Abstract: A combined algorithm for image recognition and boundary segmentation of overlapped apple fruits under natural environments was proposed. Firstly, the algorithm selects the mature apple images that are shaded by branches and leaves or overlapped with each other and preprocess the images. Then, a component of lab space, U component of YUV space, and V component of YUV space were extracted as image segmentation. The binary image was obtained by using the improved multi-channel global threshold segmentation algorithm. Finally, the Hough transform and Circular constraints are used to identify the individual apple target. The experimental results showed that the proposed algorithm could recognize the overlapped objectives under natural environments. In terms of the segmentation algorithm, the average accuracy was 95.5%, and the average false positive rate was 2.1%. It ran 8.9 times faster than Kmeans iterative algorithm. It had high precision and fast speed comparing with the direct Otsu and Kmeans iterative algorithm. In the aspect of circle recognition, the circle constraint method can eliminate the over-segmented circle after the Hough transform algorithm and accurately locate the target.