Abstract:With the rapid development of the low-altitude economy in agriculture, agricultural unmanned aerial vehicles (UAVs) have become an essential driving force in promoting agricultural modernization and the construction of smart agriculture. As the operational scale expands and task complexity increases, heavy-load agricultural UAVs—with their greater payload capacity and operational efficiency—are emerging as a key development direction for agricultural aviation equipment. The research progress and application status of heavy-load agricultural UAVs were systematically reviewed, and the development trends of key platform technologies such as aerodynamic configuration, power systems, and flight control systems were analyzed. Typical applications in pesticide spraying, seeding and fertilization, and agricultural material transportation were summarized, along with the progress and effectiveness of critical technologies such as variable spraying, spreading mechanism optimization, and sling-load control. The results indicated that heavy-load agricultural UAVs played a significant role in improving operational efficiency and expanding unmanned operation scenarios, yet still faced technical challenges in payload capacity, energy efficiency, and intelligent flight control. Future research should focus on high-efficiency power systems, intelligent decision-making and control, improvements in operational accuracy, as well as safety and reliability assessment, so as to promote the large-scale, intelligent, and sustainable development of these systems in agricultural production.

Abstract:It is crucial to achieve safe and efficient multi-target waypoint operation of crop protection drones in complex environments of mountainous orchards, reduce energy consumption of drone operations, and implement reasonable path planning. Addressing the issues of low efficiency, convoluted search paths, and slow convergence speed in traditional seagull optimization algorithm (SOA). A 3D path planning method was designed based on the improved seagull optimization algorithm (PSOA). Firstly, based on the actual scene of mountainous orchards, the vertex coordinates of the fruit tree canopy were obtained and a three-dimensional orchard model was established. Then, an energy loss model for rotary wing unmanned aerial vehicles was established in the three-dimensional path planning of a single work area to evaluate and optimize flight path performance, in order to achieve the most energy-efficient three-dimensional path planning. Finally, in response to the problem of seagull optimization algorithm easily getting stuck in local optima and slow convergence speed, the Lévy flight mechanism was introduced to expand the search range, the adaptive control factor was used to improve the search ability, and the elite retention strategy was applied to maintain population diversity, in order to obtain the globally optimal three-dimensional flight path. Comparative verification of benchmark test functions showed that PSO outperformed other mainstream optimization algorithms in terms of convergence accuracy, convergence speed, and stability. The simulation results of the experimental field showed that compared with traditional SOA, the PSO algorithm reduced the total path length by 48.74%, the total turning angle by 24.36%, the expected energy consumption by 49.10%, the expected operation time by 33.3%, and significantly reduceds the number of dangerous nodes. This method can optimize the operation path based on the crown vertex position of fruit trees and the energy consumption characteristics of drones, providing an effective solution for the three-dimensional path planning problem of crop protection drones in mountainous orchards.

Abstract:Aiming at the challenges of high path planning complexity, complex regional boundaries, and idle aircraft resources faced by multi-UAVs in grasslands, fields, and other complex scenarios, an efficient and load-balanced multi-UAV coverage path planning framework was proposed. The framework consisted of two core elements: first, an innovative two-step construction strategy of “recursive decomposition-variable neighborhood simulated annealing algorithm”. The strategy took the minimization of the total sub-polygon width as the stage goal, firstly, using the convex decomposition property of concave polygons to design the recursive decomposition method to minimize the width and localization, and then, embedding the recursive decomposition into the improved simulated annealing algorithm with variable neighborhood to achieve the width and global minimization. Secondly, a task assignment method based on UAV performance index, which calculated the performance index based on UAV speed and bypass spacing, assigned the operation area blocks accordingly and planned the multi-trip paths in combination with the range capability, aiming to equalize the task duration of each aircraft. Simulation experiments showed that the proposed two-step construction strategy can find the width and the global minimum convex decomposition scheme in all test cases, and the width were reduced by 9.072, 5.169, and 2.869 percentage points, respectively, compared with the improved genetic algorithm in the test cases of 5, 6, 7 concave vertices;the coefficients of variation of the task duration obtained from the performance index-based task assignment method were as low as 4.02%~7.33%, which can effectively realize the equalization of task duration, effectively achieved mission duration equalization.

Abstract:Plant protection unmanned aerial vehicles (UAVs) have achieved a high degree of automation in spraying agricultural chemicals, yet the process of agricultural chemical replenishment remains largely dependent on manual operation, which limits overall operational efficiency. Factors such as global navigation satellite system (GNSS) positioning inaccuracies, terrain variations, and sudden crosswinds often cause significant deviations between the actual landing point and the predefined target when the UAV returns autonomously. These deviations hinder precise docking with an automatic replenishment device. To overcome this challenge, a vision-based landing position correction system was presented for plant protection UAVs’ automatic agricultural chemicals replenishment. The system employed an OV7725 complementary metal oxide semiconductor (CMOS) imaging sensor to capture real-time images during the UAV’s landing phase. A series of image processing steps were applied, including color space conversion from red green blue (RGB) to hue saturation lightness (HSL) to improve illumination invariance, threshold segmentation for preliminary detection, and a morphological erosion algorithm for accurate boundary extraction and center localization of the UAV. An STM32 microcontroller computed the positional offset between the detected UAV center and the desired landing coordinates. Experimental validation showed that the proposed system achieved a recognition accuracy of 93.25% for the plant protection UAV and attained an average positioning error of 1.87 cm. The results confirmed the system’s capability to enable accurate alignment between the UAV and an automated liquid replenishment unit. This research offered a viable and precise technical solution for automated chemical replenishment, contributing to enhanced operational intelligence and efficiency in precision aerial agriculture. The proposed approach demonstrated strong potential for practical implementation in modern agricultural aviation systems.

Abstract:In view of the current problems such as uneven spraying, droplet drift, low effective utilization rate of fertilizers and pesticides, and serious waste and pollution caused by traditional spraying methods of unmanned aerial vehicles (UAVs), a variable angle and variable spraying device for plant protection UAVs was designed based on fruit tree canopy recognition. This device collected the canopy images of fruit trees through the camera and transmited them to Jetson Orin Nano. The recognition of the canopy boundary and the row width of fruit trees was achieved by using ROI selection and sub-interval line scanning methods. The recognition results were transmitted to the STM32 controller. Among them, the canopy boundary information was used to adjust the nozzle angle, and the pixel ratio information was used to regulate the nozzle flow rate. The feedback signals of the flowmeter and the magnetic encoder were processed by the fuzzy PID algorithm to achieve double closed-loop control of the nozzle angle and the spray volume. It was verified that the offline recognition rate of the fruit tree canopy boundary algorithm was 91.58%. The bench test results showed that the average error rate of this method in the recognition of the width of fruit tree rows was 7.32%, and the linear fitting R2=0.91. In the spray verification, when the spray angle was greater than the target area, after adjustment, the average number of deposits outside the target area could be reduced by 26.61 per cm2, and the number of droplet deposits could be reduced by 51.94%. When the spraying angle was smaller than the target area, after adjustment, the average number of external droplets deposited in the target area can be increased by 25.37 per cm2, and the number of droplets deposited can be increased by 54.8%. The kernel density estimation curve was smoother, effectively improving the uniformity of sedimentary distribution. The field test results showed that the statistical test results were t=3.29 and P=0.03,reaching a statistically significant level of P<0.05. The deposition effect was better when the device was turned on. This method realized variable angle and variable spraying control based on the width of fruit tree rows, which can effectively improve the utilization rate of droplets and provide technical support for precise pesticide application in unmanned aerial vehicle orchards.

