Reduce Yield Losses Research Articles

Diagnosis of Major Foliar Diseases in Black gram (Vigna mungo L.) using Convolution Neural Network (CNN)

Background: Proper diagnosis of a foliar disease is a prerequisite to undertaking any crop protection strategy under field conditions. Poor diagnosis and a delay in confirmation in turn decrease the crop yield and increase the cost of plant protection. In this background, advanced machine learning techniques were used for diagnosis of major foliar diseases in black gram using image detection. Casually, black gram yields are highly reduced due to anthracnose and powdery mildew diseases up to 40-67%. To address the issues, the advanced disease identification method of Convolution Neural Network (CNN) is proposed for automated diagnosis in its early stages to assist farmers. Methods: Disease infected leaf samples and their images were collected from different cultivated areas of Tanjore district, Tamil Nadu, India. The image noises were removed and enhanced to improve the accuracy of the training network. A Convolution Neural Network was built with five layers to work on disease images. The first stage of training is to load the image set for training, establish the learning rate, run the optimizer and compile the training convolution model. The final part is to save the loss and accuracy during the training process and evaluate the accuracy of the model. To improve the training learning rate, the Adam optimizer and RMSprop algorithm are used to dynamically adjust the learning rate. The image dataset holds a total of 2002 images of black gram anthracnose and powdery mildew for evaluation. Result: The experiment result showed that the accuracy of disease detection in black gram is about 92.50 per cent with a Precision: 97.14 per cent, Recall: 87.17 per cent, F1 score: 91.89 per cent which proves that convolutional neural network has a faster training speed and higher accuracy. In addition, the proposed method is less time consuming, an early detection tool for the farmers to identify the anthracnose and powdery mildew in black gram leaf which is essential for the application of proper disease management strategies and reduction of yield loss and aids in promotion of smart agriculture.

A review on Cotton Leaf Disease Detection with Balanced and Imbalanced Data

Cotton is a major crop in India, and its production is essential to the country's economy. However, cotton plants are susceptible to a number of diseases, which can cause significant yield losses. Early detection of these diseases is crucial, but manual identification can be challenging. This paper reviews the use of machine learning for cotton leaf disease detection. A number of classification methods and algorithms are discussed, including convolutional neural networks (CNNs), random forests, and SMOTE. The results show that machine learning can be used to achieve high accuracy in detecting cotton leaf diseases, even in the case of imbalanced datasets. The paper also discusses the importance of balancing class distributions in order to improve model performance. A number of data-level and classifier-level techniques are evaluated, and the results show that the Weighted Random Forest algorithm is effective in handling imbalanced datasets. The findings of this paper suggest that machine learning is a promising tool for the early detection of cotton leaf diseases. This could help to reduce yield losses and improve the profitability of cotton farming in India.

Comparison of plant growth and yield of rice under double-row and tile transplanting systems

To increase rice production for meeting community needs and reduce yield loss under the climate change, the Indonesian government has been recommending double-row transplanting systems, which are Jajar Legowo (JL) and Jejer Manten (JM). Some researchers found that those systems could increase the yield of the Indica rice cultivar without scientific investigations, such as differences in micrometeorological conditions under the different planting spaces. This study aimed to compare plant growth, yield, and solar radiation within the canopy in JL and JM with those in a standard tile transplanting system (TL) by planting Japonica rice cv. Nikomaru, from June 18 to October 7 2022. The result showed that the yield was higher in JL and JM than in TL, by 18.2% and 20.4%, respectively. In addition, the number of filled grain was higher in JL and JM than in TL, by 19.0% and 21.0%, respectively. JM and JL had higher leaf area index than TL during the generative stage causing higher photosynthesis in the unit area. Comparing JL and JM, JL had a smaller leaf area index resulting in lower absorption of solar radiation by the leaves. We suggested that JM is the best transplanting system for producing a higher yield.

Major viral diseases in grain legumes: designing disease resistant legumes from plant breeding and OMICS integration.

