查看更多>>摘要：? 2022 Elsevier B.V.It is difficult to identify similar apples diseases due to the complicated changes in color and texture of diseased parts. In order to solve this problem, an Internet of Things (IoT) system for apple disease detection based on deep multi-scale dual-channel convolutional neural network (DMCNN) was proposed in this paper. Firstly, the image was transformed into HSV and RGB color subspaces through color space transformation, the color and texture features of apple diseases were extracted respectively. Then, (1) The Color Analysis Subnet of HSV color subspace was proposed to extract the color features. (2) The Texture Analysis Subnet of RGB color subspace was proposed to extract the texture features. The attention mechanism optimized by double-factor weight was used to effectively improve the capability of texture feature extraction of this subnet. (3) DMCNN was constructed through a cross-fusing mechanism of homologous features. It can fuse the features that are extracted by color and texture analysis subnets, thereby improving its expression. Finally, an IoT detection system was constructed by combining hardware and detection model. The Experiments conducted on our self-collected database (Images were taken in natural light using a Nikon camera. After data enhancement, 1674 in total, 1341 training, 332 testing) and other scholars' database (After data enhancement, 3336 in total, 2669 training, 667 testing) show that the proposed DMCNN has attained a high detection rate that exceeds 99.5% on average.
查看更多>>摘要：? 2022 Elsevier B.V.Crop diseases and insect pests are a serious natural disaster, which needs to be predicted and monitored in time to ensure the output of crops. Due to the wide variety of pests and the similar morphology of crops in the early stages of growth, it is difficult for agricultural workers to accurately identify various types of pests. Crop insects have brought huge challenges to the prevention and control of plant diseases and insect pests. In response to this problem, we propose a way of classification of crop pests based on multi-scale feature fusion(MFFNet) to accurately recognizes and classifies crop pests. First, the multi-scale feature extraction module (MFE) is designed by using dilated convolution to obtain the multi-scale feature map of the pest image. At the same time, extracted the deep feature information of the image by the feature extraction module (DFE). Finally, the features extracted separately by the multi-scale feature extraction module (MFE) and the feature extraction module (DFE) were fused thus achieving accurately classified and identified the crops insects by the way of end-to-end. Experiments show that our proposed method has obtained excellent classification performance on the dataset of 12 types of pests, its classification accuracy rate (ACC) reached 98.2%.
查看更多>>摘要：? 2022 Elsevier B.V.Counting plant flowers is a common task with applications for estimating crop yields and selecting favorable genotypes. Typically, this requires a laborious manual process, rendering it impractical to obtain accurate flower counts throughout the growing season. The model proposed in this study uses weak supervision, based on Convolutional Neural Networks (CNNs), which automates such a counting task for cotton flowers using imagery collected from an unmanned aerial vehicle (UAV). Furthermore, the model is trained using Multiple Instance Learning (MIL) in order to reduce the required amount of annotated data. MIL is a binary classification task in which any image with at least one flower falls into the positive class, and all others are negative. In the process, a novel loss function was developed that is designed to improve the performance of image-processing models that use MIL. The model is trained on a large dataset of cotton plant imagery which was collected over several years and will be made publicly available. Additionally, an active-learning-based approach is employed in order to generate the annotations for the dataset while minimizing the required amount of human intervention. Despite having minimal supervision, the model still demonstrates good performance on the testing dataset. Multiple models were tested with different numbers of parameters and input sizes, achieving a minimum average absolute count error of 2.43. Overall, this study demonstrates that a weakly-supervised model is a promising method for solving the flower counting problem while minimizing the human labeling effort.