Abstract:The pulse-jet rice seeding method achieves rows and holes seeding in an unmanned aircraft airborne manner by pressurising the rice shoots under the wrapping of a jet liquid column, which enhances the seed ejection velocity to reduce the interference of the external rotor wind field. The one-point-multiplex diverter device is an important component to improve the operational efficiency of pulse-jet rice planter. Regarding the issue of seeding uniformity during the operation of seeding devices, focusing on the liquid-solid two-phase flow of pulse-jet seeding materials (liquid material with germ seeds), using Yuenongsimiao rice seeds as the subject. A CFD-DEM coupled simulation model of a one-in-four branch device was established;in order to select the optimal branching scheme suitable for the liquid-solid two-phase flow of pulsed jet rice seeding, based on the type of branching device (straight-split axial end-type, straight-split intermediate-type, and conical-split diversion), and the angle of the inlet and outlet axes were used as factors, and the hierarchical analysis method was employed to construct a diversion scheme. As a factor, the comprehensive evaluation indexes of diversion uniformity for each outlet flow rate, flow rate, coefficient of variation of outflow uniformity and total pressure loss were constructed by using hierarchical analysis method, and two-factor and three-level orthogonal simulation tests were carried out, and the test results showed that, when the angle of inlet and outlet axial clamping angle of the cone-type diversion device was 180°, the comprehensive evaluation indexes of its diversion uniformity was 19.71%, which showed a better diversion uniformity. In order to further optimize the geometrical parameters of the one-part-four-way diverter device, the effects of the interaction between the length of the reducing section (Lr), the distance between the manifold and the inlet of the shunt device (Li) and the length of the manifold (Lb) on the overall performance of the diverter device were explored, and a three-factor, three-level orthogonal simulation experiment was designed. The preferred parameter combinations were Li=5mm, Lb=25mm, Lr=55mm (outlet diameter 11.5mm, outlet centre distance 30mm), and the comprehensive evaluation index of diversion uniformity was 5.47%;a prototype machine was made for field verification, and the test results showed that the total seed displacement of the whole machine sowing operation of the pulsed jet unmanned drone sowing operation based on a one-point-four-way diversion device was within 0.5% of the stability coefficient of variation. The coefficient of variation of stability was between 0.5% and 0.7%, the coefficient of variation of sowing uniformity between seed holes was lower than 40%, and the coefficient of variation of sowing uniformity between seed rows was lower than 28%. The coefficient of variation of stability of the total discharge between routes of the field validation test was lower than 0.27%, and the coefficient of variation of the total sowing uniformity between routes was 3.86%, which was in line with the standard of GB/T 25418—2022, and it was in accordance with the standard of GB/T 25418—2022, and the device’s performance in the field operation was verified.

Abstract:The canopy is the primary source of photosynthesis in citrus fruit trees, and it has a direct impact on the growth, yield, and fruit quality of the trees. It is the foundation for healthy and productive fruit trees, and efficient and accurate monitoring of canopy growth is especially important. Monitoring the canopy structure allows planting management measures such as pruning, irrigation, and fertilization to be adjusted promptly, optimizing the internal environment of the canopy and promoting healthy fruit tree growth and development. A dataset of citrus fruit tree canopies in a natural environment was created and a lightweight YOLO-DRR segmentation model (YOLO v5s-seg-DSConv-RFEM-RIME) was proposed to address the issues of dense planting in citrus orchards and overlap shading between canopies, which affect fruit tree growth efficiency and yield quality. Meanwhile, the model was deployed on a portable edge computing platform to improve real-time performance, reduce power consumption, and make it easier to use in inter-orchard scenarios. Firstly, the segmentation accuracy of multi-scale targets was improved based on the YOLO v5s-seg model by using the scale-aware RFE module (RFEM) for backbone networks. Secondly, the use of the distribution shifting convolution module (DSConv) to replace the C3 module in the neck network reduced memory usage in the convolutional kernel, thereby increasing the speed of operations. Thirdly, the rime optimization algorithm (RIME) was used to optimize the hyperparameters of the YOLO-DRR model, and the iterative mechanism of swarm intelligence was utilized to further improve the model performance. Finally, the YOLO-DRR model was transplanted and implemented on the FPGA edge computing platform. The FPGA device was highly environmentally adaptable and can operate reliably in a wide range of temperature and humidity conditions, ensuring the device’s dependability and stability in the complex and changing environment of the orchard. Simultaneously, the powerful edge computing capability of FPGA was used to ensure real-time data processing, more efficient use of hardware resources, reduction of power consumption and heat dissipation issues, and the realization of the requirement for long-term real-time segmentation monitoring of citrus fruit tree canopies in complex environments. The experimental results showed that the YOLO-DRR model segmented the canopy with 86.34% precision, 88.68% recall, 93.41% mAP@0.5, and 63.13% mAP@0.5:0.95. After porting it to the edge computing platform, the detection speed was increased to 19f/s while consuming only 22W of power. This suggested that the model proposed was capable of segmenting the canopy of a citrus fruit tree in the complex context of real-time canopy segmentation, which can meet the demand for real-time monitoring of the canopy growth environment in the orchard.

Abstract:In recent years, due to the lag in farmland monitoring and management methods, abandoned land has become increasingly prevalent in some rural areas, resulting in decreased utilization efficiency of cultivated land and constraints on grain production capacity. To address this issue, a cropping recommendation strategy for abandoned farmland was proposed, which integrated crop monitoring with spatial distribution perception. The approach utilized UAV-acquired remote sensing imagery with complex farmland backgrounds as the primary research object. On the basis of the DeepLabv3+ semantic segmentation model, a lightweight MobileNetv4 network was introduced as the backbone feature extractor to reduce parameter complexity and computational cost. Additionally, an adaptive fine-grained channel attention mechanism was incorporated in the decoder to enhance the model’s sensitivity to crop boundary contours and texture details. To improve the extraction of small-scale farmland features under UAV nadir perspectives, the conventional 3×3 convolution was replaced with windmill convolution. Furthermore, a hybrid focal-dice loss function was constructed to mitigate the effects of class imbalance and the difficulty in distinguishing between visually similar crop categories. Finally, by combining the remote sensing analysis results with geolocation data and crop spatial distribution statistics, the model aggregated surrounding crop information over a broad spatial domain and recommended suitable crops for abandoned plots based on seasonal farming schedules and regional crop dominance. Experimental results demonstrated that the improved DeepLabv3+ model achieved an ACC of 96.64%, mPA of 96.37%, and MIoU of 92.82%, representing increases of 1.85, 3.71, and 6.10 percentage points, respectively, over the baseline model. This approach can provide a critical technical foundation for precision crop monitoring and abandoned land reutilization, promoting intelligent agricultural management and sustainable farmland development.

Abstract:Straw mulching is an important measure for conservation tillage in the black soil region. Straw identification is of great significance for the assessment of conservation tillage implementation effects and agricultural management decisions. To address the issue that fully supervised deep learning straw identification methods due to their reliance on a large amount of pixel-level labeled data, a weakly supervised learning straw identification method was proposed based on unmanned aerial vehicle (UAV) images and the segment anything model (SAM). By fine-tuned SAM with an adapter and a boundary-aware joint loss function, and generating high-quality pseudo-labels from bounding box weak annotations, an improved U-Net segmentation network was ultimately trained to achieve straw identification. A straw extraction experiment was conducted in the conservation tillage area of corn in Lishu County, Jilin Province as the research area. The results showed that the fine-tuned SAM achieved an MIoU of 81.04% and an F1-score of 87.85%, significantly outperforming the un-fine-tuned model. The model combining SAM weak supervision and the improved U-Net achieved a higher performance than other segmentation methods, with an F1-score of 90.6%. Ablation experiments verified the effectiveness of the joint loss function and convolutional modules in improving model performance. The research provided an efficient and cost-effective solution for remote sensing identification of straw identification in maize conservation tillage in the black soil region.