Grain legumes play a crucial role in human nutrition and as a staple crop for low-income farmers in developing and underdeveloped nations, contributing to overall food security and agroecosystem services. Viral diseases are major biotic stresses that severely challenge global grain legume production. In this review, we discuss how exploring naturally resistant grain legume genotypes within germplasm, landraces, and crop wild relatives could be used as promising, economically viable, and eco-environmentally friendly solution to reduce yield losses. Studies based on Mendelian and classical genetics have enhanced our understanding of key genetic determinants that govern resistance to various viral diseases in grain legumes. Recent advances in molecular marker technology and genomic resources have enabled us to identify genomic regions controlling viral disease resistance in various grain legumes using techniques such as QTL mapping, genome-wide association studies, whole-genome resequencing, pangenome and 'omics' approaches. These comprehensive genomic resources have expedited the adoption of genomics-assisted breeding for developing virus-resistant grain legumes. Concurrently, progress in functional genomics, especially transcriptomics, has helped unravel underlying candidate gene(s) and their roles in viral disease resistance in legumes. This review also examines the progress in genetic engineering-based strategies, including RNA interference, and the potential of synthetic biology techniques, such as synthetic promoters and synthetic transcription factors, for creating viral-resistant grain legumes. It also elaborates on the prospects and limitations of cutting-edge breeding technologies and emerging biotechnological tools (e.g., genomic selection, rapid generation advances, and CRISPR/Cas9-based genome editing tool) in developing virus-disease-resistant grain legumes to ensure global food security.

Resistance screening and in silico characterization of cloned novel RGA from multi-race resistant lentil germplasm against Fusarium wilt (Fusarium oxysporum f. sp. lentis).

Fusarium wilt caused by Fusarium oxysporum f. sp. lentis (Fol) is the most devastating disease of lentil present worldwide. Identification of multi-race fusarium wilt resistance genes and their incorporation into existing cultivars will help to reduce yield losses. In the present study, 100 lentil germplasms belonging to seven lentil species were screened against seven prevalent races of Fol, and accessions IC201561 (Lens culinaris subsp. culinaris), EC714243 (L. c. subsp. odemensis), and EC718238 (L. nigricans) were identified as resistant. The typical R gene codes for the nucleotide-binding site and leucine-rich repeats (NBS-LRR) at the C terminal are linked to either the Toll/interleukin 1-like receptor (TIR) or coiled coil (CC) at the N terminal. In the present study, degenerate primers, designed from the NBS region amplifying the P-loop to the GLPLA motif, isolated forty-five resistance gene analogues (RGAs) from identified resistant accessions. The sequence alignment identified both classes of RGAs, TIR and non-TIR, based on the presence of aspartate (D) and tryptophan (W) at the end of the kinase motif, respectively. The phylogenetic analysis grouped the RGAs into six classes, from LRGA1 to LRGA6, which determined the diversity of the RGAs present in the host. Grouping of the RGAs identified from Lens nigricans, LnRGA 2, 9, 13 with I2 revealed the structural similarity with the fusarium resistance gene. The similarity index ranged from 27.85% to 86.98% among the RGAs and from 26.83% to 49.41% among the known R genes, I2, Gpa2, M, and L6. The active binding sites present along the conserved motifs grouped the RGAs into 13 groups. ADP/ATP, being the potential ligand, determines the ATP binding and ATP hydrolysis activity of the RGAs. The isolated RGAs can be used to develop markers linked to the functional R gene. Furthermore, expression analysis and full-length gene isolation pave the path to identifying the molecular mechanism involved in resistance.

