Digital image processing techniques for detecting, quantifying and classifying plant diseases

Residual swin transformer for classifying the types of cotton pests in complex background

Background

Cotton pests have a major impact on cotton quality and yield during cotton production and cultivation. With the rapid development of agricultural intelligence, the accurate classification of cotton pests is a key factor in realizing the precise application of medicines by utilize unmanned aerial vehicles (UAVs), large application devices and other equipment.

Methods

In this study, a cotton insect pest classification model based on improved Swin Transformer is proposed. The model introduces the residual module, skip connection, into Swin Transformer to improve the problem that pest features are easily confused in complex backgrounds leading to poor classification accuracy, and to enhance the recognition of cotton pests. In this study, 2705 leaf images of cotton insect pests (including three insect pests, cotton aphids, cotton mirids and cotton leaf mites) were collected in the field, and after image preprocessing and data augmentation operations, model training was performed.

Results

The test results proved that the accuracy of the improved model compared to the original model increased from 94.6% to 97.4%, and the prediction time for a single image was 0.00434s. The improved Swin Transformer model was compared with seven kinds of classification models (VGG11, VGG11-bn, Resnet18, MobilenetV2, VIT, Swin Transformer small, and Swin Transformer base), and the model accuracy was increased respectively by 0.5%, 4.7%, 2.2%, 2.5%, 6.3%, 7.9%, 8.0%.

Discussion

Therefore, this study demonstrates that the improved Swin Transformer model significantly improves the accuracy and efficiency of cotton pest detection compared with other classification models, and can be deployed on edge devices such as utilize unmanned aerial vehicles (UAVs), thus providing an important technological support and theoretical basis for cotton pest control and precision drug application.

Image processing and impact analyses of terminal heat stress on yield of lentil

Abiotic factors, including heat stress, significantly impact the growth and development of lentil across the globe. Although these stresses impact the plant's phenotypic, genotypic, metabolic, and yield development, predicting those traits in lentil is challenging. This study aimed to construct a machine learning-based yield prediction model for lentil using various yield attributes under two different sowing conditions. Twelve genotypes were planted in open-field conditions, and images were captured 45 days after sowing (DAS) and 60 DAS to make predictions for agro-morphological traits with the assessment for the influence of high-temperature stress on lentil growth. Greening techniques like Excess Green, Modified Excess Green (ME × G), and Color Index of Plant Extraction (CIVE) were used to extract 35 vegetative indices from the crop image. Random forest (RF) regression and artificial neural network (ANN) models were developed for both the normal-sown and late-sown lentils. The ME × G-CIVE method with Otsu's thresholding provided superior performance in image segmentation, while the RF model showed the highest level of model generalization. This study demonstrated that yield per plant and number of pods per plant were the most significant attributes for early prediction of lentil production in both conditions using the RF models. After harvesting, various yield parameters of the selected genotypes were measured, showing significant reductions in most traits for the late-sown plants. Heat-tolerant genotypes like RLG-05, Kota Masoor-1, and Kota Masoor-2 depicted decreased yield and harvest index (HI) reduction than the heat-sensitive HUL-57. These findings warrant further study to correlate the data with more stress-modulating attributes.

Supplementary Information

The online version contains supplementary material available at 10.1007/s13205-024-04031-5.

Mitigating Global Challenges: Harnessing Green Synthesized Nanomaterials for Sustainable Crop Production Systems

Green nanotechnology, an emerging field, offers economic and social benefits while minimizing environmental impact. Nanoparticles, pivotal in medicine, pharmaceuticals, and agriculture, are now sourced from green plants and microorganisms, overcoming limitations of chemically synthesized ones. In agriculture, these green-made nanoparticles find use in fertilizers, insecticides, pesticides, and fungicides. Nanofertilizers curtail mineral losses, bolster yields, and foster agricultural progress. Their biological production, preferred for environmental friendliness and high purity, is cost-effective and efficient. Biosensors aid early disease detection, ensuring food security and sustainable farming by reducing excessive pesticide use. This eco-friendly approach harnesses natural phytochemicals to boost crop productivity. This review highlights recent strides in green nanotechnology, showcasing how green-synthesized nanomaterials elevate crop quality, combat plant pathogens, and manage diseases and stress. These advancements pave the way for sustainable crop production systems in the future.

Research on machine vision and deep learning based recognition of cotton seedling aphid infestation level

Aphis gossypii Glover is a major insect pest in cotton production, which can cause yield reduction in severe cases. In this paper, we proposed the A. gossypii infestation monitoring method, which identifies the infestation level of A. gossypii at the cotton seedling stage, and can improve the efficiency of early warning and forecasting of A. gossypii, and achieve precise prevention and cure according to the predicted infestation level. We used smartphones to collect A. gossypii infestation images and compiled an infestation image data set. And then constructed, trained, and tested three different A. gossypii infestation recognition models based on Faster Region-based Convolutional Neural Network (R-CNN), You Only Look Once (YOLO)v5 and single-shot detector (SSD) models. The results showed that the YOLOv5 model had the highest mean average precision (mAP) value (95.7%) and frames per second (FPS) value (61.73) for the same conditions. In studying the influence of different image resolutions on the performance of the YOLOv5 model, we found that YOLOv5s performed better than YOLOv5x in terms of overall performance, with the best performance at an image resolution of 640×640 (mAP of 96.8%, FPS of 71.43). And the comparison with the latest YOLOv8s showed that the YOLOv5s performed better than the YOLOv8s. Finally, the trained model was deployed to the Android mobile, and the results showed that mobile-side detection was the best when the image resolution was 256×256, with an accuracy of 81.0% and FPS of 6.98. The real-time recognition system established in this study can provide technical support for infestation forecasting and precise prevention of A. gossypii.

