Color image segmentation of sequence images in visual inspection of OCS is studied by modifying the conventional color difference operator. Subregion process mode is adopted for fast recognition of multi-target in sequence images. A new RGB characteristic operator, combined with pixel features in target area, is proposed. Optimal segment threshold is determined by target pixel area tracking method. The experiment, done by comparing the suggested algorithm and other typical methods based on color difference in segmenting a four-target image sampled on site, proves that the color image segmentation method based on target pixel characteristic analysis can achieve the segmentation accurately and completely.