Fuente: WIPO "processing tomato" The invention discloses a tomato segmentation method based on an improved YOLOv8 model, and the method mainly comprises the steps: searching a tomato segmentation data set with rich categories, carrying out the data enhancement processing and manual marking of the tomato segmentation data set, expanding the number of the data set, improving a YOLOv8n model, replacing a trunk with MobileNetv3, enabling the model to be light, and carrying out the manual marking of the data set. DCNv2 variable convolution is introduced into a YOLOv8n network head, so that the shape of a target can be better adapted, and a channel attention mechanism is added to the head to pay attention to channel information and space information of a feature layer. The result shows that the average precision of the improved algorithm category is 91.8%, and the method meets the practical requirements in the aspects of recognition accuracy and speed, and further promotes the development of tomato picking robots and intelligent agriculture.