Abstract:The Asian corn borer causes stem damage in the early growth stage of maize, disrupting the transport of water and nutrients. The application of non-destructive and precise detection techniques is crucial for optimizing pest control strategies and improving maize production efficiency. A multi-source feature fusion-based method for monitoring the corn borer damage level (CBDL) was proposed, integrating vegetation indices, texture features, and color indices of maize at the three-leaf stage to enhance the overall accuracy of early-stage damage assessment. UAV-mounted RGB and multispectral imaging systems were employed to acquire spectral data during the three-leaf stage. The mahalanobis distance classification (MDC) algorithm under supervised classification was used to distinguish maize from soil, followed by binary masking to remove soil background. Fourteen vegetation indices, including the excess green index (ExG) and soil-adjusted vegetation index (SAVI) were extracted;totally 32 texture features were computed from four bands based on the gray-level co-occurrence matrix (GLCM);and eight color parameters were derived. Features were selected by using the Pearson correlation coefficient (PCC), and machine learning prediction models, including random forest (RF), extreme gradient boosting (XGBoost), K-nearest neighbors (KNN), and categorical boosting (CatBoost) were constructed. Results indicated that multi-source feature fusion significantly improved model prediction overall accuracy. Among all models, the KNN model integrating vegetation, texture, and color features achieved the best overall performance, with an overall accuracy, precision, recall, F1-score, and Kappa coefficient of 91.8%, 91.9%, 91.8%, 89.5%, and 87.4%, respectively. The findings demonstrated the effectiveness of multi-source feature fusion in predicting the damage level of corn borer infestations, and it can provide a reliable technical reference for early detection and control of maize pests.

Abstract:In order to address the problems of low detection accuracy across different rice growth stages and easily missed detection under complex field environments, an improved rotated object detection model AHF-YOLO 11 was proposed based on YOLO 11. This model introduced convolutional vision module AssemFormer into YOLO 11 backbone network to construct C3k2_AssemFormer feature extraction module, significantly enhancing the local feature expression and cross reproductive global morphology learning ability in dense occlusion scenes. Furthermore, it adopted the high-level screening feature pyramid network (HSFPN) to replace the original neck structure, effectively improving the ability of the model to recognize multi-scale rice panicles under varying backgrounds. Experimental results demonstrated that the accuracy of the improved AHF-YOLO 11 model reached 90.7%, which represented an improvement of 4.3 percentage points higher than that of the original YOLO 11 model and 4.4~39 percentage points higher than the mainstream models, respectively. The results of cross-growth-stage testing further revealed that the recognition accuracy of AHF-YOLO 11 in the booting, heading, filling, and maturity stages was respectively improved by 11.88, 7.3, 4.78, and 4.3 percentage points when compared with that of the original model. Among these, the highest accuracy of 94.41% was achieved for the model at the heading stage. Further follow-up tracking studies and yield estimation experiments indicated that the period from the late heading stage to the early filling stage was the optimal period for rice yield estimation, with the lowest yield estimation error of only 4.66%. The research result can provide important technical support for rice panicle number recognition and yield estimation.

Abstract:Aiming to address the core limitations of existing rice yield estimation methods, particularly the neglect of intra-species variability and the coupled influence of planting density and single-panicle weight, which compromise accuracy, a novel approach was proposed integrating an improved YOLO v8s-obb model with the NPRP-A method. The YOLO v8s-obb architecture was enhanced by incorporating the C2f_DCNv4 module, GSConv, EPSANet, and DAT attention mechanism to strengthen multi-scale detection of rice panicles. To ensure estimation reliability, ground-truth yield data were collected through field harvesting of sample plots at maturity. Gaussian kernel density estimation and the NPRP-A-based single-panicle weight modeling were further introduced to establish a nonlinear mapping between planting density and panicle weight, capturing their interactive effects. Experimental validation was conducted across three 1m2 plots. Results showed prediction errors consistently below 5.3%, with the lowest error at 2.2%, significantly outperforming traditional methods. This demonstrated the method’s high accuracy and robustness in real-world conditions. The proposed framework not only delivered a reliable technical solution for precise and efficient rice yield estimation but also advanced crop phenotyping and yield analysis in smart agriculture. By explicitly accounting for individual plant variation and density-yield interactions, the approach bridged a critical gap in current remote sensing-based estimation practices. Its design supported scalable deployment and offered practical value for precision farming applications, highlighting strong potential for broader adoption in agricultural monitoring systems.

Abstract:Rice plant height is a core parameter for phenotypic analysis and growth state assessment, and its high-throughput detection is of significant importance for breeding and production. UAV LiDAR, with its high-precision advantages, has become a research hotspot for rice plant height detection. However, existing studies generally constructed linear regression models based on a single feature, which led to insufficient accuracy and generalization ability in complex application scenarios such as multi-variety and multi-breeding material analysis. Therefore, a multi-feature fusion strategy was adopted to build a nonlinear regression model for rice plant height, aiming to improve estimation accuracy. Field experiments were conducted in Zengcheng, Guangzhou, and Yazhou, Sanya, where plant height and laser point cloud time-series data were collected for five rice varieties and 225 breeding materials. A multi-dimensional feature system, including height percentiles, statistical parameters, and canopy profile area, was constructed. Linear and nonlinear machine learning algorithms were used to establish the detection model for rice plant height. The results showed that the accuracy of the multi-feature nonlinear prediction model was higher than that of the linear prediction model, with the highest coefficient of determination (R2) reaching 0.969 and the root mean square error (RMSE) as low as 4.73 cm. Compared with the single-feature linear model (R2= 0.905, RMSE was 8.231 cm), the R2 was increased by 7.2% and RMSE was decreased by 42.5%. Further studies on generalization ability indicated that the model built using data from the five varieties in Zengcheng, Guangzhou, showed significantly lower generalization ability compared with the model constructed from the 225 breeding materials in Yazhou, Sanya, confirming that the diversity of rice samples can effectively improve model robustness. The research result can provide a high-precision and highly applicable general technical framework for high-throughput phenotypic analysis of rice plant height.

Abstract:In plant protection unmanned aerial vehicle (UAV) operations, the interaction between a random environmental crosswind and a rotor downwash flow notably influences spray droplet deposition and drift. Obtaining consistent research results in field experiments presents challenges, and simulations based on numerical modeling methods cannot accurately describe the field velocity and spatial distribution and are time-consuming and labor-intensive. A filtered Gaussian white noise signal was designed to simulate the random environmental crosswind signal in the field, and a numerical simulation of the downwash flow field of a plant protection UAV rotor under random crosswinds was conducted based on the computational fluid dynamics method. A prediction model for the downwash flow field that utilized a cosine annealing learning rate within physics-informed neural networks (PINNs) was introduced. The downwash airflow approximated a “cylindrical” downward development within 1s, reaching a peak coverage area and exhibiting strong resistance to random crosswinds. As the crosswind speed was increased, the structure and intensity of the downwash airflow changed significantly. When the random crosswind exceeded 3m/s, the airflow coverage area and peak vertical velocity were decreased substantially, which was unfavorable for spraying operations. The PINNs prediction model demonstrated high accuracy in predicting velocity information, with overall goodness of fit (R2) values of 0.971 and 0.919 and root mean square errors of 0.364m/s and 0.253m/s for the horizontal and vertical velocity predictions, respectively. Based on the high-accuracy flow field information obtained from the model, the droplet drift trend can be further evaluated, thereby providing a physical basis for analyzing the effectiveness of spraying operations. These findings can provide valuable references for studying the influence of rotor airflow field on droplet deposition characteristics during field operations.

