In order to meet the speed requirements of harvesting robot, a method of tracking and recognition of overlapping apples is proposed in this paper. First of all, the first image should be segmented and denoised, the center and the radius is determined, the template which is used for matching is extracted according to the center and radius. Then, determine the center of ten images which are taken continuously, fit motion path of the robot according to the center of each image and predict subsequent motion path. The next processing area is determined according to the radius and predicted path. Finally, overlapping apples are identified by fast normalized cross correlation match method. Experiments prove that the new method can locate the center and radius of overlapping apples correctly. Besides, matching time is reduced by 48.1 % compared with the original one.