Vegetal-derived biostimulants distinctively command the physiological and metabolomic signatures of lettuce grown in depleted nitrogen conditions

Biostimulants are sustainable inputs that can be used to reduce chemical fertilizers dependency while reinforcing nutrient uptake, yield and quality of crops, and modulate plant metabolic processes. Protein hydrolysates (PHs) are prominent biostimulants that guarantee a reduction in yield loss under sub-optimal nitrogen (N) conditions. On these bases, a new Malvaceae-derived PH product was tested along a commercial legume-derived PH on soilless greenhouse-grown lettuce, to comparatively assess their activity under N depletion conditions (1 mM NO3−). Both PHs increased biometric parameters under optimal but to a lesser extent under depleted N conditions. Legume-derived PH promoted greater Fv/Fm, lutein and β-carotene under optimal N conditions and higher catalase and total phenolic acids. In contrast, Malvaceae-derived PH did not affect phenolic acids but increased leaf concentration of Ca, Mg and catalase while reducing H2O2. Biochemical changes were then evaluated through untargeted metabolomics. Metabolomics showed a hierarchically prevalent effect of the N level, with the PHs showing distinctive reprogramming under optimal and depleted N conditions. Among others, phenylpropanoids were mainly down-accumulated in stressed plants, while (polyunsaturated fatty acids) PUFA accumulated following the application of PHs. Notwithstanding, the severe depletion of N cannot be compensated by PHs treatment since biostimulants are used to complement fertilizers use and not to replace it.

LDRGDb - Legumes disease resistance genes database.

Legumes comprise one of the world's largest, most diverse, and economically important plant families, known for their nutritional and medicinal benefits. Legumes are susceptible to a wide range of diseases, similar to other agricultural crops. Diseases have a considerable impact on the production of legume crop species, resulting in large yield losses worldwide. Due to continuous interactions between plants and their pathogens in the environment and the evolution of new pathogens under high selection pressure; disease resistant genes emerge in plant cultivars in the field against those pathogens or disease. Thus, disease resistant genes play critical roles in plant resistance responses, and their discovery and subsequent use in breeding programmes aid in reducing yield loss. The genomic era, with its high-throughput and low-cost genomic tools, has revolutionised our understanding of the complex interactions between legumes and pathogens, resulting in the identification of several critical participants in both the resistant and susceptible relationships. However, a substantial amount of existing information about numerous legume species has been disseminated as text or is preserved across fractions in different databases, posing a challenge for researchers. As a result, the range, scope, and complexity of these resources pose challenges to those who manage and use them. Therefore, there is an urgent need to develop tools and a single conjugate database to manage genetic information for the world's plant genetic resources, allowing for the rapid incorporation of essential resistance genes into breeding strategies. Here, developed the first comprehensive database of disease resistance genes named as LDRGDb - LEGUMES DISEASE RESISTANCE GENES DATABASE comprises 10 legumes [Pigeon pea (Cajanus cajan), Chickpea (Cicer arietinum), Soybean (Glycine max), Lentil (Lens culinaris), Alfalfa (Medicago sativa), Barrelclover (Medicago truncatula), Common bean (Phaseolus vulgaris), Pea (Pisum sativum),Faba bean (Vicia faba), and Cowpea (Vigna unguiculata)]. The LDRGDb is a user-friendly database developed by integrating a variety of tools and software that combine knowledge about resistant genes, QTLs, and their loci, with proteomics, pathway interactions, and genomics (https://ldrgdb.in/).

Effect of chemical regulators on the recovery of leaf physiology, dry matter accumulation and translocation, and yield-related characteristics in winter wheat following dry-hot wind

