The mechanization of famous tea harvesting is an essential way to develop China’s tea industry. This paper centers on the detection and positioning technologies in famous tea harvesting, systematically reviewing research progress in these domains. In tea detection, traditional methods rely on color space selection and image segmentation, exhibiting limitations such as insufficient accuracy and poor generalization capability. Conversely, deep learning algorithms demonstrate superior detection accuracy and robustness. Current research focuses on enhancing detection accuracy, inference speed, and multi-variety recognition. In picking positioning, depth information measurement technology utilizing RGB-D cameras provides foundational support. Positioning methods have evolved from traditional visual processing techniques to deep learning and point cloud approaches, seeking to overcome challenges including occlusion and irregular growth patterns. Notwithstanding notable technological advancements, existing methods confront three primary limitations: difficulties in adapting to diverse growth stage characteristics, reliance on large-scale annotated datasets, and inadequate occlusion handling. Future research ought to concentrate on three directions: developing highly universal tea bud detection models, refining model training techniques for small-sample scenarios, and improving tea-picking point positioning accuracy under occluded conditions. This review aims to furnish critical references for advancing high-end intelligent tea-picking machinery, thereby facilitating the tea industry’s mechanization and intelligentization.

Tea shoot segmentation is crucial for the automation of high-quality tea plucking. However, accurate segmentation of tea shoots in unstructured and complex environments presents significant challenges due to the small size of the targets and the similarity in color between the shoots and their background. To address these challenges and achieve accurate recognition of tea shoots in complex settings, an advanced tea shoot segmentation network model is proposed based on You Only Look Once version 8 segmentation (YOLOv8-seg) network model. Firstly, to enhance the model’s segmentation capability for small targets, this study designed a feature fusion network that incorporates shallow, large-scale features extracted by the backbone network. Subsequently, the features extracted at different scales by the backbone network are fused to obtain both global and local features, thereby enhancing the overall information representation capability of the features. Furthermore, the Efficient Channel Attention mechanism was integrated into the feature fusion process and combined with a reparameterization technique to refine and improve the efficiency of the fusion process. Finally, Wise-IoU with a dynamic non-monotonic aggregation mechanism was employed to assign varying gradient gains to anchor boxes of differing qualities. Experimental results demonstrate that the improved network model increases the AP50 of box and mask by 4.33% and 4.55%, respectively, while maintaining a smaller parameter count and reduced computational demand. Compared to other classical segmentation algorithms models, the proposed model excels in tea shoot segmentation. Overall, the advancements proposed in this study effectively segment tea shoots in complex environments, offering significant theoretical and practical contributions to the automated plucking of high-quality tea.

