Development of Target Person Detecting and Tracking Method for Mobile Robots Based on Gradual Updating Multi Color HS-histogram

Abstract

This paper presents that the novel target person detecting and tracking method for mobile robots based on updating multi color HS-histogram. To detect and track a specific person, the image processing with color camera is common method for vision based navigation system of mobile robot. The image processing for the target person detecting and tracking by robot vision has many problems such as the image from the vision device located on robot is often influenced by the change of target size, target moving direction and lighting condition. Moreover, it is difficult to detect a target person in the crowd. To solve these problems, we propose “GRUMIC-HS-hist method (Gradual Updating Multi Color HS-histogram method)”. In the proposed method, the colors of the tracking object are extracted by HSV color space. To decrease the influence of the illumination changes, a HS-histogram is proposed. The extracting colors on HS-histogram are clustered by Nearest Neighbor Method. The colors on HS-histogram are weighted by the anisotropic asymmetric Gaussian distribution to update the HS-histogram without spreading. This proposed “GRUMIC-HS-hist method” achieves the robust target person detecting and tracking by a mobile robot. In addition, the quantitative evaluation method is also proposed to evaluate the proposed target person detecting and tracking method. The quantitative evaluation method is composed with “Extraction Rate”, “Object Detecting Rate” and “Tracking Error”. The quantitatively experimental results show that the proposed “GRUMIC-HS-hist method ” is the robust method for many problems.