Abstract:With the popularization of plant protection unmanned aerial vehicles (UAVs) in tropical fruit tree spraying, problems such as uneven droplet deposition and low pesticide utilization caused by banana’s structural characteristics (e.g., large, overlapping leaves) have become prominent. To optimize spraying parameters, the orthogonal design field tests were conducted by using multi-rotor UAVs on bananas from vegetative to reproductive growth stages. Taking flight height, speed, and droplet size as key factors, combined with vertical canopy stratification and horizontal leaf zoning, the influence of parameters on droplet deposition was analyzed. Results showed that the three factors significantly affected droplet deposition on the adaxial surface of upper, middle, and lower canopy leaves (P<0.05), with distinct differences in their impact on abaxial deposition across canopy layers and stronger regulatory effects on the adaxial surface. Meanwhile, the three factors exerted an extremely significant influence on ground loss ratio (P<0.001). Based on hierarchical loss ratio control goals, two optimized parameter combinations were proposed: 30%~45% loss ratio corresponded to lower flight height (2m), medium-low speed (1.5m/s and 2.5m/s), and medium-to-large droplets (230μm and 310μm);10%~30% loss ratio can use higher flight height (3m and 4m). The correlation between spraying parameters, droplet deposition, and pesticide loss was clarified, providing quantitative basis for precise banana UAV spraying and support for droplet distribution mechanism research on tropical large-leaf fruit trees.

Abstract:Curvularia leaf spot of maize severely threatens corn yield in cold regions. However, research on emergency UAV spraying during the mid-to-late stages of disease outbreaks remains limited, with insufficient analysis of deposition characteristics and yield. Aiming to clarify the effects of typical factors including pesticide type, nozzle type, and adjuvant addition on droplet deposition within the canopy of mid-to-late stage infected corn plants and subsequent control efficacy (as measured by yield), thereby providing technical guidance for emergency management during large-scale disease outbreaks. Using agricultural UAV as the spraying platform, incorporating 13 treatments, including a blank control CK. Three variables: pesticide (Yangcai Propiconazole azoxystrobin (PA), DuPont@FaTuo@Fungicide (DF), and Pyraclostrobin Tebuconazole (PT)), nozzles (XR80015, IDK90015), and adjuvant addition (FlyWin D) were specifically designed. Precise measurements of spray deposition data, including droplet coverage rate, deposition density, and droplet size were taken at susceptible locations across corn canopy layers. Yields were continuously monitored and measured for each treatment at corn maturity. PA (suspo-emulsion) demonstrated significantly superior average deposition density (20.77 drops/cm2) and coverage (2.63%) compared with DF (suspension concentrate) and PT (suspension concentrate) (P<0.05), though its deposition uniformity was slightly inferior. Based on the above pesticide applications, the addition of the adjuvant (FlyWin D) failed to effectively improve droplet deposition, with deposition efficiency actually decreasing in some treatments. It was also found that the XR nozzle produced smaller droplet sizes and higher average deposition density (19.63 drops/cm2), while the IDK nozzle produced relatively larger droplets. Although it possessed anti-drift capabilities, its measured effective deposition was lower. Regarding corn yield after emergency control, statistical analysis showed that half of the treatments (six treatments) yielded higher than the blank control group (11313.0kg/hm2). However, since the disease had progressed to the mid-to-late stages, affecting yield, only four treatments achieved normal yield levels (12750.0kg/hm2). In summary, under the experimental conditions, the best deposition effect was achieved by using the pesticide PA without adjuvants and the XR nozzle. However, specific UAV spraying strategies, particularly pesticide selection, should consider actual yield outcomes. The two treatments (T6 and T8) using the pesticide DF achieved the highest yields despite not having the optimal droplet deposition effect.

Abstract:Mechanized production of dry direct-seeded and degradable film-mulched draught-resistant rice is challenged by the complexity of the seeding implements, the difficult control of mulched soil layers above the film, and the large deviation of soil thickness. A soil-bin investigation was conducted to illustrate soil movement induced by rotary till. Based on the recognized characteristic of soil movement, compromised design of rotary till, film application and soil mulching mechanisms was made to formulize the rotary-till soil-casting technology, which was further adapted to the development of a film-mulching dry direct-seeder for draught-resistant rice. Soil movement induced by rotary-till was theoretically investigated to formulize the model of soil casting and redirecting. Key parameters of the soil-redirecting plate were identified and single-parameter experiment was made on each parameters. The results from both the experiment and theoretical analysis clarified the selective ranges of the three parameters, i.e., dimensions of the soil-redirecting plate respecting to both the rotary blade rotation center and the ground level. By introducing the above-film soil covering rate and variation of the soil layer as criteria, orthogonal experiment of three factors in three levels was followed to illustrate the optimum combination of the three influencing factors. The rotary-till soil-casting technology was clarified and was then adapted to the development of a film-mulching dry direct-seeder for draught-resistant rice. Field test was then conducted to evaluate the performance of the developed seeder. Results showed that the proposed three parameters had significant influence on the performance of rotary-till soil-casting technology. The optimum combination of the design parameters were 38° for the tilting angle of the plate, 111mm of height of the plate and 362mm from the plate to the rotary blade rotation center. Such a combined set of parameters yielded the best performance of 94.2% soil covering rate and 14.8% of soil thickness variation in soil bin test. While in the field evaluation they were 94.5% and 14.8%, respectively. Rotary-till soil-casting technology was featured by 50mm shallow soil engaging and 500r/min high speed rotating, which achieved the purposes of soil-cutting and high casting, overtopping the film-applying mechanisms and soil mulching on the applied film. The findings of parameterized design of soil-redirecting plate in this work can be applied as a reference for related research in the future.

Abstract:Aiming to address the issue of unstable conversion of sugarcane segments between the transfer chain and achieve uniform filling and seeding of sugarcane segments, a baffle belt type transversal sowing seeder was designed, which was realized by a conveyor belt with partitions driven by a motor to rotate. The effects of conveyor belt speed, conveyor belt angle and the bottom angle of sugarcane collecting box on the performance of filling were investigated by analyzing the motion of sugarcane segments on the conveyor belt and simulating the motion of the seeder based on ADAMS. By using the prototype test platform, the speed of conveyor belt, the number of sugarcane segments in sugarcane collecting box and the angle of conveyor belt were studied. The results showed that the speed of the conveyor belt A and the number of sugarcane segments in sugarcane collecting box B had significant effects on the qualified rate of filling seed. The belt angle C had significant effect on the qualified rate of filling seed. The angle of the bottom angle of sugarcane collecting box had no significant effect on the qualified rate of filling seed. The order of the major and secondary influences of A, B and C on the qualified rate of filling seed was B, A and C. The optimal parameter combination B2A2C4, that was, the speed of conveyor belt, the number of sugarcane segments in sugarcane collecting box and the angle of conveyor belt were 8.16m/min, 20 and 55°, respectively. Under this combination of parameters, the qualified rate of seed filling, distance between sugarcane segments and deflection angle of sugarcane segments were 95.4%, 99.4% and 96.3%, all of which met the agronomic requirements.

Abstract:Aiming to address key operational issues in automatic transplanting machines, such as missed planting, seedling blockages and abnormal seedling planting state, an optimized lightweight detection model, YOLO v5n-GE, was proposed for real-time monitoring of seedling conditions within transplanting equipment. The research began by collecting images of both single and multiple pepper seedlings under varying lighting conditions (front and backlighting) using a camera. To reduce computational load and latency, traditional convolutions were replaced with Ghost convolutions on the basis of YOLO v5n model, and the main feature extraction modules were substituted with improved FastGhost and SimAMGhost modules. EMA attention mechanism was applied to enhance the network’s focus on important detail information, effectively improving the model’s recognition results for highly overlapping pepper seedlings, reducing sensitivity to occlusions, and increasing recognition accuracy. Additionally, Shape-IoU loss was used to replace CIoU loss, addressing the influence of the bounding box shape on bounding box regression and improving bounding box regression accuracy. Experimental results on the self-built dataset demonstrated that the improved YOLO v5n-GE model achieved an mAP of 95.3%, representing a 0.3 percentage points improvement over the original model. The model’s parameter count and computational load were reduced by 52.5% and 51.2%, respectively, detection speed was increased by 12.2%. These enhancements enabled efficient, real-time detection of pepper seedlings while maintaining high accuracy, demonstrating the improved algorithm’s effectiveness. The research result can provide technical support for seedling recognition in transplanting machine components with limited hardware resources.

