Assessing the efficiency of UAV for pesticide application in disease management of peanut crop.

BACKGROUND: Effective utilization of plant protection UAVs in peanut cultivation management necessitates a comprehensive grasp of how application volume rates and pesticides influence peanut leaf spot and rust control. This study aimed to compare the effects of application volume rates and pesticides on droplet deposition, disease, leaf retention rate and peanut yield. A T20 plant protection unmanned aerial vehicle (UAV) sprayer was used to apply four various pesticide doses. In comparison, a knapsack sprayer was used to spray with an application volume rate of 450 L ha-1. RESULTS: The results showed a significant difference in droplet deposition between the plant protection UAVs and the electric knapsack sprayer. In the pesticide treatment with an application volume rate of 15.0 L ha-1, there was no significant difference in the deposition on the peanut canopy of each pesticide treatment, but there was a significant difference in the deposition on the ground in the treatment with adding vegetable oil adjuvant. The treatment with added vegetable oil additives showed the worst performance. The treatment with an application volume rate of 22.5 L ha-1 showed the best performance, with the leaf spot control effect being only 0.3% lower than that of the electric knapsack sprayer. CONCLUSION: Plant protection UAV spraying is feasible to control peanut diseases. Considering the operational effectiveness of the plant protection UAV and application volume rate, it is recommended to use an application volume rate of 22.5 L ha-1 without adding vegetable oil adjuvants for field operations. © 2024 Society of Chemical Industry.

Evaluating the use of unmanned aerial vehicles for spray applications in mountain Nanguo pear orchards.

BACKGROUND: Nanguo pear is a distinctive pear variety in northeast China, grown mainly in mountainous areas. Due to terrain limitations, ground-based pesticide application equipment is difficult to use. This limitation could be overcome by using unmanned aerial vehicles (UAVs) for pesticide application in Nanguo pear orchards. This study evaluated the spraying performance of two UAVs in the Nanguo pear orchards and compared them with a manually used backpack electric sprayer (BES). The study also analyzed the effect of canopy size on droplet deposition and ground loss, and evaluated two sampling methods, leaf sampling and telescopic rod sampling. RESULTS: Compared to BESs, droplet deposition is lower for UAVs, but the actual pesticide active ingredient deposition is not necessarily lower given the solution concentration. The droplet deposition varies among different UAVs due to structural differences. Under the same UAV operating parameters, droplet deposition on trees with smaller canopy sizes is typically greater than that on trees with larger canopy sizes, and the ground loss was also more severe. Although telescopic rod sampling is a quick and convenient method, it can only reflect the trend of droplet deposition, and the data error is greater compared with leaf sampling. CONCLUSION: UAVs can achieve better droplet deposition in mountainous Nanguo pear orchards and does almost no harm to the operators compared with the BES. However, canopy size needs to be considered to adjust the application volume rate. Telescopic rods can be used for qualitative analyses, but are not recommended for quantitative analyses. © 2024 Society of Chemical Industry.

Triboelectric micro-flexure-sensitive fiber electronics.

Developing fiber electronics presents a practical approach for establishing multi-node distributed networks within the human body, particularly concerning triboelectric fibers. However, realizing fiber electronics for monitoring micro-physiological activities remains challenging due to the intrinsic variability and subtle amplitude of physiological signals, which differ among individuals and scenarios. Here, we propose a technical approach based on a dynamic stability model of sheath-core fibers, integrating a micro-flexure-sensitive fiber enabled by nanofiber buckling and an ion conduction mechanism. This scheme enhances the accuracy of the signal transmission process, resulting in improved sensitivity (detectable signal at ultra-low curvature of 0.1 mm-1; flexure factor >21.8% within a bending range of 10°.) and robustness of fiber under micro flexure. In addition, we also developed a scalable manufacturing process and ensured compatibility with modern weaving techniques. By combining precise micro-curvature detection, micro-flexure-sensitive fibers unlock their full potential for various subtle physiological diagnoses, particularly in monitoring fiber upper limb muscle strength for rehabilitation and training.

Electronic eye and electronic tongue data fusion combined with a GETNet model for the traceability and detection of Astragalus.