Corn is one of the most important grain crops with a wide range of planting areas in recent years. The planting mode of corn is divided mainly into the field direct seeding and the pot seedling transplanting. Among them, the corn pot seedling transplanting is one of the dryland transplanting ways to obtain a high yield. However, manual transplanting cannot fully meet the large-scale requirement of the planting quality, due to the high cost, high labor intensity, and low efficiency. Fortunately, mechanized transplanting can be expected to reduce the labor intensity for the high yield. Therefore, there is an urgent need for mechanized equipment for corn pot seedling transplanting in the corn industry. In this study, a clipping-stem, non-circular gear and five-bar transplanting mechanism was designed for the corn pot seedling, in order to realize the integrated transplanting trajectory of corn pot seedling picking and planting. Taking the corn pot seedling as the research object, a systematic analysis was performed on the physical and mechanical properties of pot seedlings, including the moisture content, physical morphological characteristics, the force of picking seedlings, and the force of breaking the stem. The kinematics and mathematical models were established to clarify the working principle of high accuracy, four poses, non-circular gear and five-bar mechanism. The optimization target of the transplanting mechanism was determined to combine the physical and mechanical properties of corn pot seedlings with the agronomic requirements of transplanting. The auxiliary software of optimization design was also compiled using Matlab platform. A series of mechanism parameters were optimized to fully meet the target values, including the bar length of the mechanism, the non-circular gear transmission ratio, the gear pitch curve, and the theoretical analysis trajectory. The simulation experiment of the virtual prototype was then carried out in Adams software using a three-dimensional model of the mechanism. The simulation of static trajectory was basically consistent with the theoretical analysis. The idle experiment was also performed on the transplanting and bench test. The transplanting experiment was carried out on a self-designed experiment bench, where the transplanting frequencies were 40, 50, and 60 plants/min, respectively. A tray of corn pot seedlings was used in each group of experiments. The results showed that there was a relatively consistent performance in the actual trajectory of the transplanting mechanism, the simulation, and the theoretical. Furthermore, the success rate of picking seedlings decreased, whereas, the lodging rate and the injury rate of seeding increased, and the final qualified rate of planting decreased with the increase of transplanting frequency. Specifically, the success rate of picking seedlings was above 95%, while the lodging rate of seedlings was within 4%, and the injury rate of seedlings was within 2% under the transplanting frequency of 40-60 plants/min. In the bench experiment, the success rate of picking seeding was 95.04%, the lodging rate of seeding was 3.31%, the injury rate of seeding was 1.65%, and the qualified rate of planting was 90.08% when the transplanting plant spacing was 300 mm and the frequency was 60 plants/min. The optimization fully met the harsh requirements of dryland planting machinery. A better transplanting performance was achieved with the optimal seed spacing of transplanting and the high uprightness for the corn pot seedling. The feasibility of the actual transplanting mechanism was also verified to ensure the correctness of the comprehensive theoretical design. The finding can provide an important reference to optimize the integrated transplanting machine for corn pot seedling picking and planting.

Three-dimensional elevated strawberry cultivation can effectively avoid soil-borne diseases and pests to decline the soil fertility under climate change, compared with the conventional soil-based ones. However, the low planting density and uneven light distribution have limited the growth, development, and final yield of strawberries. In this study, a rotational strawberry cultivation device was developed with self-supplementary lighting, in order to improve the advantages of the conventional three-dimensional cultivation. A planting experiment was conducted to compare this device and the conventional "H"-shaped elevated strawberry cultivation frame. Firstly, the physical parameters of strawberry plants were measured to determine the structural dimensions of the planting trough frame. The feasible solution domain was calculated for the radius, number of troughs, and angle of the cultivation frame under a specific planting environment. A radius of 0.85 m, 10 troughs, and an angle of 36° were selected as the basic dimensional parameters. The main structure of the rotating disk was designed to fully meet the requirements of the rotational irrigation and split-mounted planting troughs. Meanwhile, a fan-shaped supplementary lighting device was also combined to adjust light distribution, according to the radiation behavior of the surface light sources in the space. The rotational schemes were proposed to solve the light obstruction in the different periods. The reason was that the illuminance was varied to the first decrease and then increase during the planting cycle, whereas the daily solar irradiance first gradually increases and then decreases, due to the solar trajectory in the solar longitude on different dates. In rainy and low-light weather, artificial light sources were introduced for supplementary lighting. Firstly, the photosynthetic response curve of strawberries was measured to determine the light compensation point and light saturation point of strawberries. The light intensity of artificial supplementary lighting was then provided for the subsequent sections. The orthogonal simulation experiments were carried out to optimize the key parameters of the supplementary lighting device. An optimal combination of the parameters was as follows. The lamp installation position was 300 mm away from the rotation point of the device, the lamp tilt angle was 7°, and the middle lamp position was numbered 1. Simulation results showed that the average illuminance was 11701 lx, and the light uniformity was 55.98%. Actual measurements of supplementary lighting were conducted using the optimized parameters. The results showed an average illuminance of 11926 lx and a light uniformity of 56.89%. The uniformity of light distribution was improved by 11.37 percentage points, compared with the commonly used supplementary lighting in greenhouses. A strawberry planting experiment was carried out on the whole device. There was a slight difference in the net photosynthetic rate of strawberry plants on the rotational cultivation frame, compared with the "H"-shaped frame. The rotational state effectively alleviated the "midday depression of photosynthesis" in strawberries, particularly for the light distribution. The average growth multiple of plant height in the rotational frame cultivation was 0.518, which was close to 0.537 of the conventional "H"-shaped ones. The photosynthetic parameters were also relatively close to those of the "H"-shaped frame. The cultivation performance of the rotational frame reached a level comparable to that of the "H"-shaped frame. This finding can also provide a strong reference for the three-dimensional cultivation mode of the cash crops, such as the strawberries, and the self-supplementary lighting of the devices.

