Precision Operation Technology and Intelligent Equipment in Farmland

The accurate and rapid acquisition of fruit tree canopy parameters is fundamental for achieving precision operations in orchard robotics, including accurate spraying and precise fertilization. In response to the issue of inaccurate citrus tree canopy segmentation in complex orchard backgrounds, this paper proposes an improved DeepLabv3+ model for fruit tree canopy segmentation, facilitating canopy parameter calculation. The model takes the RGB-D (Red, Green, Blue, Depth) image segmented canopy foreground as input, introducing Dilated Spatial Convolution in Atrous Spatial Pyramid Pooling to reduce computational load and integrating Convolutional Block Attention Module and Coordinate Attention for enhanced edge feature extraction. MobileNetV3-Small is utilized as the backbone network, making the model suitable for embedded platforms. A citrus tree canopy image dataset was collected from two orchards in distinct regions. Data from Orchard A was divided into training, validation, and test set A, while data from Orchard B was designated as test set B, collectively employed for model training and testing. The model achieves a detection speed of 32.69 FPS on Jetson Xavier NX, which is six times faster than the traditional DeepLabv3+. On test set A, the mIoU is 95.62%, and on test set B, the mIoU is 92.29%, showing a 1.12% improvement over the traditional DeepLabv3+. These results demonstrate the outstanding performance of the improved DeepLabv3+ model in segmenting fruit tree canopies under different conditions, thus enabling precise spraying by orchard spraying robots. Full article

The maize field environment is complex. Weeds and maize have similar colors and may overlap, and lighting and weather conditions vary. Thus, many methods for the automated differentiation of maize and weeds achieve poor segmentation or cannot be used in real time. In this paper, a weed recognition model based on improved Swin-Unet is proposed. The model first performs semantic segmentation of maize seedlings and uses the resulting mask to identify weeds. U-Net acts as the semantic segmentation framework, and a Swin transformer module is introduced to improve performance. DropBlock regularization, which randomly hides some blocks in crop feature maps, is applied to enhance the generalization ability of the model. Finally, weed areas are identified and segmented with the aid of an improved morphological processing algorithm. The DeepLabv3+, PSANet, Mask R-CNN, original Swin-Unet, and proposed models are trained on a dataset of maize seedling images. The proposed Swin-Unet model outperforms the others, achieving a mean intersection over union of 92.75%, mean pixel accuracy of 95.57%, and inference speed of 15.1 FPS. Our model could be used for accurate, real-time segmentation of crops and weeds and as a reference for the development of intelligent agricultural equipment. Full article

The suspension velocity is the core of the cleaning and sorting mechanisms that utilize a combination of a fan and vibrating sieve. To investigate this, various experimental subjects, such as peanuts with different kernels and clay-heavy clods in different states, were used. The experiment involved simulating the suspension velocity of materials through numerical calculations using fluid dynamics and particle discrete element coupling. The Eularian model was employed to study the coupled gas-solid two-phase flow. The experiment measured the suspension velocities of single and double kernel peanuts, which were found to be 8.34~9.40 m/s and 8.13~9.51 m/s, respectively. Under 20.4% water content and lumpy conditions, the suspension velocities of smaller clods, side by side clods, and larger clods were 12.61~14.30 m/s, 14.16~15.76 m/s and 16.44~18.72 m/s, respectively; under 20.4% water content and smaller clods, the suspension velocities of lumpy and strip of clods were 12.61~14.30 m/s, 11.90~14.13 m/s, respectively; under lumpy and smaller clods, the suspension velocity at 17.6%, 20.4%, and 23.9% water content ranged from 12.38 to 14.20 m/s, 12.61 to 14.30 m/s, and 12.62 to 14.49 m/s, respectively. The simulations showed that the suspension velocity for different types of peanuts, clod sizes, shapes, and water contents was less different from the actual experiments. Specifically, the relative errors in suspension velocity for single-kernel peanuts, double-kernel peanuts, smaller clods, side-by-side clods, larger clods, lumpy clods, strips of clods, and clods with 17.3%, 20.4%, and 23.9% water content were 1.2%, 4.1%, 0.4%, 2.0%, 4.4%, 0.4%, 5.1%, 5.4%, 0.4%, and 1.9%, respectively, compared to actual experiment measurements. The results indicate a significant difference in the suspension velocity between peanuts and clay-heavy clods, which can be distinguished from each other based on this difference. Furthermore, the simulation results have been found to be consistent with the experimental results, thus verifying the feasibility of measuring the material suspension velocity using CFD-DEM gas-solid coupling. Full article