BACKGROUND: Astragalus is a widely used traditional Chinese medicine material that is easily confused due to its quality, price and other factors derived from different origins. This article describes a novel method for the rapid tracing and detection of Astragalus via the joint application of an electronic tongue (ET) and an electronic eye (EE) combined with a lightweight convoluted neural network (CNN)-transformer model. First, ET and EE systems were employed to measure the taste fingerprints and appearance images, respectively, of different Astragalus samples. Three spectral transform methods - the Markov transition field, short-time Fourier transform and recurrence plot - were utilized to convert the ET signals into 2D spectrograms. Then, the obtained ET spectrograms were fused with the EE image to obtain multimodal information. A lightweight hybrid model, termed GETNet, was designed to achieve pattern recognition for the Astragalus fusion information. The proposed model employed an improved transformer module and an improved Ghost bottleneck as its backbone network, complementarily utilizing the benefits of CNN and transformer architectures for local and global feature representation. Furthermore, the Ghost bottleneck was further optimized using a channel attention technique, which boosted the model's feature extraction effectiveness. RESULTS: The experiments indicate that the proposed data fusion strategy based on ET and EE devices has better recognition accuracy than that attained with independent sensing devices. CONCLUSION: The proposed method achieved high precision (99.1%) and recall (99.1%) values, providing a novel approach for rapidly identifying the origin of Astragalus, and it holds great promise for applications involving other types of Chinese herbal medicines. © 2024 Society of Chemical Industry.

Zinc-oxide nanoparticles ameliorated the phytotoxic hazards of cadmium toxicity in maize plants by regulating primary metabolites and antioxidants activity.

Cadmium stress is a major threat to plant growth and survival worldwide. The current study aims to green synthesis, characterization, and application of zinc-oxide nanoparticles to alleviate cadmium stress in maize (Zea mays L.) plants. In this experiment, two cadmium levels (0, 0.6 mM) were applied to check the impact on plant growth attributes, chlorophyll contents, and concentration of various primary metabolites and antioxidants under exogenous treatment of zinc-oxide nanoparticles (25 and 50 mg L-1) in maize seedlings. Tissue sampling was made 21 days after the zinc-oxide nanoparticles application. Our results showed that applying cadmium significantly reduced total chlorophyll and carotenoid contents by 52.87% and 23.31% compared to non-stress. In comparison, it was increased by 53.23%, 68.49% and 9.73%, 37.53% with zinc-oxide nanoparticles 25, 50 mg L-1 application compared with cadmium stress conditions, respectively. At the same time, proline, superoxide dismutase, peroxidase, catalase, and ascorbate peroxidase contents were enhanced in plants treated with cadmium compared to non-treated plants with no foliar application, while it was increased by 12.99 and 23.09%, 23.52 and 35.12%, 27.53 and 36.43%, 14.19 and 24.46%, 14.64 and 37.68% by applying 25 and 50 mg L-1 of zinc-oxide nanoparticles dosages, respectively. In addition, cadmium toxicity also enhanced stress indicators such as malondialdehyde, hydrogen peroxide, and non-enzymatic antioxidants in plant leaves. Overall, the exogenous application of zinc-oxide nanoparticles (25 and 50 mg L-1) significantly alleviated cadmium toxicity in maize. It provides the first evidence that zinc-oxide nanoparticles 25 ~ 50 mg L-1 can be a candidate agricultural strategy for mitigating cadmium stress in cadmium-polluted soils for safe agriculture practice.

The effect of Aceria litchii (Keifer) infestation on the surface properties of litchi leaf hosts.

BACKGROUND: The wettability of target crop surfaces affects pesticide wetting and deposition. The structure and properties of the leaf surface of litchi leaves undergo severe changes after infestation by Aceria litchii (Keifer). The objective of this study was to systematically investigate the surface texture and wettability of litchi leaves infested. RESULTS: Firstly, the investigation focused on the surface structure and physicochemical properties of litchi leaves infested with Aceria litchii. Subsequently, different levels of Contact Angle (CA) were measured individually on the infested litchi leaves. Lastly, Surface Free Energy (SFE) and its polar and dispersive components were calculated using the Owens-Wendt-Rabel-Kaelble (OWRK) method. The outcomes revealed distinctive 3D surface structures of the erineum at various stages of mycorrhizal growth. At stage NO. 1, the height of the fungus displayed a peaked appearance, with the skewness value indicating a surface characterized by more crests. In contrast, at stages NO. 2 and NO. 3, the surface appeared relatively flat. Furthermore, post-infestation of litchi leaves, the CA of droplets on the abaxial surface of diseased leaves exhibited an increase, while the SFE value on the abaxial surface of leaves decreased significantly, in contrast to the abaxial surface of healthy leaves. CONCLUSION: The infestation behavior of Aceria litchii changed the surface structure and chemistry of litchi leaves, which directly affected the CA value of foliar liquids and the SFE value of leaves, changing the surface wettability of litchi leaves from hydrophobic to superhydrophobic. This study provides useful information for improving the wetting and deposition behavior of liquid droplets on the surface of infested leaves. © 2024 Society of Chemical Industry.