Abstract:The realization of autonomous and safe operation for orchard mobile robots relies critically on efficient path planning technology. To address the prevalent challenges of low planning efficiency, excessive path turning points, and poor smoothness in existing path planning methods within complex orchard environments, an autonomous path planning method was proposed, integrating an adaptive A* algorithm with trajectory optimization, improving the autonomous navigation and operation performance of robots. Firstly, an orchard grid map model was constructed as the foundation for global planning. Secondly, a real-time optimization mechanism combining cost-weighted and centerline offset functions enhanced the adaptive A* algorithm, and a dynamic five-neighborhood search strategy was introduced for comprehensive global path searching. Subsequently, third-order Bézier curves were applied for adaptive path smoothing, generating a curvature-continuous navigation trajectory that met the operational requirements of orchard robots. Simulation and field experiments conducted in representative orchard environments demonstrated that compared with an improved A* algorithm, the proposed method significantly reduced the average path planning time by 23.8% (17.15ms) and 23.1% (16.09ms) in obstacle-free and obstacle-present scenarios, respectively, while achieving a reduction in path average curvature by 10.7% (0.011m-1) and 15.8% (0.028m-1). Field tests further validated the reductions in average planning time by 26.4% (19.01ms) and 27.4% (21.28ms), along with decreases in path average curvature by 7.3% (0.009m-1) and 8.7% (0.013m-1) under the respective scenarios. The proposed method significantly enhanced path planning efficiency and smoothness, effectively meeting the practical operational demands of orchard robots and demonstrating strong potential for practical application.

Abstract:Aiming to improve the cleaning performance of maize grain harvester cleaning device under the condition of large feeding mass, a multi-tetragonal conical sieve was designed to make the maize grains on the sieve retained and quickly penetrate the sieve through the non-planar structure. The key structural parameters of the multi-tetragonal conical sieve were determined by theoretical analysis. CFD-DEM method was used for numerical simulating the motion of gas-solid two-phase in multi-tetragonal conical sieve cleaning device. A high speed airflow belt could be formed in front of the multi-tetragonal conical sieve, which was beneficial to make maize mixtures separated. The air velocity in the middle of the sieve body was low, which was conducive to the retention of maize grain with the structure of multi-tetragonal conical sieve. The maize grain on the sieve completed the process of ‘impact sieve-retention-leaping-impact sieve-through the sieve’ under the movement of multi-tetragonal conical sieve. No build-up of the maize mixture on the sieve. The maize grain on the sieve completed rapid penetration of the sieve. The height of multi-tetragonal conical sieve unit, the airflow velocity at the inlet of the cleaning device and the vibrating amplitude of the vibrating sieve were selected as the test factors. The impurity rate, loss rate and screening efficiency of maize grain were selected as test indexes. The quadratic orthogonal rotation combination test was carried out. The mathematical models between factors and indicators were established. The best combination of parameters was obtained as follows, the height of the multi-tetragonal conical sieve unit was 11.26mm, the airflow velocity of the cleaning device was 9.6m/s, the vibration amplitude of the vibrating screen was 15.13mm. The motion law of maize grains on the sieve was verified through the bench validation test. At the maize mixture feeding mass of 7kg/s, the impurity rate of maize grain after screening by the multi-tetragonal conical sieve cleaning device was 1.68%, the loss rate of maize grain was 1.33%, and the screening efficiency of maize grain was 3.66kg/s. The multi-tetragonal conical sieve cleaning device could meet the requirements of cleaning maize mixtures with large feed mass. The research result had a certain reference value for improving the working performance of maize grain combine harvester.

Abstract:Aiming to deeply study the working mechanism and the technical approaches for improving the working quality of the potato soil separation device during harvesting, and solve the problems of high injury rate and skin breakage rate, low efficiency and inconvenient data collection in field tests of potato harvesting machinery, a potato harvesting loss reduction rule test bench was designed. The test bench mainly consisted of a separation device, a vibration frequency adjustment device, an amplitude adjustment device, an angle adjustment device, a material monitoring device, and a full-angle observation device. The test bench adopted a visual design, allowing for all-round observation of the separation process. Through theoretical analysis, a dynamic model of the potato-soil mixture during the conveying and separation process was established. High-speed photography technology was used to analyze the occurrence and evolution of mechanical behaviors such as potatosoil collision, rolling backflow, and rupture. Through curve fitting, the velocity and acceleration curves of the potato-soil mixture were obtained. Through singlefactor experiments, the influence intervals of the main working parameters, including linear velocity, device angle, vibration frequency, and amplitude, were clarified. Response surface experiments were conducted to find the optimal working parameters. The results showed that when the linear velocity was 1.2m/s, the device angle was 20°, the vibration frequency was 12.5Hz, and the amplitude was 50mm, the injury rate of potatoes was 0.81%, the skin breakage rate was 1.36%, and the impurity rate was 2.256%. All working indicators met the requirements of potato mechanical harvesting operations, providing theoretical basis and technical support for the design and optimization of low-loss and high-efficiency potato harvesting equipment.

Abstract:The existing crushing devices cannot cope with the hardness and shape of walnut shell, resulting in poor crushing effect and low screening efficiency. Combined with the shell material characteristics and crushing methods, a hammer type walnut shell crusher was designed to suppress the circulation layer effect. The equipment can complete the crushing, screening and collection operations as one of the working procedures. As one of the key components of crushing walnut shell, the threedimensional structure of hitting end face was designed based on the curve equation theory, and the arc tooth plate auxiliary parts were designed to complete the crushing process. The gas-solid coupling method was used to analyze the motion and crushing process of walnut shell under the action of new hammer and arc-shaped tooth plate, and the structural strength of key components was verified by static simulation. The initial hammering angle, hammerscreen clearance and steering angle were further used as test factors, the regression model of productivity and power consumption per ton of material was established by the Box-Behnken test method, and parameter optimization were carried out to obtain the optimal combination as follows. When the initial hammering angle was 43.89°, the hammer-screen clearance was 13.87mm, and the guiding angle was 117.95°, the productivity was 98.12kg/h, and the power consumption per ton of material was 3.27kW·h/t, which was 24.55% higher than the traditional crushing efficiency, 15.69% higher than the traditional productivity, and 11.52% lower than the power consumption per ton of material. It not only met the operation requirements of efficient walnut shell crushing, but also provided theoretical basis and technical support for nut core shell crushing.

Abstract:In order to explore the influence of gate slow start on the hydraulic stability during the start-up process of the pump unit, a full-flow geometric model of the pump unit was established, and the gate movement and runner rotation were realized by using the laying mesh and dynamic mesh technology. The three-dimensional numerical simulation of the start-up process of the pump unit at different gate opening speeds was carried out. Three control schemes with different gate opening speeds were set up to explore the influence of gate slow start on the hydraulic stability of the pump startup process from the perspectives of external parameter changes, pressure pulsation characteristics and internal flow characteristics. Combined with the entropy production theory, the energy loss in different regions was quantified, and the change of energy loss in the pump starting process was further analyzed. The results showed that different control schemes showed similar evolution trends in external parameters, pressure fluctuation and head loss, but there were differences in time process and change amplitude. During the starting process, slowing down the opening speed of the gate would slow down the flow change and the falling rate of the external characteristic parameters. The inlet flow of Scheme 1, Scheme 2 and Scheme 3 reached the operating flow value at 12.2s, 25.4s and 56.4s, respectively. It would also increase the fluctuation amplitude of torque and axial force, and increase the peak pressure and pressure pulsation peak-to-peak value in the unit. In terms of head loss, the influence of reducing gate speed on energy loss had obvious time-domain duality: in the acceleration stage, the energy loss was mainly increased, while in the later stage, the loss was reduced.

