81%。 An apple recognition method based on K-means algorithm is proposed for the green apples that have similar color with leaves. The image is divided into 8×8 pixel blocks and the block is taken as the segmentation unit by the algorithm. Color difference R-B was selected as color feature and mean value, standard deviation and regional entropy of gray scale images are selected as texture features. The feature vectors including color feature and texture features are extracted. Gap statistic is applied to calculate the best number of the clusters. The recognition experiment is conducted to test the algorithm with 200 sample images taken in different illumination conditions. The experimental results show that the apple fruits can be recognized successfully both in front light conditions and back light conditions. The recognition rate reaches 81%.