Visible and NIR microscopic hyperspectrum reconstruction from RGB images with deep convolutional neural networks.

We investigate the microscopic hyperspectral reconstruction from RGB images with a deep convolutional neural network (DCNN) in this paper. Based on the microscopic hyperspectral imaging system, a homemade dataset consisted of microscopic hyperspectral and RGB image pairs is constructed. For considering the importance of spectral correlation between neighbor spectral bands in microscopic hyperspectrum reconstruction, the 2D convolution is replaced by 3D convolution in the DCNN framework, and a metric (weight factor) used to evaluate the performance reconstructed hyperspectrum is also introduced into the loss function used in training. The effects of the dimension of convolution kernel and the weight factor in the loss function on the performance of the reconstruction model are studied. The overall results indicate that our model can show better performance than the traditional models applied to reconstruct the hyperspectral images based on DCNN for the public and the homemade microscopic datasets. In addition, we furthermore explore the microscopic hyperspectrum reconstruction from RGB images in infrared region, and the results show that the model proposed in this paper has great potential to expand the reconstructed hyperspectrum wavelength range from the visible to near infrared bands.

Tryptophan Seed Treatment Improves Morphological, Biochemical, and Photosynthetic Attributes of the Sunflower under Cadmium Stress.

Tryptophan, as a signal molecule, mediates many biotic and environmental stress-induced physiological responses in plants. Therefore, an experiment was conducted to evaluate the effect of tryptophan seed treatment in response to cadmium stress (0, 0.15, and 0.25 mM) in sunflower plants. Different growth and biochemical parameters were determined to compare the efficiency of the treatment agent. The results showed that cadmium stress reduced the growth attributes, including root and shoot length, dry and fresh weight, rate of seed germination, and the number of leaves. Cadmium stress also significantly reduced the contents of chlorophyll a, b, and total chlorophyll, carotenoid contents, phenolics, flavonoids, anthocyanin, and ascorbic acid. Whereas cadmium stress (0.15 and 0.25 mM) enhanced the concentrations of malondialdehyde (45.24% and 53.06%), hydrogen peroxide (-11.07% and 5.86%), and soluble sugars (28.05% and 50.34%) compared to the control. Tryptophan treatment decreased the effect of Cd stress by minimizing lipid peroxidation. Seed treatment with tryptophan under cadmium stress improved the root (19.40%) and shoot length (38.14%), root (41.90%) and shoot fresh weight (13.58%), seed germination ability (13.79%), average leaf area (24.07%), chlorophyll b (51.35%), total chlorophyll (20.04%), carotenoids (43.37%), total phenolic (1.47%), flavonoids (19.02%), anthocyanin (26.57%), ascorbic acid (4%), and total soluble proteins (12.32%) compared with control conditions. Overall, the tryptophan seed treatment showed positive effects on sunflower plants' growth and stress tolerance, highlighting its potential as a sustainable approach to improve crop performance.

Explanatory Object Part Aggregation for Zero-Shot Learning.

Zero-shot learning (ZSL) aims to recognize objects from unseen classes only based on labeled images from seen classes. Most existing ZSL methods focus on optimizing feature spaces or generating visual features of unseen classes, both in conventional ZSL and generalized zero-shot learning (GZSL). However, since the learned feature spaces are suboptimal, there exists many virtual connections where visual features and semantic attributes are not corresponding to each other. To reduce virtual connections, in this paper, we propose to discover comprehensive and fine-grained object parts by building explanatory graphs based on convolutional feature maps, then aggregate object parts to train a part-net to obtain prediction results. Since the aggregated object parts contain comprehensive visual features for activating semantic attributes, the virtual connections can be reduced by a large extent. Since part-net aims to extract local fine-grained visual features, some attributes related to global structures are ignored. To take advantage of both local and global visual features, we design a feature distiller to distill local features into a master-net which aims to extract global features. The experimental results on AWA2, CUB, FLO, and SUN dataset demonstrate that our proposed method obviously outperforms the state-of-the-arts in both conventional ZSL and GZSL tasks.