查看更多>>摘要：? 2022 Elsevier B.V.Training of convolutional neural networks for semantic segmentation of fruit tree branches requires accurate pixel-wise labeling which requires large amounts of human effort. The human-in-the-loop method, where a human annotator corrects the outputs of a neural network, reduces labeling effort; however, it requires human intervention for each image. This paper describes an iterative training methodology for semantic segmentation, Automating-the-Loop. This aims to replicate the manual adjustments of the human-in-the-loop method with an automated process, hence, drastically reducing labeling effort. Using the application of detecting partially occluded apple tree segmentation, we compare manually labeled annotations, self-training, human-in-the-loop, and Automating-the-Loop methods in both the quality of the trained convolutional neural networks, and the effort needed to create them. The convolutional neural network (U-Net) performance is analyzed using traditional metrics and a new metric, Complete Grid Scan. It is shown that in our application, the new Automating-the-Loop method greatly reduces the labeling effort while producing comparable performance to both human-in-the-loop and complete manual labeling methods.
查看更多>>摘要：? 2022Fish locomotion analysis can provide important information for aquaculture managers to make effective production decision. However, the existing aquaponic system work mainly focuses on the transformation of nitrogen elements and are limited because their complex inspection procedures, it is rarely reported to explore effects of aquaponic system on fish locomotion. Using automated methods to explore the effect of aquaponic system on fish locomotion is a challenging task. To address this research gap, we examined and compared fish locomotion in aquaculture with different systems, i.e., aquaponic system and recirculating aquaculture system based on computer vision. The statistical results of fish locomotion in the two systems showed that the fish locomotion in aquaponic systems is significantly higher than that in recirculating aquaculture system, primarily due to the certain purification effect of plants and microorganisms on water quality. In addition, under the same conditions in the aquaculture system aquaponic system had stronger nitrogen absorption capacity than recirculating aquaculture system, the average multiple of the water quality parameter concentration of nitrate nitrogen, total nitrogen and total phosphorus in RAS are 1.79, 1.58, 1.81 times that of APS, from the stability of the experimental system to the end of the experiment. The error of our proposed method of calculating fish locomotion is less than 5%. Overall, this is the first automated intensive monitoring method that examined the role aquaponic system played in fish locomotion, which could assess fish health based on the locomotion to further fine tune the planting density of aquaponics to achieve the maximum benefit.
查看更多>>摘要：? 2022 Elsevier B.V.The constitutive model of corn has an intense influence on the mechanical structure design and machine performance. In this paper, a creep property formula of corn-based on the Burgers model was proposed, which described the change process of the stress–strain trait of corn grain with time. Then, the mathematical model of the “corn-mechanical” interaction system was then established based on the Hertz-Mindlin no-slip model. The kinematic and dynamic formulas of corn ear under the action of threshers were deduced, and the “corn-corn” contact characteristics based on Hertz-Mindlin with bonding model were analyzed. Moreover, a new discrete element model construction method of the corn ear was proposed through particle filling and splicing methods. The simulation experiment was carried out in EDEM to verify the constitutive model and discrete element model of corn ear. The results showed that the particle filling model could better explain the mechanism of corn ear threshing and obtain the force of each corn grain during threshing. The average velocity of corn ear was 55.2 m/s, the maximum value was 216.7 m/s, and the minimum value was 16.2 m/s, while the average normal force was 201.7 N, the maximum force was 342.0 N, and the minimum force was 104.2 N respectively, it verified the feasibility and accuracy of models. Finally, the field experiment was carried out to verify the simulation results, and the corn grain broken rate was 2.16%. These results fill the blank of fundamental theoretical research on the design of corn ear threshing devices and provide a reference for the optimal design of other crop production equipment.
查看更多>>摘要：? 2022 Elsevier B.V.Stress-crack detection is important for determining seed quality. This paper presents both a machine-vision-based method and a prototype of an industrial hardware design for stress-crack detection in maize kernels. Specifically, we present a cascade model incorporating kernel-status classification, region-of-interest segmentation, and crack-detection models. The status-classification model selects kernels with the correct camera orientation, whereas the region-of-interest segmentation model locates the main axis of the kernel and supplies a kernel mask for endosperm segmentation. Further, we utilise the EDLines algorithm to detect cracks and apply three novel constraints to distinguish real cracks from noise. The results of experiments conducted indicate that our integrated hardware–software system can detect stress cracks in maize kernels effectively and automatically. The observed precision and recall of the overall system were 92.7% and 94.4%, respectively.
