Accurately obtaining the posture and spatial position of tea buds through machine vision and other technologies is difficult due to the small size, different shapes, and complex growth environment of tea buds. Therefore, end effectors are prone to problems, such as picking omission and picking error. This study designs a picking end effector based on negative pressure guidance for famous tea. This end effector uses negative pressure to guide tea buds in a top-down manner, thereby correcting their posture and spatial position. Therefore, the designed end effector has deviation tolerance performance that can improve the picking success rate. The pre-experiment is designed, the tip of apical bud is referred to as the descent position, and the negative pressure range is determined to be 0.6 to 0.9 kPa. A deviation tolerance orthogonal experiment is designed. Experimental results show that various experimental factors are ranked in terms of the significance level of the effect on the average success rate, and the significance ranking is as follows: negative pressure (P) > pipe diameter (D) > descent speed (V). An evaluation method of deviation tolerance performance is presented, and the optimal experiment factor-level combination is determined as: P = 0.9 kPa, D = 34 mm, V = 20 mm/s. Within the deviation range of a 10 mm radius, the average success rate of the negative pressure guidance of the end effector is 97.36%. The designed end effector can be applied to the intelligent picking of famous tea. This study can provide a reference for the design of similar picking end effectors for famous tea.

To address the problem of a lack of accurate parameters in the discrete element simulation study of the machine-picked fresh tea leaf mechanized-sorting process, this study used machine-picked fresh tea leaves as the research object, established discrete element models of different fresh tea leaf components in EDEM software version 7.0.0. based on the bonded particle model using three-dimensional scanning inverse-modeling technology, and calibrated the simulation parameters through physical tests and virtual simulation tests. Firstly, the intrinsic parameters of machine-picked tea leaves were measured using physical tests; the physical-stacking tea leaf test was conducted using the cylinder lifting method, the tea leaf repose angle being 32.62° as measured from the stacking images using CAD. With the physical repose angle as the target value, the Plackeet–Burman test, the steepest-ascent test and the Box–Behnken optimization test were conducted in turn, and the results showed that the static friction coefficient between tea leaves, the rolling friction coefficient between tea leaves and the static friction coefficient between tea leaves and PVC have a major effect on the repose angle, and the optimal combination of the three significant parameters was determined. Finally, five simulations were conducted using the optimal combination of parameters, the relative error between the repose angle measured by the simulation test and the physical repose angle being just 0.28%. Moreover, the t-test obtained p > 0.05, indicating that there was no significant difference between the simulation test results and the physical test results. The results showed that the calibrated discrete element simulation parameters obtained could provide a reference for the discrete element simulation study of fresh tea leaves.

The hyperspectral remote sensing images of agricultural crops contain rich spectral information, which can provide important details about crop growth status, diseases, and pests. However, existing crop classification methods face several key limitations when processing hyperspectral remote sensing images, primarily in the following aspects. First, the complex background in the images. Various elements in the background may have similar spectral characteristics to the crops, and this spectral similarity makes the classification model susceptible to background interference, thus reducing classification accuracy. Second, the differences in crop scales increase the difficulty of feature extraction. In different image regions, the scale of crops can vary significantly, and traditional classification methods often struggle to effectively capture this information. Additionally, due to the limitations of spectral information, especially under multi-scale variation backgrounds, the extraction of crop information becomes even more challenging, leading to instability in the classification results. To address these issues, a semantic-guided transformer network (SGTN) is proposed, which aims to effectively overcome the limitations of these deep learning methods and improve crop classification accuracy and robustness. First, a multi-scale spatial–spectral information extraction (MSIE) module is designed that effectively handle the variations of crops at different scales in the image, thereby extracting richer and more accurate features, and reducing the impact of scale changes. Second, a semantic-guided attention (SGA) module is proposed, which enhances the model’s sensitivity to crop semantic information, further reducing background interference and improving the accuracy of crop area recognition. By combining the MSIE and SGA modules, the SGTN can focus on the semantic features of crops at multiple scales, thus generating more accurate classification results. Finally, a two-stage feature extraction structure is employed to further optimize the extraction of crop semantic features and enhance classification accuracy. The results show that on the Indian Pines, Pavia University, and Salinas benchmark datasets, the overall accuracies of the proposed model are 98.24%, 98.34%, and 97.89%, respectively. Compared with other methods, the model achieves better classification accuracy and generalization performance. In the future, the SGTN is expected to be applied to more agricultural remote sensing tasks, such as crop disease detection and yield prediction, providing more reliable technical support for precision agriculture and agricultural monitoring.