Dry-hot wind stress causes losses in wheat productivity in major growing regions worldwide, especially winter wheat in the Huang-Huai-Hai Plain of China, and both the occurrence and severity of such events are likely to increase with global climate change. To investigate the recovery of physiological functions and yield formation using a new non-commercial chemical regulator (NCR) following dry-hot wind stress, we conducted a three-year field experiment (2018–2021) with sprayed treatments of tap water (control), monopotassium phosphate (CKP), NCR at both the jointing and flowering stages (CFS), and NCR only at the jointing stage (FSJ) or flowering stage (FSF). The leaf physiology, biomass accumulation and translocation, grain-filling process, and yield components in winter wheat were assessed. Among the single spraying treatments, the FSJ treatment was beneficial for the accumulation of dry matter before anthesis, as well as larger increases in the maximum grain-filling rate and mean grain-filling rate. The FSF treatment performed better in maintaining a high relative chlorophyll content as indicated by the SPAD value, and a low rate of excised leaf water loss in flag leaves, promoting dry matter accumulation and the contribution to grain after anthesis, prolonging the duration of grain filling, and causing the period until the maximum grain-filling rate reached earlier. The CFS treatment was better than any other treatments in relieving the effects of dry-hot wind. The exogenous NCR treatments significantly increased grain yields by 12.45–18.20% in 2018–2019, 8.89–13.82% in 2019–2020, and 8.10–9.00% in 2020–2021. The conventional measure of the CKP treatment only increased grain yield by 6.69% in 2020–2021. The CFS treatment had the greatest mitigating effect on yield loss under dry-hot wind stress, followed by the FSF and FSJ treatments, and the CKP treatment only had a minimal effect. In summary, the CFS treatment could be used as the main chemical control measure for wheat stress resistance and yield stability in areas with a high incidence of dry-hot wind. This treatment can effectively regulate green retention and the water status of leaves, promote dry matter accumulation and efficient translocation, improve the grain-filling process, and ultimately reduce yield losses.

Faster region-based convolutional neural network for plant-parasitic and non-parasitic nematode detection

Nematodes represent very abundant and the largest species diversity in the world. Nematodes, which live in a soil environment, possess several functions in agricultural systems. There are two huge groups of soil nematodes, a non-parasitic nematode, which contributes positively to ecological processes, and a plant-parasitic nematode, which cause various disease and reduces yield losses in the agricultural system. Early detection and classification in the agricultural area infected with plant-parasitic nematode and interpreting the soil level condition in this area required a fast and reliable detection system. However, nematode identification is challenging and time-consuming due to their similar morphology. This study applied a pre-trained faster region-based convolutional neural network (RCNN) for plant-parasitic and non-parasitic nematodes detection. These deep learning-based object detection models gave satisfactory results as the accuracy reached 87.5%.

Application of image processing and transfer learning for the detection of rust disease

Plant diseases introduce significant yield and quality losses to the food production industry, worldwide. Early identification of an epidemic could lead to more effective management of the disease and potentially reduce yield loss and limit excessive input costs. Image processing and deep learning techniques have shown promising results in distinguishing healthy and infected plants at early stages. In this paper, the potential of four convolutional neural network models, including Xception, Residual Networks (ResNet)50, EfficientNetB4, and MobileNet, in the detection of rust disease on three commercially important field crops was evaluated. A dataset of 857 positive and 907 negative samples captured in the field and greenhouse environments were used. Training and testing of the algorithms were conducted using 70% and 30% of the data, respectively where the performance of different optimizers and learning rates were tested. Results indicated that EfficientNetB4 model was the most accurate model (average accuracy = 94.29%) in the disease detection followed by ResNet50 (average accuracy = 93.52%). Adaptive moment estimation (Adam) optimizer and learning rate of 0.001 outperformed all other corresponding hyperparameters. The findings from this study provide insights into the development of tools and gadgets useful in the automated detection of rust disease required for precision spraying.

Crop rotation with Meloidogyne-resistant germplasm is useful to manage and revert the (a)virulent populations of Mi1.2 gene and reduce yield losses.

