The image segmentation algorithm based on the fuzzy C-average clustering (FCM) needs initial cluster number and cluster center in advance, which make the algorithm easy to fall into local optimum. An image segmentation method based on improved artificial swarm optimization fuzzy clustering was proposed. The optimization of proposed method was conducted on the basis of the traditional artificial colony. The fitness function of artificial colony was improved by using objective function of FCM algorithm. With the collaboration of bee colony, follow bees and computerized bee, the optimal initial clustering center could be solved quickly. Then the optimal initial clustering center was input into FCM and image segmentation was finally realized by using maximum membership principle. The fruit segmentation experiment was carried out with 300 ‘summer black’ grape photos taken under frontlight, backlight and normal light illumination conditions. The experiment proves that the proposed method can identify fruit from the natural environment quickly. The average time for segmentation was 0.2193s per photo and accuracy was 90.33%. The time consuming was shorter and the accuracy was higher than OTSU and traditional FCM algorithm. It can meet the real-time requirement of picking robot and fruit grading system.