Evaluation of the deposition and distribution of spray droplets in citrus orchards by plant protection drones.

Plant protection drone spraying technology is widely used to prevent and control crop diseases and pests due to its advantages of being unaffected by crop growth patterns and terrain restrictions, high operational efficiency, and low labor requirements. The operational parameters of plant protection drones significantly impact the distribution of spray droplets, thereby affecting pesticide utilization. In this study, a field experiment was conducted to determine the working modes of two representative plant protection drones and an electric backpack sprayer as a control to explore the characteristics of droplet deposition with different spray volumes in the citrus canopy. The results showed that the spraying volume significantly affected the number of droplets and the spray coverage. The number of droplets and the spray coverage area on the leaf surface were significantly increased by increasing the spray volume from 60 L/ha to 120 L/ha in plant protection drones. Particularly for the DJI T30, the mid-lower canopy showed a spray coverage increase of 52.5%. The droplet density demonstrated the most significant variations in the lower inner canopy, ranging from 18.7 droplets/cm2 to 41.7 droplets/cm2 by XAG V40. From the deposition distribution on fruit trees, the plant protection drones exhibit good penetration ability, as the droplets can achieve a relatively even distribution in different canopy layers of citrus trees. The droplet distribution uniformity inside the canopy is similar for XAG V40 and DJI T30, with a variation coefficient of approximately 50%-100%. Compared to the plant protection drones, the knapsack electric sprayer is suitable for pest and disease control in the mid-lower canopy, but they face challenges of insufficient deposition capability in the upper canopy and overall poor spray uniformity. The distribution of deposition determined in this study provides data support for the selection of spraying agents for fruit trees by plant protection drones and for the control of different pests and diseases.

An efficient approach to detect and track winter flush growth of litchi tree based on UAV remote sensing and semantic segmentation.

The immature winter flush affects the flower bud differentiation, flowering and fruit of litchi, and then seriously reduces the yield of litchi. However, at present, the area estimation and growth process monitoring of winter flush still rely on manual judgment and operation, so it is impossible to accurately and effectively control flush. An efficient approach is proposed in this paper to detect the litchi flush from the unmanned aerial vehicle (UAV) remoting images of litchi crown and track winter flush growth of litchi tree. The proposed model is constructed based on U-Net network, of which the encoder is replaced by MobeilNetV3 backbone network to reduce model parameters and computation. Moreover, Convolutional Block Attention Module (CBAM) is integrated and convolutional layer is added to enhance feature extraction ability, and transfer learning is adopted to solve the problem of small data volume. As a result, the Mean Pixel Accuracy (MPA) and Mean Intersection over Union (MIoU) on the flush dataset are increased from 90.95% and 83.3% to 93.4% and 85%, respectively. Moreover, the size of the proposed model is reduced by 15% from the original model. In addition, the segmentation model is applied to the tracking of winter flushes on the canopy of litchi trees and investigating the two growth processes of litchi flushes (late-autumn shoots growing into flushes and flushes growing into mature leaves). It is revealed that the growth processes of flushes in a particular branch region can be quantitatively analysed based on the UAV images and the proposed semantic segmentation model. The results also demonstrate that a sudden drop in temperature can promote the rapid transformation of late-autumn shoots into flushes. The method proposed in this paper provide a new technique for accurate management of litchi flush and a possibility for the area estimation and growth process monitoring of winter flush, which can assist in the control operation and yield prediction of litchi orchards.

Global Analysis of Lysine Lactylation of Germinated Seeds in Wheat.