A rotation sequence of ungrafted and grafted tomato-melon-pepper-watermelon on resistant rootstocks 'Brigeor', Cucumis metuliferus, 'Oscos' and Citrullus amarus, respectively, was carried out in a plastic greenhouse, ending with a susceptible or resistant tomato crop. The rotation was conducted in plots infested by an avirulent (Avi) or a partially virulent (Vi) Meloidogyne incognita population to the Mi1.2 gene. At the beginning of the study, the reproduction index (RI, relative reproduction in the resistant respect susceptible tomato) of Avi and Vi populations was 1.3% and 21.6%, respectively. Soil nematode density at transplanting (Pi) and at the end (Pf) of each crop, disease severity and crop yield were determined. Moreover, the putative virulence selection and fitness cost were determined at the end of each crop in pot tests. In addition, a histopathological study was carried out 15 days after nematode inoculation in pot test. The volume and number of nuclei per giant cell (GC) and the number of GC, their volume and the number of nuclei per feeding site in susceptible watermelon and pepper were compared with C. amarus and resistant pepper. At the beginning of the study, the Pi of Avi and Vi plots did not differ between susceptible and resistant germplasm. At the end of the rotation, the Pf of Avi was 1.2 the Pi in susceptible and 0.06 in resistant, the cumulative yield of grafted crops was 1.82 times higher than that of the ungrafted susceptible ones, and the RI in resistant tomato less than 10% irrespective of the rotation sequence. Concerning the Vi, Pf was below the detection level at the end of the rotation in resistant and 3 times Pi in the susceptible. The cumulative yield of grafted crops was 2.83 times higher than that of the ungrafted and the RI in resistant tomato was 7.6%, losing the population's virulence. In the histopathological study, no differences in number of GC per feeding site were observed in watermelon compared to C. amarus, but they were more voluminous and contained higher number of nuclei per GC and per feeding site. Regarding pepper, Avi population did not penetrate resistant rootstock.

Deep transfer learning model for disease identification in wheat crop

Wheat rusts, caused by pathogenic fungi, are responsible for significant losses in Wheat production. Leaf rust can cause around 45–50% crop loss, whereas stem and stripe rust can cause up to 100% crop loss under suitable weather conditions. Early treatment is crucial in reducing yield loss and improving the effectiveness of phytosanitary measures. In this study, an EfficientNet architecture-based model for Wheat disease identification is proposed for automatically detecting major Wheat rusts. We prepared a dataset, referred to as WheatRust21, consisting of 6556 images of healthy and diseased leaves from natural field conditions. We attempted several classical CNN-based models such as VGG19, ResNet152, DenseNet169, InceptionNetV3, and MobileNetV2 for Wheat rust disease identification and obtained accuracy ranging from 91.2 to 97.8%. To further improve accuracy, we experimented with eight variants of EfficientNet architecture and discovered that our fine-tuned EfficientNet B4 model achieved a testing accuracy of 99.35%, a result that has not been reported in the literature so far to the best of our knowledge. This model can be easily integrated into mobile applications for use by stakeholders for image-based wheat disease identification in field conditions.

Ridge intertillage alters rhizosphere bacterial communities and plant physiology to reduce yield loss of waterlogged cotton

Ridge intertillage alters rhizosphere bacterial communities and plant physiology to reduce yield loss of waterlogged cotton

EFEKTIVITAS PESTISIDA NABATI EKOENZIM DALAM MENGENDALIKAN SERANGAN KUTU DAUN (Aphis gossypii GLOVER) PADA TANAMAN CABAI (Capsicum annuum L.)

Chili plants are one of the important horticultural commodities that have many benefits. Aphids are one of the plant pests as well as a viral vector on chili plants. One control that can be applied to control it is by utilizing eco enzymes. The study used a randomized group design with seven treatments and three repeats. The treatment used is without the application of eco enzyme vegetable pesticides (A0), tobacco eco enzyme application (A1), onion eco enzyme (A2), tembelekan leaf eco enzyme (A3), red ginger eco enzyme (A4), noni fruit ecoenzyme (A5) and synthetic insecticides (A6). The observation variables consist of the intensity of the attack and the number of fruits. Observational data were analyzed using fingerprint analysis. If the treatment has a real effect, it is followed by the Smallest Real Difference test at a confidence level of 95%. Tobacco-based eco enzyme is effective in controlling aphid infestation (A. gossypii Glover) and reducing yield losses in large chili peppers (C. annuum L.).

Genomic Insights into the Origin of a Thermotolerant Tomato Line and Identification of Candidate Genes for Heat Stress.

