The application of agricultural robot is the inevitable trend of the development of intelligent agriculture. The main difficulties in the application of fruit picking robots are the difficulties of fruit recognition and location caused by fruit occlusion and uneven light. In order to solve the problem of inaccurate fruit recognitionin tomato picking robot, an image processing algorithm based on geometric morphology and iterative random circle was proposed, which can effectively segment and recognize the adhesive fruit in the image. Firstly, taking the tomato called Jiaxina as the research object, digital RGB camera was used to collect image. Then the image was preprocessed through Canny edge detection to obtain the fruit edge contour points. After obtaining the fruit edge contour points, the fruit contour points were obtained through geometric morphology processing, which can filter out nonfruit contour points. Finally, the fruit contour points were grouped and processed by iteration random circle, and the fruit recognition contour was obtained. The correct rate and accuracy rate of the algorithm was calculated after applying the proposed method on 80 images with 302 fruits. The results showed that the correct rate of fruit recognition was 85.1%, and the accuracy rate of fruit recognition was 79.1%. It was showed that the algorithm solved the problem of fruit segmentation in complex environments where multiple fruits were adhered or occluded by a small amount.