in which the search window size and orientation are determined by calculating moments. The experiments show that the proposed algorithm can track multicolor target and effectively attenuate the effect of disturbance including light and color.