At present,domestic tea picking methods are mainly manual picking,su pplemented by mechanical picking. Manual picking has low efficiency and high economic cost.Although mechanical pi cking is efficient,it will damage the buds.The artificial intelligence that has emerged in recent years and applied t o tea picking can effectively solve the problem of mechanical picking and damaged buds,one of the most important tasks is to realize automatic and efficient identification of tea buds.The purpose of this paper is to study the method of tea bud recognition based on the depth network model.According to the grade and quality requirements of tea,the tea b uds are divided into one bud,one leaf and one bud,two leaves.Because the growth posture of tea is very different,th e classification of occlusion is added to the tea bud recognition model.The Faster R-CNN deep network model based on the VGG-16,ResNet-50and ResNet- 101feature extraction network is selected to train the tea bud data samples res pectively.At the same time,the method is compared with the SSD deep network model of the three same feature extraction ne tworks.The experimental results show that the Faster R-CNN deep network model based on the VGG-16feature extractio n network has a good recognition effect. The accuracy of tea bud recognition model is 85.14%,recall rate is 78.90,and m ap is 82.17%.The depth network model can effectively identify the target of tea buds and provide technical support for intelligent tea picking.