Protein lactylation is a newly discovered posttranslational modification (PTM) and is involved in multiple biological processes, both in mammalian cells and rice grains. However, the function of lysine lactylation remains unexplored in wheat. In this study, we performed the first comparative proteomes and lysine lactylomes during seed germination of wheat. In total, 8000 proteins and 927 lactylated sites in 394 proteins were identified at 0 and 12 h after imbibition (HAI). Functional enrichment analysis showed that glycolysis- and TCA-cycle-related proteins were significantly enriched, and more differentially lactylated proteins were enriched in up-regulated lactylated proteins at 12 HAI vs. 0 HAI through the KEGG pathway and protein domain enrichment analysis compared to down-regulated lactylated proteins. Meanwhile, ten particularly preferred amino acids near lactylation sites were found in the embryos of germinated seeds: AA*KlaT, A***KlaD********A, KlaA**T****K, K******A*Kla, K*Kla********K, KlaA******A, Kla*A, KD****Kla, K********Kla and KlaG. These results supplied a comprehensive profile of lysine lactylation of wheat and indicated that protein lysine lactylation played important functions in several biological processes.

Prediction and visualization of gene modulated ultralow cadmium accumulation in brown rice grains by hyperspectral imaging.

Monitoring (including prediction and visualization) the gene modulated cadmium (Cd) accumulation in rice grains is one of the most important steps for identification of key transporter genes responsible for grain Cd accumulation and breeding low grain-Cd-accumulating rice cultivars. A method to predict and visualize the gene modulated ultralow Cd accumulation in brown rice grains based on the hyperspectral image (HSI) technology is proposed in this study. Firstly, the Vis-NIR HSIs of brown rice grain samples with 48Cd content levels induced by gene modulation (ranging from 0.0637 to 0.1845 mg/kg) are collected using HSI system. Then, Kernel-ridge (KRR) and random forest (RFR) regression models based on full spectral data and the data after feature dimension reduction (FDR) with kernel principal component analysis (KPCA) and truncated singular value decomposition (TSVD) algorithms are established to predict the Cd contents. RFR model shows poor performance due to the over-fitting based on the full spectral data, while the KRR model can obtain a good predict accuracy with Rp2 of 0.9035, RMSEP of 0.0037 and RPD of 3.278. After the FDR of the full spectral data, the RFR model combined with TSVD reaches the optimum prediction accuracy with Rp2 of 0.9056, RMSEP of 0.0074 and RPD of 3.318, and the best prediction precision of KRR model can also be further enhanced by TSVD with Rp2 of 0.9224, RMSEP of 0.0067 and RPD of 3.512. Finally, the visualization of the predicted Cd accumulation in brown rice grains are realized based on the best regression model (KRR + TSVD). The results of this work indicate that Vis-NIR HSI has great potential for detection and visualization gene modulation induced ultralow Cd accumulation and transport in rice crops.

Detection of Rice Pests Based on Self-Attention Mechanism and Multi-Scale Feature Fusion.

In recent years, the occurrence of rice pests has been increasing, which has greatly affected the yield of rice in many parts of the world. The prevention and cure of rice pests is urgent. Aiming at the problems of the small appearance difference and large size change of various pests, a deep neural network named YOLO-GBS is proposed in this paper for detecting and classifying pests from digital images. Based on YOLOv5s, one more detection head is added to expand the detection scale range, the global context (GC) attention mechanism is integrated to find targets in complex backgrounds, PANet is replaced by BiFPN network to improve the feature fusion effect, and Swin Transformer is introduced to take full advantage of the self-attention mechanism of global contextual information. Results from experiments on our insect dataset containing Crambidae, Noctuidae, Ephydridae, and Delphacidae showed that the average mAP of the proposed model is up to 79.8%, which is 5.4% higher than that of YOLOv5s, and the detection effect of various complex scenes is significantly improved. In addition, the paper analyzes and discusses the generalization ability of YOLO-GBS model on a larger-scale pest data set. This research provides a more accurate and efficient intelligent detection method for rice pests and others crop pests.

Evaluation of aerial spraying application of multi-rotor unmanned aerial vehicle for Areca catechu protection.

Multi-rotor unmanned aerial vehicle (UAV) is a new chemical application tool for tall stalk tropical crop Areca catechu, which could improve deposit performance, reduce operator healthy risk, and increase spraying efficiency. In this work, a spraying experiment was carried out in two A. catechu fields with two leaf area index (LAI) values, and different operational parameters were set. Spray deposit quality, spray drift, and ground loss were studied and evaluated. The results showed that the larger the LAI of A. catechu, the lesser the coverage of the chemical deposition. The maximum coverage could reach 4.28% and the minimum 0.33%. At a flight speed of 1.5 m/s, sprayed droplets had the best penetration and worst ground loss. The overall deposition effect was poor when the flight altitudes were greater than 11.09 m and the flight speed was over 2.5 m/s. Comparing flight speed of 2.5 to 1.5 m/s, the overall distance of 90% of the total drift increased to double under the same operating parameters. This study presents reference data for UAV chemical application in A. catechu protection.