Abstract:Modal characteristic analysis is a crucial step in the design of hydraulic machinery. When the hydraulic excitation characteristics match the natural modes of the structure, it may lead to intense resonance phenomena, threatening the safe and stable operation of the power station. The hydrofoil, a simplified model of the impeller, was taken as the research object. The acoustic fluid-structure interaction method was adopted to analyze the dry and wet modal characteristics of the hydrofoil, and the effects of the angle of attack and trailing edge modifications on the modal characteristics were investigated. It was found that the natural frequency of the hydrofoil was decreased significantly under the effect of the added mass, and the mode shape of the wet mode also changed obviously. When the angle of attack changed linearly, the natural frequency of the hydrofoil showed a non-linear variation trend, with the maximum frequency occurring at an angle of attack of 2°. When the angle of attack increased to 10° and 15°, the natural frequency of the hydrofoil decreased significantly, and the frequency was the smallest at 15°. The change in the angle of attack had little effect on the spanwise modal displacement of the first four-order modes and the chordwise modal displacement of the torsional modes of the hydrofoil, while the difference in the chordwise modal displacement of the bending modes fs1 and fs3 was increased. In the case of trailing edge modification, the natural frequency of the hydrofoil was increased, with the maximum frequency corresponding to the 30° modification. The trailing edge modification had an obvious effect on the deformation of the hydrofoil’s trailing edge: suppressing the bending deformation of fs1 and fs3, and significantly reducing the range of modal displacement changes. Moreover, there was a slight decrease in the chordwise modal displacement values of the tail edges of torsional modes fs2 and fs4, as well as the spanwise modal displacement values of fs3.

Abstract:The rapid development of China’s protected horticulture industry has led to a surge in demand for intelligent knowledge services. However, the current fragmented and loosely connected knowledge systems, along with imprecise and inefficient knowledge service methods, pose significant challenges in guiding production. Moreover, practitioners’ descriptions of issues are often incomplete, further complicating the resolution of protected horticulture problems. To address these issues in protected horticulture production, integrating knowledge graph (KG) and large language model (LLM) were proposed to create a multi-source knowledge-enhanced question-answering model. Initially, a knowledge dataset for protected horticulture was constructed, encompassing over 60 commonly cultivated categories in protected horticulture and containing nearly 1.5 million words. Through semantic segmentation, totally 26349 textual blocks were obtained and stored in a vector database. Additionally, textual knowledge related to production techniques was extracted from the dataset to construct a knowledge graph. Concurrently, a semantic information enhancement model was proposed based on KG entity matching. Subsequently, a retrieval-augmented generation method was designed, in which the KG and related textual information were input into the prompt template to improve the LLM’s problem-analysis capabilities. Furthermore, to enhance its adaptability in the field of protected horticulture, the LLM was fine-tuned on relevant question-answering corpora by using low-rank adaptation (LoRA) method. Based on this, a multi-source knowledge-enhanced LLM (named PengKGPT) was developed to reason and respond to issues in protected horticulture production. Finally, the case studies revealed that PengKGPT attained score and accuracy rates of 91.2% and 82.10%, respectively, marking improvements of 36.6 and 32.53 percentage points compared with the base model. This enhancement significantly augmented the large language models analytical capabilities for questions in vertical domains. When benchmarked against classic commercial models such as ERNIE 4.0 Turbo and GPT-4o, PengKGPT demonstrated increases of 10.2 and 14 percentage points in score rate, along with improvements of 10.4 and 12.69 percentage points in accuracy rate, respectively. These results indicated that PengKGPT exhibited superior professionalism and reliability in addressing challenges within protected horticulture production. The results indicated that this approach can provide auxiliary support for protected horticulture production.

Abstract:The alpine meadow ecosystem has a strong carbon sink capacity, and accurately estimating the gross primary productivity (GPP) is essential to grasp the global carbon cycle. Solar-induced chlorophyll fluorescence (SIF) is a nondestructive probe indicating the photosynthetic process of plants, and red SIF (RSIF) contains more information about PSⅡ. To explore the response characteristics of RSIF to the GPP of the alpine meadow ecosystem, integrating RSIF, environmental variables and canopy structure parameters, and respectively constructed GPP prediction models based on random forest regression (RFR), multiple linear regression (MLR), and simple linear regression (SLR) methods. The results showed that both canopy RSIF and FRSIF were significantly positively correlated with GPP, and the correlation between RSIF and GPP was 23.53% higher than that of FRSIF, which had greater advantages than FRSIF in the prediction of alpine meadow GPP. In the training data set, the RFR and MLR models constructed by combining RSIF, environmental variables and canopy structure parameters increased the average R2 between predicted GPP and measured GPP by 5.79% and 12.69%, respectively, compared with FRSIF, and the average RMSE was decreased by 16.37% and 30.56%, respectively. Compared with FRSIF, the average R2 between the predicted GPP and the measured GPP was increased by 31.02% and the average RMSE was decreased by 34.28% in SLR model with a single RSIF as the independent variable. In the validation data set, the average R2 between predicted GPP and measured GPP predicted by the RFR model with RSIF, environment variables and canopy structure parameters as independent variables was increased by 1.86% compared with MLR, and by 6.62% compared with the single SLR model with RSIF as independent variable, and the corresponding average RMSE was decreased by 1.04% and 17.13%, respectively. RSIF had greater potential than FRSIF for GPP monitoring in alpine meadow ecosystem, and the results also had important reference value for GPP monitoring in other ecosystems.

Abstract:High spatial and temporal resolution NDVI data is of great significance in the application of agricultural remote sensing. Spatiotemporal fusion (STF) models can serve as an effective approach to enhance the spatiotemporal resolution of NDVI data. An STF model, UBiaSTF, was proposed which integrated the Unet framework into BiaSTF, and applied it to the spatiotemporal NDVI fusion of Landsat 8 and Sentinel-2 with MODIS imagery in the Jiefangzha Irrigation District. The model was compared with ESTARFM and BiaSTF models to analyze its effectiveness in the reconstruction of remote sensing time series NDVI. The results indicated that the UBiaSTF model performed excellently in the reconstruction of NDVI time series, with the coefficient of determination R2 significantly improved compared with that of other models, reaching a maximum of 0.930. Additionally, the UBiaSTF model demonstrated strong stability in long time series data fusion tasks, effectively overcoming the impact of reference image temporal interval changes on prediction accuracy. Furthermore, the UBiaSTF model showed the lowest fusion error in the reconstruction of time series NDVI across different vegetation coverage categories compared with ESTARFM and BiaSTF, closely matching the actual changes. This model can serve as an effective tool for the reconstruction of time series NDVI in areas with vegetation coverage.

Abstract:Accurate delineation of small and spatially fragmented paddy fields from remote sensing imagery remains a challenging task. Traditional inputs based on linear spectral combinations are limited in their ability to capture nonlinear couplings across spectral bands, while direct stacking of multi-band imagery often introduces redundant information and increases computational complexity. To address these limitations, an enhanced U-Net-based segmentation framework was proposed. The network employed a dual-input strategy, integrating both RGB and NRG false-color images derived from the Gaofen-2 satellite, and incorporated a dual-encoder architecture to extract complementary multimodal feature representations. To further strengthen the discrimination of fine-scale objects, a local pyramid attention (LPA) module that enabled hierarchical aggregation of local contextual cues was designed, thereby improving the model’s sensitivity to small paddy patches with irregular boundaries. In addition, an adaptive multi-scale attention dynamic feature fusion (AMSADFF) module was introduced to dynamically integrate features across multiple scales, mitigating redundancy while preserving the most informative spatial and spectral patterns. By synergizing these mechanisms, the proposed framework—termed DFAU-Net—achieved a robust balance between local detail preservation and global contextual understanding. Experimental evaluations were conducted on a dedicated paddy field dataset constructed from high-resolution Gaofen-2 imagery. Results demonstrated that DFAU-Net consistently outperformed several state-of-the-art segmentation models. Specifically, it achieved a dice coefficient (Dice) of 77.54%, a mean intersection over union (mIoU) of 86.34%, and an overall accuracy (Acc) of 91.48%. These results highlighted the superiority of the method in capturing fragmented and smallscale field patterns, where conventional models tended to fail. Overall, DFAU-Net provided a promising solution for accurate agricultural parcel mapping, with potential applications in crop monitoring, yield estimation, and precision agriculture.