Climate change represents the main problem for agricultural crops, and the constitution of heat-tolerant genotypes is an important breeder's strategy to reduce yield losses. The aim of the present study was to investigate the whole genome of a heat-tolerant tomato genotype (E42), in order to identify candidate genes involved in its response to high temperature. E42 presented a high variability for chromosomes 1, 4, 7 and 12, and phylogenetic analysis highlighted its relationship with the wild S. pimpinellifolium species. Variants with high (18) and moderate (139) impact on protein function were retrieved from two lists of genes related to heat tolerance and reproduction. This analysis permitted us to prioritize a subset of 35 candidate gene mapping in polymorphic regions, some colocalizing in QTLs controlling flowering in tomato. Among these genes, we identified 23 HSPs, one HSF, six involved in flowering and five in pollen activity. Interestingly, one gene coded for a flowering locus T1 and mapping on chromosome 11 resides in a QTL region controlling flowering and also showed 100% identity with an S. pimpinellifolium allele. This study provides useful information on both the E42 genetic background and heat stress response, and further studies will be conducted to validate these genes.

BotanicX-AI: Identification of Tomato Leaf Diseases Using an Explanation-Driven Deep-Learning Model

Early and accurate tomato disease detection using easily available leaf photos is essential for farmers and stakeholders as it help reduce yield loss due to possible disease epidemics. This paper aims to visually identify nine different infectious diseases (bacterial spot, early blight, Septoria leaf spot, late blight, leaf mold, two-spotted spider mite, mosaic virus, target spot, and yellow leaf curl virus) in tomato leaves in addition to healthy leaves. We implemented EfficientNetB5 with a tomato leaf disease (TLD) dataset without any segmentation, and the model achieved an average training accuracy of 99.84% ± 0.10%, average validation accuracy of 98.28% ± 0.20%, and average test accuracy of 99.07% ± 0.38% over 10 cross folds.The use of gradient-weighted class activation mapping (GradCAM) and local interpretable model-agnostic explanations are proposed to provide model interpretability, which is essential to predictive performance, helpful in building trust, and required for integration into agricultural practice.

In Silico Identification of Cassava Genome-Encoded MicroRNAs with Predicted Potential for Targeting the ICMV-Kerala Begomoviral Pathogen of Cassava.

Cassava mosaic disease (CMD) is caused by several divergent species belonging to the genus Begomovirus (Geminiviridae) transmitted by the whitefly Bemisia tabaci cryptic species group. In India and other parts of Asia, the Indian cassava mosaic virus-Kerala (ICMV-Ker) is an emergent begomovirus of cassava causing damage that results in reduced yield loss and tuber quality. Double-stranded RNA-mediated interference (RNAi) is an evolutionary conserved mechanism in eukaryotes and highly effective, innate defense system to inhibit plant viral replication and/or translation. The objective of this study was to identify and characterize cassava genome-encoded microRNAs (mes-miRNA) that are predicted to target ICMV-Ker ssDNA-encoded mRNAs, based on four in silico algorithms: miRanda, RNA22, Tapirhybrid, and psRNA. The goal is to deploy the predicted miRNAs to trigger RNAi and develop cassava plants with resistance to ICMV-Ker. Experimentally validated mature cassava miRNA sequences (n = 175) were downloaded from the miRBase biological database and aligned with the ICMV-Ker genome. The miRNAs were evaluated for base-pairing with the cassava miRNA seed regions and to complementary binding sites within target viral mRNAs. Among the 175 locus-derived mes-miRNAs evaluated, one cassava miRNA homolog, mes-miR1446a, was identified to have a predicted miRNA target binding site, at position 2053 of the ICMV-Ker genome. To predict whether the cassava miRNA might bind predicted ICMV-Ker mRNA target(s) that could disrupt viral infection of cassava plants, a cassava locus-derived miRNA-mRNA regulatory network was constructed using Circos software. The in silico-predicted cassava locus-derived mes-miRNA-mRNA network corroborated interactions between cassava mature miRNAs and the ICMV-Ker genome that warrant in vivo analysis, which could lead to the development of ICMV-Ker resistant cassava plants.

Overexpression of an Antioxidant Enzyme APX1 in cpr5 Mutant Restores its Pleiotropic Growth Phenotype.

