Robust ball color auto-calibration for tracking

Abstract

In Automatic Soccer Robot Competition, ball tracking is the most important task. A common method used to track a ball in the field is by using color based method. The main problem of color based method is that the color values of the ball may be vary according to the lighting condition of that environment. The lighting problem made operator need to manually change the color range used to track the ball. This way is ineffective and unoptimal as operator need to test it several times. This paper presents an approach where the robot can automatically define the color range of the ball. The idea is to use hough circle to find the ball in the image and then get all the ball's pixel values. Then standard deviation is used to the ball's color values data to reject outliers. Then from the new data, minimum and the maximum value for the range is extracted and used in the color-based object tracking.