For an apple harvesting robot, the machine vision system is a very important part which is used to recognize and locate apples. In this paper, an automatic apple recognition method based on image processing technology is developed for apple harvesting robot. Firstly, 2R-G-B grey value transform is performed, and the adaptive threshold Otsu algorithm is used for image segmentation. Then, a series of morphological operations are adopted to eliminate image noise, and edge detection is utilized to obtain the apple edges. Finally, the optimized Hough transform is presented to recognize apples. Experimental results indicate that the proposed method can segment the adjacent and overlapped apples effectively, where the recognition rate can reach 91 %. Meanwhile, the centers and radii of the apples could also be accurately extracted.