The film-covered sweet potato transplanting method requires ensuring the transplantation conditions of small planting holes and large lateral displacement. In the soil insertion phase, the transplantation machine requires a mechanism design with multiple timed poses, and the existing design methods are still imperfect. For this reason, this article proposes a multi-target trust region parameter-guided optimization algorithm. This algorithm aims to achieve multi-objective optimization design with more timed pose conditions starting from individual timed pose conditions. First, multi-target problems are decomposed into multiple subproblems, and the parameter arrays are kept with the minimum polymerization value of each subproblem. Then, the approximate function value reduction for each target is calculated using this parameter set, and the step size for the next iteration of each subproblem is determined by comparing this approximate reduction with the actual reduction. After many iteration calculations, the parameter arrays end the calculation when the parameter group is no longer updated. This paper uses the design of a film-covered sweet potato transplanting mechanism as a complex optimized application example. The algorithm is used to obtain the optimization results of the target values of eight groups of institutions. The smallest hole is 2.99 mm, and the horizontal transplanting distance is 108.40 mm. The maximum hole is 17.64 mm, and the horizontal transplanting distance is 124.97 mm. Considering the size of the hole and the horizontal transplanting distance of sweet potato transplanting, the mechanism’s target value of the horizontal transplanting distance at 119.92 mm and the hole size at 0.31 mm were selected to design the sweet potato transplanting machine. The correctness of the results is verified, which reflects the practicability of the algorithm.

A significant compromise can often be found between the generalization of models and the insufficient original data of tea leaves in the field of deep learning, leading to a substantial decline in the detection of tender tea shoots. In this study, a Tea DCGAN (tea deep convolution generative adversarial networks) was proposed with its data augmentation. Initially, a 64×64×64 layer was added into both the generator and discriminator of the DCGAN (deep convolution generative adversarial networks), in order to enhance the perception and learning of low-dimensional features. Additionally, the LeakyReLU (leaky rectified linear unit) function in the DCGAN was replaced with the more linearly controllable ELU (exponential linear units) function, thereby improving the stability and accuracy of model training. Subsequently, a Fake Tea data augmentation framework was constructed using the Tea DCGAN network. The distribution of real tender tea shoots in the existing dataset was analyzed to determine the underlying patterns. According to these patterns, the sample images generated by the Tea DCGAN network were distributed into real tea tree images. A deep learning dataset was formed automatically. Finally, several experiments were carried out on data augmentation, including adversarial generative network ablation tests, and rare tea variety control tests. A comparison was also performed on various data augmentation at different scales. The ablation test results indicated that Tea DCGAN performed best, in terms of the FID (Frechet Inception Distance) metric. Especially after 100 000 training epochs, the FID values for the Zijuan and Longjing 43 tea varieties dropped from 322.10 to 265.63, and from 396.38 to 323.09, indicating the significantly high enhancement in the quality of the generated images. Fake Tea framework was outperformed over the various experiments of the detection model with multiple data augmentation. Specifically, the Faster R-CNN model was achieved in the mAP of 42.71% and 38.46% on the datasets with 25 Longjing 43 and 25 Zijuan tea pictures, respectively. The performance of the models was improved, as the dataset size increased. But the Fake Tea was consistently maintained on the highest mAP value across all dataset sizes. Notably, when the original dataset consisted of 200 images, the mAP value reached 89.41% suitable for intelligent tea harvesting. Therefore, there were the high effectiveness and superiority of Tea DCGAN and the Fake Tea data augmentation in the tea leaf image generation and object detection. The Tea DCGAN and Fake Tea data augmentation effectively enhanced the data acquisition to avoid the scarcity of samples. The high accuracy of detection was achieved in the tender tea shoot under various scenarios with limited samples.