Abstract:The main issues in extracting forest parameters and estimating stand volume using UAV laser scanning (UAV-LS) are insufficient accuracy in individual tree segmentation and the inability to directly obtain diameter at breast height (DBH) parameters. To address this limitation, the UAV-LS was utilized to collect high-density point cloud data from Eucalyptus and Chinese fir plantations. By improving the mean shift algorithm (IMSA), a method capable of accurately obtaining the diameter at any height of standing trees was proposed, thereby calculating tree volume and achieving accurate estimation of stand volume from the perspective of individual tree segmentation. The results showed that the improved mean shift algorithm effectively handled dense noise near the tree trunk, significantly enhancing detection accuracy. The accuracy of determining and fitting the edge points of the trunk was optimal, with coefficients of determination R2>0.93 for estimating diameters at heights of 1.3m and 2m, and average relative errors of 2.41% (Eucalyptus) and -4.05% (Chinese fir). The model can effectively estimate the diameter and volume of individual trees in Eucalyptus and Chinese fir plantations, with performance for Chinese fir being optimized compared with Eucalyptus. Point cloud density significantly affected the estimation performance of the model. When using point cloud densities of 50% or less of the original density for individual tree measurements, the omission rate was increased significantly;when using only 10% of the original density, the maximum absolute error exceeded 86%. The research result can provide technical support and theoretical basis for the timely, accurate, and efficient estimation of stand volume at the individual tree scale by using UAV-LS, while also offering a reference for high-precision forest resource assessment under resource-limited conditions.

Abstract:Carotenoid content in wheat grain is the key index to measure the nutritional value and breeding quality of wheat. In order to realize the rapid and nondestructive detection of carotenoid content in wheat grain, a near-infrared spectrum rapid acquisition device for wheat grain was designed, which can realize the equal sample size of the sample to be measured, and quickly collect near-infrared spectrum data for many times, so as to improve the spectral acquisition efficiency. Taking 180 wheat grains as the research object, the near-infrared spectral data in the range of 900~1700nm were obtained. Savitzky-Golay (SG), standard normal variate transformation (SNV), multiplicative scatter correction (MSC), trend correction (TC), SG+TC, SG+MSC, SG+SNV, Savitzky-Golay+first derivative (SG+1D), Savitzky-Golay+second derivative (SG+2D), Savitzky-Golay+third derivative (SG+3D) ten pretreatment methods, four feature selection algorithms, relief algorithm (Relief), genetic algorithm (GA), variance threshold (VT) and successive projections algorithm (SPA), were used to establish three mathematical models of partial least squares regression (PLSR), support vector regression (SVR) and adaptive boosting regression (ABR) to predict carotenoid content in wheat seeds. The results showed that the prediction effect of ABR model based on SNV preprocessing, relief feature selection algorithm and CSO-ACF improved strategy was the best. The determination coefficient R2C of correction set was 0.90, the root mean square error (RMSEC) was 0.29μg/g, the determination coefficient R2P of prediction set was 0.90, the root mean square error (RMSEP) was 0.32μg/g, and the RPD was 3.16. Therefore, the device combined with the model algorithm can achieve rapid and nondestructive prediction of carotenoid content in wheat.

Abstract:The integrated monitoring of airborne pathogenic spores using intelligent spore trapping devices has become a crucial approach for early online warning of crop airborne diseases. To address issues such as image defocus blur and low spore detection accuracy caused by fixedfocus microscopic imaging under complex working conditions, focusing on urediospores of wheat stripe rust, an automatic focusing system for spore microscopic imaging was designed. This system integrated YOLO v5n object detection with a spore morphology-adaptive dynamic step search strategy to achieve adaptive tracking and counting of urediospores in complex backgrounds. Firstly, a low-cost portable microscopic image acquisition device was constructed by using a Raspberry Pi microcontroller and CMOS image sensor. A stepper motor drived the lens barrel vertically (1/8 microstepping mode, step size 0.625μm) to capture multifocal urediniospore image sequences. Secondly, an improved spore focusing evaluation function was innovatively proposed by combining the YOLO v5n model with the traditional squared modified Laplacian (SML) gradient evaluation function, effectively solving the misjudgment of focal planes caused by background impurities. Finally, a spore morphology-adaptive dynamic step search strategy (coarse search: 10 micrometers per step;fine search: 2.5 micrometers per step) was implemented to optimize focusing efficiency. Experimental results demonstrated that the proposed evaluation function achieved 97.44% spore counting accuracy, representing a 56.54 percentagepoint improvement over that of traditional gradient-based methods. The automatic focusing success rate reached 98% with an average focusing time of 116.49s. The developed autofocusing algorithm (high accuracy/robustness) combined with the low-cost, fastresponse portable imaging device significantly advanced intelligent spore trap automation, offering key technological solutions for cross-regional management of airborne crop pathogens.

Abstract:Fine-grained classification of growth stages is a prerequisite for achieving intelligent and precise environmental control in seafood mushroom cultivation. However,due to the subtle phenotypic differences between adjacent growth stages and the high granularity of stage division required for regulation,accurate classification remains challenging. An enhanced YOLO 11-based method for fine-grained growth stage classification was proposed. Firstly, a global attention mechanism(GAM) was integrated into the YOLO 11 backbone network to enhance channel and spatial attention, thereby improving the extraction of discriminative features. Secondly,the activation function was replaced with Mish to strengthen the network’s nonlinear representation capability.Finally, the original convolution was optimized to GhostConv,simplifying the model architecture while maintaining high detection accuracy and computational efficiency.Experimental results demonstrated that the improved algorithm achieved a recognition accuracy of 96.97%, a recall rate of 96.73%, a mean average precision (mAP) of 96.58%, and a precision of 96.81%.Furthermore, the inference time and model parameters were reduced by 4.28% and 21.69%, respectively, outperforming that of RF-SVM,ResNet50, YOLO v8,and the original YOLO 11. These results indicated that the proposed method exhibited superior comprehensive performance and can be effectively applied to fine-grained classification of seafood mushroom growth stages,providing a robust foundation for intelligent environmental regulation in mushroom cultivation.

Abstract:Automated counting of silkworms is crucial for intelligent sericulture management. However, due to small individual size, morphological variations, and severe occlusion in dense farming environments. This impeded precise distribution of mulberry leaves and lime powder for optimal feeding and disinfection. To address dense silkworm counting challenges and enhance detection and segmentation accuracy, DLC-YOLO, an improved instance segmentation algorithm was proposed based on YOLO 11n. Firstly, in the backbone network, a DysnakeConv and C3k2 fused structure was designed, utilizing dynamic deformable convolution kernels to adaptively trace silkworm contours, strengthening feature extraction for slender, curved morphologies. Secondly, in the neck network, a CGAfusion content-guided attention mechanism replaced the original feature fusion, integrating multi-scale cross-layer features from the backbone and feature pyramid at the detection head input. This significantly improved multi-scale feature fusion efficiency, particularly for dense small targets. Finally, an LSCSBD-Seg segmentation head incorporated local semantic constraints and boundary refinement modules to enhance edge segmentation precision in dense clusters. A dedicated dense silkworm instance segmentation dataset covering varied rearing densities and lighting conditions was constructed for validation. Experimental results demonstrated that DLC-YOLO outperformed YOLO 11n, with detection precision (mAP50, mAP50:95) increasing by 2.3 and 5.1 percentage points, and segmentation precision (mAP50, mAP50:95) improving by 3.0 and 2.9 percentage points, respectively.

