Abstract: To grade fruit surface color by automatic computer vision detecting system,　an indoor computer vision　system　was set up.　By using this system the RGB values of apple hues were converted into HLS coordinate， and the cluster hues distribution curves can be determined by experiments． Because there are four fruit overlapping hue distribution curves, it is difficult to be used for grading. Otherwise in the frequency picture of cluster hues distribution, different hues are separated in the region between 0～80°, we can grade color according to the percentage of clustering color area under appropriate hues. The experiments with RedStar apple revealed that the way can improve the average consistence between computer vision grading and manual grading to over 88 %.