In order to improve the intelligent level of grape bagging and reduce the amount of human labor, a robot system for grape bagging was developed. The hardware structure of robot system was designed based on cylindrical coordinate. A method was proposed in order to get target feature extraction and recognition method by genetic algorithm. Compared with the proposed algorithm, the segmentation method using morphology algorithm and BP neural network under the HSI color space needed more time. The result of experiments showed that when the velocity of platform was 0.3m/s, the recognition rate for grape was up to 95% and the average time-consuming was about 136ms. The grape bagging robot success rate was 85% and the average time-consuming for a cluster of grapes was 39.6s.