High-performance fertilization equipment with high uniformity is essential for the improvement of fertilizer use efficiency in paddies. The performance of these fertigation systems might be affected by the initial field conditions and fertilizer doses. In this study, the uniformity of fertilization by an automatic system (SF) was investigated; the investigation had two initial field water conditions and fertilizer doses, and manual fertilization by farmers (FF) was used as the control. After fertilization, the Christiansen uniformity coefficient (CU) in the SF paddies was higher than in the FF paddies, and the SF in the non-flooded paddies (SFN) was the highest. With time, the CU of treatments with poor fertilization uniformity was improved; it was driven by the osmotic potential of fertilizer ions, but it was far from exceeding that of the treatments originally conducted with higher CU. For the SF treatments, the fertilizer dose did not affect fertilization uniformity significantly; so, an SF can match the efficient fertilization strategies more precisely. As water-saving irrigation (WSI) is conducive to the production of non-flooded field conditions and the promotion of the efficient use of topdressing, the use of automatic fertilization systems to implement efficient fertilization management practices in WSI paddy fields is suggested. Full article

Nitrogen is one of the most important nutrients affecting the growth and fruit quality of walnut trees. Rapid and accurate monitoring of nitrogen concentration in the walnut canopy can provide a digital basis for its nutritional diagnosis and precision fertilization. Consequently, the main purpose of this study was to use Unmanned Aerial Vehicle (UAV) remote sensing technology to monitor the nitrogen concentration of walnut canopies. In this study, UAV multispectral images of the canopies of nine walnut orchards with different management levels in Wensu County, South Xinjiang, China, were collected during the fast-growing (20 May), sclerotization (25 June), and near-maturity (27 August) periods of walnut fruit, and canopy nitrogen concentration data for 180 individual plants were collected during the same periods. The validity of the information extracted via the outline canopy and simulated canopy methods was compared. The accuracy of nitrogen concentration inversion for three modeling methods, partial least squares regression (PLSR), support vector machine (SVM), and random forest (RF), was analyzed; the effects of different combinations of variables on model accuracy were compared; and the spatial distribution of the nitrogen concentration in the walnut canopy was numerically mapped using the optimal model. The results showed that the accuracy of the model created using the single plant information extracted from the outlined canopy was better than that of the simulated canopy method, but the simulated canopy method was more efficient in extracting effective information from the single plant canopy than the outlined canopy. The simulated canopy method overcame the difficulty of mismatching the spectral information of individual plants extracted, by outlining the canopy in the original image for nitrogen distribution mapping with the spectral information of image elements in the original resolution image. The prediction accuracy of the RF model was better than that of the SVM and PLSR models; the prediction accuracy of the model using a combination of waveband texture information and vegetation index texture information was better than that of the single-source model. The coefficients of determination (R²) values of the RF prediction model built using the band texture information extracted via the simulated canopy method with the vegetation index texture information were in the range of 0.61–0.84, the root mean square error (RMSE) values were in the range of 0.27–0.43 g kg⁻¹, and the relative analysis error (RPD) values were in the range of 1.58–2.20. This study shows that it is feasible to monitor the nitrogen concentration of walnut tree canopies using UAV multispectral remote sensing. This study provides a theoretical basis and methodological reference for the rapid monitoring of nutrients in fruit trees in southern Xinjiang. Full article

Counting the soybean pods automatically has been one of the key ways to realize intelligent soybean breeding in modern smart agriculture. However, the pod counting accuracy for whole soybean plants is still limited due to the crowding and uneven distribution of pods. In this paper, based on the VFNet detector, we propose a deformable attention recursive feature pyramid network for soybean pod counting (DARFP-SD), which aims to identify the number of soybean pods accurately. Specifically, to improve the feature quality, DARFP-SD first introduces the deformable convolutional networks (DCN) and attention recursive feature pyramid (ARFP) to reduce noise interference during feature learning. DARFP-SD further combines the Repulsion Loss to correct the error of predicted bboxse coming from the mutual interference between dense pods. DARFP-SD also designs a density prediction branch in the post-processing stage, which learns an adaptive soft distance IoU to assign suitable NMS threshold for different counting scenes with uneven soybean pod distributions. The model is trained on a dense soybean dataset with more than 5300 pods from three different shapes and two classes, which consists of a training set of 138 images, a validation set of 46 images and a test set of 46 images. Extensive experiments have verified the performance of proposed DARFP-SD. The final training loss is 1.281, and an average accuracy of 90.35%, an average recall of 85.59% and a F1 score of 87.90% can be achieved, outperforming the baseline method VFNet by 8.36%, 4.55% and 7.81%, respectively. We also validate the application effect for different numbers of soybean pods and differnt shapes of soybean. All the results show the effectiveness of the DARFP-SD, which can provide a new insight into the soybean pod counting task. Full article