Breeding crops with enhanced immunity is an effective strategy to reduce yield loss caused by pathogens. The constitutive expresser of pathogenesis-related genes (cpr5) mutant shows enhanced pathogen resistance but retarded growth; thus, it restricts the application of cpr5 in breeding crops with disease resistance. Reactive oxygen species (ROS) play important roles in plant growth and defense. In this study, we determined that the cpr5 mutant exhibited excessive ROS accumulation. However, the mutation of respiratory burst oxidase homolog D (RBOHD), a reduced nicotinamide adenine dinucleotide phosphate (NADPH) oxidase responsible for the production of ROS signaling in plant immunity, did not suppress excessive ROS levels in cpr5. Furthermore, the cpr5 mutant showed low levels of ascorbate peroxidase 1 (APX1), an important cytosolic ROS-scavenging enzyme. APX1 overexpression in the cpr5 background removed excessive ROS and restored the pleiotropic growth phenotype. Notably, APX1 overexpression did not reduce the resistance of cpr5 mutant to virulent strain Pseudomonas syringae pv. tomato (Pst) DC3000 and avirulent strain Pst DC3000 (avrRpt2). These results suggest that the removal of excessive ROS by APX1 overexpression restored the cpr5 growth phenotype while conserving pathogen resistance. Hence, our study provides a theoretical and empirical basis for utilizing CPR5 in the breeding of crops with disease resistance by effective oxidative stress management via APX1 expression.

Fine-grained Potato Disease Identification Based on Contrastive Convolutional Neural Networks

ABSTRACT Early identification of potato diseases is of great significance for reducing yield losses. The identification of different types of diseases has achieved great success. However, for different periods of different disease, it is difficult to distinguish due to similar symptoms and fine-grained, so there are few related studies. In this study, we proposed a convolutional neural network based on contrastive learning to identify fine-grained potato diseases. Different from the previous unsupervised contrastive learning used in pre-training, the proposed model adds a projection head to the backbone network of Vgg16 to extract the contrastive representation features, and then integrates the contrastive loss with the classification loss to form a joint loss. Finally, an end-to-end supervised contrastive convolutional neural network is constructed, which is easier to train while reducing the transmission error. Experimental results show that the proposed model achieves an average recognition accuracy of 97.24%, which is higher than 90.28% of Resnet50, 90.62% of Resnet101, 93.06% of AlexNet, 94.44% of Inception V3, and 94.79% of Vgg16. It shows that the model has an obvious effect on classification task with similar features, and has practical significance for fine-grained potato disease identification.

Detection Method for Tomato Leaf Mildew Based on Hyperspectral Fusion Terahertz Technology.

Leaf mildew is a common disease of tomato leaves. Its detection is an important means to reduce yield loss from the disease and improve tomato quality. In this study, a new method was developed for the multi-source detection of tomato leaf mildew by THz hyperspectral imaging through combining internal and external leaf features. First, multi-source information obtained from tomato leaves of different disease grades was extracted by near-infrared hyperspectral imaging and THz time-domain spectroscopy, while the influence of low-frequency noise was removed by the Savitzky Golay (SG) smoothing algorithm. A genetic algorithm (GA) was used to optimize the selection of the characteristic near-infrared hyperspectral band. Principal component analysis (PCA) was employed to optimize the THz characteristic absorption spectra and power spectrum dimensions. Recognition models were developed for different grades of tomato leaf mildew infestation by incorporating near-infrared hyperspectral imaging, THz absorbance, and power spectra using the backpropagation neural network (BPNN), and the models had recognition rates of 95%, 96.67%, and 95%, respectively. Based on the near-infrared hyperspectral features, THz time-domain spectrum features, and classification model, the probability density of the posterior distribution of tomato leaf health parameter variables was recalculated by a Bayesian network model. Finally, a fusion diagnosis and health evaluation model of tomato leaf mildew with hyperspectral fusion THz was established, and the recognition rate of tomato leaf mildew samples reached 97.12%, which improved the recognition accuracy by 0.45% when compared with the single detection method, thereby achieving the accurate detection of facility diseases.