查看更多>>摘要：? 2022 Elsevier B.V.Watermelon cultivators often encounter various challenges of the varietal mixing of triploid, diploid, and tetraploid seeds, thus hindering the watermelon industry due to the uncertainty in the ploidy seed nomenclature. These circumstances indirectly impose negative effects on the income of farmers and the development of companies specializing in watermelon seeds. Therefore, high seed purity is a necessity for all seed breeders and firms, as the performance of a given seed variety can be standardized. In this study, we employed machine vision techniques to classify triploid watermelon seeds from diploid and tetraploid seeds. The major objective of the research was to illustrate the potential of the discrimination of triploid watermelon seeds with multivariate machine learning classification, and, thereafter, deep learning techniques. Watermelon ploidy seed images were acquired by RGB camera, and discrimination models were constructed with multivariate machine learning methods using one-class classification with the DD-SIMCA and SVM quadratic methods. One-class classification with the DD-SIMCA and the SVM-quadratic models yielded triploid discrimination accuracies of 69.5% and 84.3%, respectively. To further improve the ploidy-class discrimination accuracy, deeplabv3 + and Resnet18 deep learning models produced accuracy of 95.5%. The deep learning model results demonstrated a higher discrimination accuracy, and, thus, these results show the potential for automation and application to online systems for real-time ploidy seed discrimination and sorting.
查看更多>>摘要：? 2022 Elsevier B.V.In recent years, deep learning has greatly improved the performance of named entity recognition models in various fields, especially in the agricultural domain. However, most existing works only utilize word embedding models to generate the context-independent embeddings, which is limited in modeling polysemous words. Moreover, the abundant morphological information in agricultural texts has not been fully utilized. Besides, the local context information needs to be further extracted. To solve the aforementioned issues, a novel enhanced contextual embeddings and glyph features-based model was proposed. First, the contextual embeddings were dynamically generated by the fine-tuned Bidirectional Encoder Representation from Transformers (BERT) on the domain-specific corpus (e.g., agricultural texts), and then the multi-granularity information was obtained from the layers of BERT. Thus, the contextual embeddings not only contain domain-specific knowledge but also include multi-grained semantic information. Second, a novel 3-dimension convolutional neural network-based framework was designed to capture the contextual glyph features for each character from the image perspective. Third, a channel-wised fusion architecture was also introduced to further improve the ability of the convolutional neural network layer to capture local context features. Experimental results showed that our proposed model achieved the best F1-scores of 95.02% and 96.51% on AgCNER and Resume datasets, which indicated the effectiveness and generalization of our model to identify the entities in cross-domain texts. The ablation study in many aspects also demonstrated the better performance of the proposed model.
查看更多>>摘要：? 2022 Elsevier B.V.Meeting an increasing demand for food while preserving the environment is one of the most important challenges of the 21st century. To meet this challenge, conservation agriculture can rely on the age-old practice of crop rotation. The objective of this article is to develop a methodology for predicting and visualizing crop rotations, supporting discussions between agronomists and producers. Based on crop history data, the 6-phase methodology, uses Markov chains for the prediction of the N most likely crops grown in year n + 1. Process mining and Directly-Follows Graphs (DFG) enables modelling and visualization of the results. Generalisation and filtering operations highlight the frequent behaviors of producers. Applied to analyse the crop history of 10,376 fields from 409 field crop farms in Quebec, Canada, the methodology is competitive with the performance of various recurrent neural networks (LSTM, RNN, GRU) with a successful prediction rate that exceeds 90%, while allowing for an intelligibility of results and a relative computational simplicity.
NSTL
Elsevier