Litchi leaf diseases and pests can lead to issues such as a decreased Litchi yield, reduced fruit quality, and decreased farmer income. In this study, we aimed to explore a real-time and accurate method for identifying Litchi leaf diseases and pests. We selected three different orchards for field investigation and identified five common Litchi leaf diseases and pests (Litchi leaf mite, Litchi sooty mold, Litchi anthracnose, Mayetiola sp., and Litchi algal spot) as our research objects. Finally, we proposed an improved fully convolutional one-stage object detection (FCOS) network for Litchi leaf disease and pest detection, called FCOS for Litch (FCOS-FL). The proposed method employs G-GhostNet-3.2 as the backbone network to achieve a model that is lightweight. The central moment pooling attention (CMPA) mechanism is introduced to enhance the features of Litchi leaf diseases and pests. In addition, the center sampling and center loss of the model are improved by utilizing the width and height information of the real target, which effectively improves the model’s generalization performance. We propose an improved localization loss function to enhance the localization accuracy of the model in object detection. According to the characteristics of Litchi small target diseases and pests, the network structure was redesigned to improve the detection effect of small targets. FCOS-FL has a detection accuracy of 91.3% (intersection over union (IoU) = 0.5) in the images of five types of Litchi leaf diseases and pests, a detection rate of 62.0/ms, and a model parameter size of 17.65 M. Among them, the detection accuracy of Mayetiola sp. and Litchi algal spot, which are difficult to detect, reached 93.2% and 92%, respectively. The FCOS-FL model can rapidly and accurately detect five common diseases and pests in Litchi leaf. The research outcome is suitable for deployment on embedded devices with limited resources such as mobile terminals, and can contribute to achieving real-time and precise identification of Litchi leaf diseases and pests, providing technical support for Litchi leaf diseases’ and pests’ prevention and control. Full article

Litchi florescence has large flower spikes and volume; reasonable control of the ratio of male to female litchi flowers is the key operational aspect of litchi orchards for preserving quality and increasing production. To achieve the rapid detection of male and female litchi flowers, reduce manual statistical errors, and meet the demand for accurate fertilizer regulation, an intelligent detection method for male and female litchi flowers suitable for deployment to low-power embedded platforms is proposed. The method uses multi-teacher pre-activation feature distillation (MPFD) and chooses the relatively complex YOLOv4 and YOLOv5-l as the teacher models and the relatively simple YOLOv4-Tiny as the student model. By dynamically learning the intermediate feature knowledge of the different teacher models, the student model can improve its detection performance by meeting the embedded platform application requirements such as low power consumption and real-time performance. The main objectives of this study are as follows: optimize the distillation position before the activation function (pre-activation) to reduce the feature distillation loss; use the LogCosh-Squared function as the distillation distance loss function to improve distillation performance; adopt the margin-activation method to improve the features of the teacher model passed to the student model; and propose to adopt the Convolution and Group Normalization (Conv-GN) structure for the feature transformation of the student model to prevent effective information loss. Moreover, the distilled student model is quantified and ported for deployment to a field-programmable gate array (FPGA)-embedded platform to design and implement a fast, intelligent detection system for male and female litchi flowers. The experimental results show that compared with an undistilled student model, the mAP of the student model obtained after MPFD feature distillation is improved by 4.42 to 94.21%; the size of the detection model ported and deployed to the FPGA-embedded platform is 5.91 MB, and the power consumption is only 10 W, which is 73.85% and 94.54% lower than that of the detection models on the server and PC platforms, respectively, and it can better meet the application requirements of rapid detection and accurate statistics of male and female litchi flowers. Full article

Citrus psyllid is the main vector of Huanglongbing, and as such, it is responsible for huge economic losses across the citrus industry. The small size of this pest, difficulties in data acquisition, and the lack of target detection algorithms suitable for complex occlusion environments inhibit detection of the pest. The present paper describes the construction of a standard sample database of citrus psyllid in multi-focal lengths and out-of-focus states in the natural environment. By integrating the attention mechanism and optimizing the key module of BottleneckCSP, YOLOv5s-BC, we have created an accurate detection algorithm for small targets. Based on YOLOv5s, our algorithm incorporates an SE-Net channel attention module into the Backbone network and improves the detection of small targets by guiding the algorithm to the channel characteristics of small-target information. At the same time, the BottleneckCSP module in the neck network is improved, and extraction of multiple features of recognition targets is improved by the addition of a normalization layer and SiLU activation function. Experimental results based on a standard sample database show the recognition accuracy (intersection over union (IoU) = 0.5) of the YOLOv5s-BC algorithm for citrus psyllid to be 93.43%, 2.41% higher than that of traditional YOLOv5s. The accuracy and recall rates are also increased by 1.31% and 4.22%, respectively. These results confirm that the YOLOv5s-BC algorithm has good generalization ability in the natural context of citrus orchards, and it offers a new approach for the control of citrus psyllid. Full article