Design and validation of a multi-objective waypoint planning algorithm for UAV spraying in orchards based on improved ant colony algorithm.

Introduction: Current aerial plant protection with Unmanned Aerial Vehicles (UAV) usually applies full coverage route planning, which is challenging for plant protection operations in the orchards in South China. Because the fruit planting has the characteristics of dispersal and irregularity, full-coverage route spraying causes re-application as well as missed application, resulting in environmental pollution. Therefore, it is of great significance to plan an efficient, low-consumption and accurate plant protection route considering the flight characteristics of UAVs and orchard planting characteristics. Methods: This study proposes a plant protection route planning algorithm to solve the waypoint planning problem of UAV multi-objective tasks in orchard scenes. By improving the heuristic function in Ant Colony Optimization (ACO), the algorithm combines corner cost and distance cost for multi-objective node optimization. At the same time, a sorting optimization mechanism was introduced to speed up the iteration speed of the algorithm and avoid the influence of inferior paths on the optimal results. Finally, Multi-source Ant Colony Optimization (MS-ACO) was proposed after cleaning the nodes of the solution path. Results: The simulation results of the three test fields show that compared with ACO, the path length optimization rate of MS-ACO are 3.89%, 4.6% and 2.86%, respectively, the optimization rate of total path angles are 21.94%, 45.06% and 55.94%, respectively, and the optimization rate of node numbers are 61.05%, 74.84% and 75.47%, respectively. MS-ACO can effectively reduce the corner cost and the number of nodes. The results of field experiments show that for each test field, MS-ACO has a significant optimization effect compared with ACO, with an optimization rate of energy consumption per meter of more than 30%, the optimization rate of flight time are 46.67%, 56% and 59.01%, respectively, and the optimization rate of corner angle are 50.76%, 61.78% and 71.1%, respectively. Discussion: The feasibility and effectiveness of the algorithm were further verified. The algorithm proposed in this study can optimize the spraying path according to the position of each fruit tree and the flight characteristics of UAV, effectively reduce the energy consumption of UAV flight, improve the operating efficiency, and provide technical reference for the waypoint planning of plant protection UAV in the orchard scene.

A Method of Setting the LiDAR Field of View in NDT Relocation Based on ROI.

LiDAR placement and field of view selection play a role in detecting the relative position and pose of vehicles in relocation maps based on high-precision map automatic navigation. When the LiDAR field of view is obscured or the LiDAR position is misplaced, this can easily lead to loss of repositioning or low repositioning accuracy. In this paper, a method of LiDAR layout and field of view selection based on high-precision map normal distribution transformation (NDT) relocation is proposed to solve the problem of large NDT relocation error and position loss when the occlusion field of view is too large. To simulate the real placement environment and the LiDAR obstructed by obstacles, the ROI algorithm is used to cut LiDAR point clouds and to obtain LiDAR point cloud data of different sizes. The cut point cloud data is first downsampled and then relocated. The downsampling points for NDT relocation are recorded as valid matching points. The direction and angle settings of the LiDAR point cloud data are optimized using RMSE values and valid matching points. The results show that in the urban scene with complex road conditions, there are more front and rear matching points than left and right matching points within the unit angle. The more matching points of the NDT relocation algorithm there are, the higher the relocation accuracy. Increasing the front and rear LiDAR field of view prevents the loss of repositioning. The relocation accuracy can be improved by increasing the left and right LiDAR field of view.

Feature Distillation in Deep Attention Network Against Adversarial Examples.

