2cm。An apple recognition method with normalized color difference (R-G)/(R+G) was presented. The apples in the images taken under different lighting conditions were recognized. Some pretreatments were applied to get contour images. The shape features of the apples were extracted from the contour images based on the method of random circle ring. A matching algorithm based on area feature and epipolar geometry was discussed to position the apples. The apples with similar areas were matched according to the principle of left-right order consistency by calculating the maximum value of cross-correlation function of vertical projections. The experiments results show that the proposed recognition method can eliminate influence of shade, and soil, to a certain extent. The recognition rate reaches 92%. The circle centers and radius are extracted precisely and rapidly based on the method of random circle ring. The errors are less than 2cm when the measure distance is between 60cm and 150cm. 