Detecting litchis in a complex natural environment is important for yield estimation and provides reliable support to litchi-picking robots. This paper proposes an improved litchi detection model named YOLOv5-litchi for litchi detection in complex natural environments. First, we add a convolutional block attention module to each C3 module in the backbone of the network to enhance the ability of the network to extract important feature information. Second, we add a small-object detection layer to enable the model to locate smaller targets and enhance the detection performance of small targets. Third, the Mosaic-9 data augmentation in the network increases the diversity of datasets. Then, we accelerate the regression convergence process of the prediction box by replacing the target detection regression loss function with CIoU. Finally, we add weighted-boxes fusion to bring the prediction boxes closer to the target and reduce the missed detection. An experiment is carried out to verify the effectiveness of the improvement. The results of the study show that the mAP and recall of the YOLOv5-litchi model were improved by 12.9% and 15%, respectively, in comparison with those of the unimproved YOLOv5 network. The inference speed of the YOLOv5-litchi model to detect each picture is 25 ms, which is much better than that of Faster-RCNN and YOLOv4. Compared with the unimproved YOLOv5 network, the mAP of the YOLOv5-litchi model increased by 17.4% in the large visual scenes. The performance of the YOLOv5-litchi model for litchi detection is the best in five models. Therefore, YOLOv5-litchi achieved a good balance between speed, model size, and accuracy, which can meet the needs of litchi detection in agriculture and provides technical support for the yield estimation and litchi-picking robots. Full article

Since plug seedlings play a key role in automatic transplanting, this work aimed to explore the interaction between the biochar rate, water content, and N–fertilization in the substrate on the cultivation of cucumber seedlings before and after transplanting. The research showed that most of the factors obtained significant individual and interaction effects by measuring and analyzing the growth parameters of seedlings before transplanting. Most growth parameters significantly decreased with the increase in biochar rate except Water Use Efficiency which obtained the highest value of 2.06 g/L when the biochar rate was 10%. Furthermore, some growth parameters increased significantly with the increase in water content, while the Total Dry Matter and Water Use Efficiency reached their highest values, 0.778 g and 1.94 g/L, respectively, when the water content was 65%. All growth parameters reached their highest values when the N–fertilization was 50%; too high or too low of N–fertilization was not conducive to the growth of seedlings. The growth parameters and photosynthesis indices of seedlings cultivated after transplanting indicated that the seedlings with superior growth before transplanting performed better than other treatments in regard to growth and photosynthesis after transplanting. The interactions were in general optimal when the biochar rate was 5%, water content was 80%, and N–fertilization was 50% in the substrate, and seedlings cultivated under this treatment could not only meet the requirements of automatic transplanting, but also ensure rapid growth after transplanting. This study thus provides some guidance for the effective cultivation of vegetable plug seedlings. Full article

It is necessary to develop automatic picking technology to improve the efficiency of litchi picking, and the accurate segmentation of litchi branches is the key that allows robots to complete the picking task. To solve the problem of inaccurate segmentation of litchi branches under natural conditions, this paper proposes a segmentation method for litchi branches based on the improved DeepLabv3+, which replaced the backbone network of DeepLabv3+ and used the Dilated Residual Networks as the backbone network to enhance the model’s feature extraction capability. During the training process, a combination of Cross-Entropy loss and the dice coefficient loss was used as the loss function to cause the model to pay more attention to the litchi branch area, which could alleviate the negative impact of the imbalance between the litchi branches and the background. In addition, the Coordinate Attention module is added to the atrous spatial pyramid pooling, and the channel and location information of the multi-scale semantic features acquired by the network are simultaneously considered. The experimental results show that the model’s mean intersection over union and mean pixel accuracy are 90.28% and 94.95%, respectively, and the frames per second (FPS) is 19.83. Compared with the classical DeepLabv3+ network, the model’s mean intersection over union and mean pixel accuracy are improved by 13.57% and 15.78%, respectively. This method can accurately segment litchi branches, which provides powerful technical support to help litchi-picking robots find branches. Full article

Grain quality involves the appearance, nutritional, and safety attributes of grains. With the improvement of people’s living standards, problems pertaining to the quality of grains have received greater attention. Modern quality detection techniques feature unique advantages including rapidness, non-destructiveness, accuracy, and efficiency in detecting grain quality. This review summarizes research progress of these techniques in detection of quality indices of grains. Particularly, the review focuses on detection techniques based on physical properties including acoustic, optical, thermal, electrical, and mechanical properties, and those simulating sensory analysis such as electronic noses, electronic tongues, and electronic eyes. According to the current technological development and application, the challenges and prospects of these techniques are demonstrated. Full article