Deep neural networks (DNNs) are easily fooled by adversarial examples. Most existing defense strategies defend against adversarial examples based on full information of whole images. In reality, one possible reason as to why humans are not sensitive to adversarial perturbations is that the human visual mechanism often concentrates on most important regions of images. A deep attention mechanism has been applied in many computer fields and has achieved great success. Attention modules are composed of an attention branch and a trunk branch. The encoder/decoder architecture in the attention branch has potential of compressing adversarial perturbations. In this article, we theoretically prove that attention modules can compress adversarial perturbations by destroying potential linear characteristics of DNNs. Considering the distribution characteristics of adversarial perturbations in different frequency bands, we design and compare three types of attention modules based on frequency decomposition and reorganization to defend against adversarial examples. Moreover, we find that our designed attention modules can obtain high classification accuracies on clean images by locating attention regions more accurately. Experimental results on the CIFAR and ImageNet dataset demonstrate that frequency reorganization in attention modules can not only achieve good robustness to adversarial perturbations, but also obtain comparable, even higher classification, accuracies on clean images. Moreover, our proposed attention modules can be integrated with existing defense strategies as components to further improve adversarial robustness.

Point Cloud Completion of Plant Leaves under Occlusion Conditions Based on Deep Learning.

The utilization of 3-dimensional point cloud technology for non-invasive measurement of plant phenotypic parameters can furnish important data for plant breeding, agricultural production, and diverse research applications. Nevertheless, the utilization of depth sensors and other tools for capturing plant point clouds often results in missing and incomplete data due to the limitations of 2.5D imaging features and leaf occlusion. This drawback obstructed the accurate extraction of phenotypic parameters. Hence, this study presented a solution for incomplete flowering Chinese Cabbage point clouds using Point Fractal Network-based techniques. The study performed experiments on flowering Chinese Cabbage by constructing a point cloud dataset of their leaves and training the network. The findings demonstrated that our network is stable and robust, as it can effectively complete diverse leaf point cloud morphologies, missing ratios, and multi-missing scenarios. A novel framework is presented for 3D plant reconstruction using a single-view RGB-D (Red, Green, Blue and Depth) image. This method leveraged deep learning to complete localized incomplete leaf point clouds acquired by RGB-D cameras under occlusion conditions. Additionally, the extracted leaf area parameters, based on triangular mesh, were compared with the measured values. The outcomes revealed that prior to the point cloud completion, the R2 value of the flowering Chinese Cabbage's estimated leaf area (in comparison to the standard reference value) was 0.9162. The root mean square error (RMSE) was 15.88 cm2, and the average relative error was 22.11%. However, post-completion, the estimated value of leaf area witnessed a significant improvement, with an R2 of 0.9637, an RMSE of 6.79 cm2, and average relative error of 8.82%. The accuracy of estimating the phenotypic parameters has been enhanced significantly, enabling efficient retrieval of such parameters. This development offers a fresh perspective for non-destructive identification of plant phenotypes.

Citrus green fruit detection via improved feature network extraction.

Introduction: It is crucial to accurately determine the green fruit stage of citrus and formulate detailed fruit conservation and flower thinning plans to increase the yield of citrus. However, the color of citrus green fruits is similar to the background, which results in poor segmentation accuracy. At present, when deep learning and other technologies are applied in agriculture for crop yield estimation and picking tasks, the accuracy of recognition reaches 88%, and the area enclosed by the PR curve and the coordinate axis reaches 0.95, which basically meets the application requirements.To solve these problems, this study proposes a citrus green fruit detection method that is based on improved Mask-RCNN (Mask-Region Convolutional Neural Network) feature network extraction. Methods: First, the backbone networks are able to integrate low, medium and high level features and then perform end-to-end classification. They have excellent feature extraction capability for image classification tasks. Deep and shallow feature fusion is used to fuse the ResNet(Residual network) in the Mask-RCNN network. This strategy involves assembling multiple identical backbones using composite connections between adjacent backbones to form a more powerful backbone. This is helpful for increasing the amount of feature information that is extracted at each stage in the backbone network. Second, in neural networks, the feature map contains the feature information of the image, and the number of channels is positively related to the number of feature maps. The more channels, the more convolutional layers are needed, and the more computation is required, so a combined connection block is introduced to reduce the number of channels and improve the model accuracy. To test the method, a visual image dataset of citrus green fruits is collected and established through multisource channels such as handheld camera shooting and cloud platform acquisition. The performance of the improved citrus green fruit detection technology is compared with those of other detection methods on our dataset. Results: The results show that compared with Mask-RCNN model, the average detection accuracy of the improved Mask-RCNN model is 95.36%, increased by 1.42%, and the area surrounded by precision-recall curve and coordinate axis is 0.9673, increased by 0.3%. Discussion: This research is meaningful for reducing the effect of the image background on the detection accuracy and can provide a constructive reference for the intelligent production of citrus.