Abstract:Intelligent and green is the current development trend of agricultural environmental protection equipment. In order to achieve long-term stable and efficient operation of spiral extrusion solid-liquid separator and reduce labor cost, a set of adaptive discharge end pressure device and its automatic control system was developed, which can automatically adjust the size of the outlet according to the change of outlet pressure. The contact curve of the pressure valve was designed based on the principle of bionics, and the structure of the pressure valve at the discharge port was optimized by combining discrete elements with experiments. The contact parameters of cow dung were calibrated through friction characteristic test and collision test, and the discrete element model of cow dung particles in the process of spiral extrusion dehydration was established. Comparing the opening size of the pressure valve and the mass flow rate of the discharge port, three kinds of pressure valves with different structures were selected. Through orthogonal test, the influences of pressure valve structure, initial pressure at discharge end and rotation speed of spiral shaft on separation efficiency, moisture content of solid after separation and effective energy consumption of the machine were studied. The results showed that under the bionic (vole claw toe) secondary fitting pressure valve, the initial pressure at discharge end of the test prototype was 1000N and rotation speed of spiral shaft was 60r/min. The separation efficiency reached 2279.76kg/h, the moisture content of cow manure was 58.73%, the effective power was 2.83kW, the separation efficiency was the highest, the energy consumption per ton of cow manure was the lowest, and the effective power was reduced by 30% compared with that before optimization.

Abstract:Aiming to address the issues of high cost and subjectivity in the industrial grading of preserved egg gel quality, a multi-model feature fusion classification framework tailored for visible/near-infrared spectral data was proposed. Firstly, key spectral wavelengths were extracted by using the competitive adaptive reweighted sampling (CARS) algorithm, and a support vector machine (SVM) model was built, achieving a classification accuracy of 90.8%. Secondly, a one-dimensional efficient channel attention module (ECA_1D) was designed and integrated into a residual-connected one-dimensional convolutional neural network (1DCNN), resulting in the 1DCNN_ECA model, which achieved an accuracy of 92.8% by extracting deep spectral features. Additionally, a long short-term memory (LSTM) network was enhanced with a self-attention mechanism to construct the LSTM_Self model, effectively capturing long-range dependencies in spectral data and reaching an accuracy of 92.1%. These three feature representations, derived from the CARS algorithm, the 1DCNN_ECA model, and the LSTM_Self model, were further fused to develop the TripleFusion model, which achieved a grading accuracy of 95.0%, outperforming all dual-model fusion configurations. The results demonstrated that multi-model feature fusion can compensate for the limitations of individual models in feature representation, significantly improving classification performance. This work can effectively address the challenge of non-destructive grading of preserved egg gel quality and provide a novel and robust approach for visible/near-infrared spectral data analysis and modeling.

Abstract:Based on spot check data of livestock and poultry products in China from 2016 to 2023, the risk characteristics of these products were evaluated by using descriptive statistical methods. A combined method of analytical hierarchy process and entropy weighting method was employed for risk ranking. Additionally, global Moran’s I and local Moran’s I were applied to investigate the spatial distribution characteristics of risk. The results provided a clear depiction of how risk levels varied across product types and regions. Specifically, the findings revealed that veterinary drugs and illegal additives were the primary sources of risk, and poultry and pork products exhibited a relatively high overall risk. Notably, regions such as Shandong, Henan, and Chongqing showed a prominent risk for residues of veterinary drugs, while problems related to illegal additives were particularly outstanding in Henan, Hunan, Chongqing, and Guizhou. Enrofloxacin, florfenicol, and ofloxacin posed the highest overall risk nationally. Moreover, high-risk veterinary drug residues were concentrated in East China, with Shandong reporting the highest diversity of illegal additives. The risks associated with enrofloxacin, chlorpromazine, nicarbazin, and four types of illegal additives demonstrated significant spatial clustering. Specific areas exhibited high-high clusters of veterinary drug residues such as enrofloxacin and doxycycline, as well as clenbuterol. Consequently, there was a need to strengthen risk prevention measures in these regions. The research result can provide valuable insights for the risk prevention and control of livestock and poultry products, serving as a reference for policymakers.

Abstract:Aiming to address the issues of significant resource consumption, data loss, and limited fault monitoring capabilities in field reliability testing of tractors, an intelligent tractor operation information monitoring system for reliability test analysis was developed. Based on reliability assessment standards and evaluation systems, the key parameters and calculation methods required for monitoring tractor unit reliability were determined, the hardware design of the tractor operation information monitoring system was developed, including components such as sensors and a data acquisition terminal. The software component and remote monitoring platform of the system were designed by using LabVIEW and the Windows 10 IoT system. Bench test, test track trials, and field tests were carried out, and the results showed that the monitoring system achieved an R2 of 0.9582 for measured tractor fuel consumption compared with actual values, with a maximum error of 0.47% for measured engine speed and a maximum error of only 3.1% for measured operation area compared with actual area, meeting the accuracy requirements for reliability test analysis. During a 75-hour field test, the developed system reliably monitored tractor operation reliability information. Specifically, the tractor’s high-load operation time was 54.56h, cumulative fuel consumption was 512.96kg, operation area was 57.75hm2, average hourly fuel consumption was 6.83kg/h, average fuel consumption per unit workload was 8.88kg/hm2, average productivity was 0.77hm2/h, and the average field operation load coefficient was 58%. Additionally, the monitoring system identified one fault that was undetectable by the operator. The monitoring system developed can provide an effective technical tool for the analysis and validation of tractor reliability tests.

Abstract:In order to realize non-destructive and fast grasping of soft and fragile spherical fruit, tomatoes were taken as the research object, and a type of flexible contour-profiling smart gripper was designed based on magnetorheological (MR) fluid smart materials. At first, the structure of the smart gripper was designed according to the size parameters of tomato. The smart MR gripper mainly consisted of four parts: the base, the electric parallel driving part, the connecting plate and the MR module. The MR fluid was constrained by a flexible film in the MR module, which can flexibly and tightly fit the surface of the tomato fruit, thereby achieving non-destructive grasping. Then the rationality of the mechanical structure of the MR module was verified by finite element simulation, and the optimal dimensional configuration was determined. To determine how the applied current and set position affected the gripper stiffness and maximum grasping quality, the stiffness test and standard sphere grasping test were carried out. The results showed that the current was positively correlated with the gripper stiffness and maximum grasping quality, while the position was negatively correlated with the maximum grasping mass and positively correlated with the gripper stiffness. Finally, in order to verify the actual grasping performance of the gripper, the tomato grasping test was carried out. The test results showed that the success rate was 100%, the direct damage rate of grasping was 0, and the average task time was 5.39s, indicating that the gripper could quickly and non-destructively grasp tomatoes of different sizes.

Abstract:Aiming to investigate the active control methods for cavitation in the gap between hydrofoils, numerical simulations were conducted by using a coupled LES model, Schnerr-Sauer cavitation model, and phenomenological plasma model. Under the conditions of an actuation voltage of 15kV, a hydrofoil angle of attack of 8°, and an incoming flow velocity of 10m/s, the effects of plasma actuation on the gap cavitation characteristics of an NACA66(MOD) hydrofoil were analyzed. The results showed that without plasma actuation, in the case of small gaps, intense sheet cavitation and shear flow suppressed the development of tip leakage vortex cavitation, limiting its propagation to the mid and lower part of the hydrofoil sidewall. As the gap increased, sheet cavitation and shear flow were weakened, reducing the suppression effect on the tip leakage vortex, allowing the cavitation to develop further. Plasma actuation weakened sheet cavitation and cloud cavitation for small-gap hydrofoils, but the intense and unstable vortex structures hindered effective suppression of tip leakage vortex cavitation. For large-gap hydrofoils, plasma actuation increased the pressure on hydrofoil sidewall, disrupting the conditions for cavitation development, suppressing the development of separation vortices and the generation of induced vortices, and enhancing the stability of the tip leakage vortex, and achieving significant suppression of gap cavitation. This led to significant suppression of cavitation in the hydrofoil gap and improves, to some extent, the hydrodynamic performance of the hydrofoil.