Athlete Number Localization Based on Edge Detection and Color Features

Aimed at the problem for low accuracy of athlete number positioning under the complicated background, an algorithm of localization which based on edge detection and color features was presented. First, the edge detection method on the basic of the character features of athlete number was applied to the athlete number image. Then, the binary image processing was carried on by selecting an optimal threshold, which could make the number region stand out. After that, the connected boundary tracking method was used to remove the image noises. Finally, the athlete number region was positioned precisely based on the color features of black character on white background in number area. From the results of experiment, it was found that the algorithm proposed can accurately locate the athlete number. It also had an ability of positioning multiple athletes´ numbers in one image.