A Method to Search for Color Segmentation Threshold in Traffic Sign Detection

Color segmentation is the most commonly used method in traffic sign detection. Therefore we propose a method to search for color segmentation threshold in video. Firstly, we classify every image's pixels into positive samples and negative samples after color segmentation using some experience thresholds. Secondly, all positive samples are assembled into a new collection as ¿positive-sample-set¿, and all negative samples are assembled into ¿negative-sample-set¿. Thirdly, the difference between positive-sample-set and negative-sample-set both in RGB and HSV color space are gathered statistics. Then better thresholds for color segmentation could be found. Experiments show the color thresholds are more accurate and appropriate to detect traffic signs. And the method could also be used in other color analysis research.