A tea tip detection method suitable for tea pickers based on YOLOv4 network

In order to realize the high-efficiency and high-precision automatic collection of famous and high-quality teas, a tea tip detection method based on YOLOv4 network suitable for tea pickers is proposed. First of all, this method innovatively proposes a segmented tea picker structure, and uses this structure for image acquisition, which greatly reduces the occlusion of leaves and solves the problem that the tea tip is close to the background color and is difficult to separate. Afterwards, the improved YOLOv4 network was used to predict the bud tip. Finally, the prediction frame of the bud tip is obtained, and the HSV image segmentation and convex hull detection algorithm are used to calculate the picking points. The average accuracy of the training model is 93.08%, the accuracy is 93.24 %, the recall rate is 86.94%, and the F -measure is 90.00%. The result is higher than the general similar detection methods. The algorithm has sufficient recognition accuracy and provides new technical ideas for intelligent bud picking.