Tea is one of the most consumed beverages in the whole world. Premium tea is a kind of tea with high nutrition, quality, and economic value. This study solves the problem of detecting premium tea buds in automatic plucking by training a modified Mask R-CNN network for tea bud detection in images. A new anchor generation method by adding additional anchors and the CIoU loss function were used in this modified model. In this study, the keypoint detection branch was optimized to locate tea bud keypoints, which, containing a fully convolutional network (FCN), is also built to locate the keypoints of bud objects. The built convolutional neural network was trained through our dataset and obtained an 86.6% precision and 88.3% recall for the bud object detection. The keypoint localization had a precision of 85.9% and a recall of 83.3%. In addition, a dataset for the tea buds and picking points was constructed in study. The experiments show that the developed model can be robust for a range of tea-bud-harvesting scenarios and introduces the possibility and theoretical basis for fully automated tea bud harvesting.

In order to achieve the recognition and detection of pear tree trunks in natural envi-ronment and solve the problem of low recognition rate caused by the complex recogni-tion environment, this study uses binocular camera to construct a visual localisation system for pear tree trunks, and proposes a pear tree trunk recognition and localisation model based on the improved YOLOv5s model. The model replaces C3 with C3TR module and embeds the CA attention mechanism, replaces the standard convolution in the Neck part with GSConv, and replaces it with the bidirectional feature pyramid network BiFPN structure; combined with the acquired depth information of the pear tree trunk, the 3D spatial coordinate information of the pear tree trunk is obtained by using the stereo imaging principle of the binocular camera. In order to verify the effec-tiveness of the proposed method, the effectiveness of YOLOv5s-pear is compared and analysed with the other three models, YOLOv4, YOLOv4-Tiny and YOLOv5s, in terms of the detection results, and the experimental results show that the accuracy of YOLOv5s-pear is improved compared with YOLOv5s, YOLOv4-Tiny and YOLOv4 by 2.1%, 6.2% and 7.7%; the positioning test shows that the error rate between the shoot-ing distance and the actual distance is 1.48%. The improved YOLOv5s model, YOLOv5s-pear, improves the model's performance in dealing with target detection in multi-scale and complex environments, and can realise the rapid identification and precise positioning of pear tree trunks, which can provide a reference for the research and development of autonomous navigation devices.