A novel deep learning model based on Yolov2 for green tea stalk detection made by machine-plucking fresh leaves

In summer and autumn, the color difference between green tea leaves processed by machine and tea stems is small, and it is difficult to recognize tea stalks through color selection. Present research proposed a method based on Yolov2 for tea stalk detection. By collecting data images of green tea stalk detection made by machine-plucking fresh leaves, image enhancement and imported into Yolov2 model for tea stalks detection; in order to achieve a rapid response of stalk picking in the process of tea refining. The proposed model consists of two components. The first component is based on the Yolov2 model of Matlab to realize transfer learning. In contrast, the second component is designed to detect green tea stalks based on Yolov2. The average IoU has been used to estimate the optimal number of anchor boxes and improve the detection performance. The test results showed that the average accuracy of Adam function model was 99.43%, the recall rate was 99.68%, which is better than Sgdm. It is proved that the model has high recognition accuracy for green tea stalk images made by machine-plucking fresh leaves, and is of great significance for promoting the development of intelligent stalk picking equipment.