In order to improve the adaptability of the planting mechanism for different plant spacings, a variable differential gear train planting mechanism based on precise pose and trajectory control was proposed by combining the open chain 2R rod group and the variable differential gear train. According to the pose requirements of receiving seedling point, transporting seedling point and planting point, three precise pose points of constrained planting trajectory were determined. Through the three-position motion generation structural synthesis method, combined with computer-aided optimization design software, a set of mechanism parameters that meet the planting requirements were optimized. Based on the optimized mechanism parameters, by only changing the coordinates of two trajectory shape control points, three planting trajectories with key point position information adapted to 300 mm, 400 mm and 500 mm plant spacing were obtained by interpolation, and three pairs of total transmission ratio of three groups of variable differential gear trains were calculated. When distributing the total transmission ratio of the mechanism, the fixed axis gear train and the differential gear train are combined. The fixed axis gear train included a pair of non-circular gear pairs and a pair of positive gear pairs, which were convenient for disassembly and assembly. The former drives the sun gear at variable speed, and the latter drives the planet carrier at uniform speed. Based on this structure, the transmission ratio of the positive gear pair is –1, and the transmission ratio of the differential gear train is 0.5. The sub-transmission ratio of the single-stage non-circular gear pair was calculated and the pitch curves of three pairs of non-circular gears were solved. Three pairs of non-circular gear pairs with different transmission ratios were replaced in turn and three sets of planting mechanisms were modeled in three dimensions. The virtual prototype motion simulation was completed by ADAMS software, and the physical prototype was built for vegetable pot seedling planting test. The theoretical solution was consistent with the attitude and trajectory of the actual test. When the test sample size was 100 plants, the actual average plant spacing was measured to be 303 mm, 402 mm, and 503 mm, with errors of 1.3%, 1.25%, and 1.88%. The width of the moving hole was 72 mm, 70 mm, and 71 mm, and the planting success rate were 94%, 96%, and 95%. The test results verified the correctness of the optimization design results of the mechanism, indicating that the variable differential gear train planting mechanism can adapt to a variety of plant spacing and has good planting effect.

In China, tea products made from fresh leaves characterized by one leaf with one bud (1L1B) are classified as “Famous Tea”, which has better taste and higher economic value, but suffers from a labor shortage. Aiming at picking automation, existing studies focus on visual detection of 1L1B, but algorithm validation is limited to a specific variety of tea sprouting in a certain harvest season at a certain location, which limits the engineering application of developed tea picking robots working in various natural tea fields. To address this gap, a deep learning model DMT (detecting multispecies of tea) based on YOLOX-S was proposed in this paper. The DMT network takes YOLOX-S as a baseline and adds ECA-Net to the CSP Darknet and FPN of YOLOX-S. The average precision (AP), precision, and recall of DMT are 94.23%, 93.39%, and 88.02%, respectively, for detecting 1L1B sprouting in spring; 93.92%, 93.56%, and 87.88%, respectively, for detecting 1L1Bsprouting in autumn. These experimental results are better than those of the five current object detection models. After fine-tuning the DMT network with another dataset composed of multiple tea varieties, the DMT network can detect 1L1B for different varieties of tea in multiple picking seasons. The results can promote the engineering application of picking automation of fresh tea leaves.