Advancements in Imaging Sensors and AI for Plant Stress Detection: A Systematic Literature Review

Integrating imaging sensors and artificial intelligence (AI) have contributed to detecting plant stress symptoms, yet data analysis remains a key challenge. Data challenges include standardized data collection, analysis protocols, selection of imaging sensors and AI algorithms, and finally, data sharing. Here, we present a systematic literature review (SLR) scrutinizing plant imaging and AI for identifying stress responses. We performed a scoping review using specific keywords, namely abiotic and biotic stress, machine learning, plant imaging and deep learning. Next, we used programmable bots to retrieve relevant papers published since 2006. In total, 2,704 papers from 4 databases (Springer, ScienceDirect, PubMed, and Web of Science) were found, accomplished by using a second layer of keywords (e.g., hyperspectral imaging and supervised learning). To bypass the limitations of search engines, we selected OneSearch to unify keywords. We carefully reviewed 262 studies, summarizing key trends in AI algorithms and imaging sensors. We demonstrated that the increased availability of open-source imaging repositories such as PlantVillage or Kaggle has strongly contributed to a widespread shift to deep learning, requiring large datasets to train in stress symptom interpretation. Our review presents current trends in AI-applied algorithms to develop effective methods for plant stress detection using image-based phenotyping. For example, regression algorithms have seen substantial use since 2021. Ultimately, we offer an overview of the course ahead for AI and imaging technologies to predict stress responses. Altogether, this SLR highlights the potential of AI imaging in both biotic and abiotic stress detection to overcome challenges in plant data analysis.

Apple Disease Recognition Based on Convolutional Neural Networks With Modified Softmax

Accurate and rapid identification of apple diseases is the basis for preventing and treating the apple diseases, and is very significant for assessing disease disaster. Apple disease recognition from its diseased leaf images is one of the interesting research areas in computer and agriculture field. An apple disease recognition method is proposed based on modified convolutional neural networks (MCNN). In MCNN, Inception is introduced into MCNN, global average pooling (GAP) operator is employed instead of several fully connected layers to speedup training model, and modified Softmax classifier is used in the output layer to improve the recognition performance. The modified Softmax classifier uses the modified linear element as the activation function in the hidden layer and adds the local response normalization in MCNN to avoid the gradient disappearance problem effectively. A series of experiments are conducted on two kinds of apple disease image datasets. The results show the feasibility of the algorithm.

Strawberry Fungal Leaf Scorch Disease Identification in Real-Time Strawberry Field Using Deep Learning Architectures

Plant health is the basis of agricultural development. Plant diseases are a major factor for crop losses in agriculture. Plant diseases are difficult to diagnose correctly, and the manual disease diagnosis process is time consuming. For this reason, it is highly desirable to automatically identify the diseases in strawberry plants to prevent loss of crop quality. Deep learning (DL) has recently gained popularity in image classification and identification due to its high accuracy and fast learning. In this research, deep learning models were used to identify the leaf scorch disease in strawberry plants. Four convolutional neural networks (SqueezeNet, EfficientNet-B3, VGG-16 and AlexNet) CNN models were trained and tested for the classification of healthy and leaf scorch disease infected plants. The performance accuracy of EfficientNet-B3 and VGG-16 was higher for the initial and severe stage of leaf scorch disease identification as compared to AlexNet and SqueezeNet. It was also observed that the severe disease (leaf scorch) stage was correctly classified more often than the initial stage of the disease. All the trained CNN models were integrated with a machine vision system for real-time image acquisition under two different lighting situations (natural and controlled) and identification of leaf scorch disease in strawberry plants. The field experiment results with controlled lightening arrangements, showed that the model EfficientNet-B3 achieved the highest classification accuracy, with 0.80 and 0.86 for initial and severe disease stages, respectively, in real-time. AlexNet achieved slightly lower validation accuracy (0.72, 0.79) in comparison with VGGNet and EfficientNet-B3. Experimental results stated that trained CNN models could be used in conjunction with variable rate agrochemical spraying systems, which will help farmers to reduce agrochemical use, crop input costs and environmental contamination.

Janus-Faced Molecules against Plant Pathogenic Fungi

The high cytotoxicity of the secondary metabolites of mycotoxins is capable of killing microbes and tumour cells alike, similarly to the genotoxic effect characteristic of Janus-faced molecules. The "double-edged sword" effect of several cytotoxins is known, and these agents have, therefore, been utilized only reluctantly against fungal infections. In this review, consideration was given to (a) toxins that could be used against plant and human pathogens, (b) animal models that measure the effect of antifungal agents, (c) known antifungal agents that have been described and efficiently prevent the growth of fungal cells, and (d) the chemical interactions that are characteristic of antifungal agents. The utilization of apoptotic effects against tumour growth by agents that, at the same time, induce mutations may raise ethical issues. Nevertheless, it deserves consideration despite the mutagenic impact of Janus-faced molecules for those patients who suffer from plant pathogenic fungal infections and are older than their fertility age, in the same way that the short-term cytotoxicity of cancer treatment is favoured over the long-term mutagenic effect.

Quantitative dissection of color patterning in the foliar ornamental coleus

Coleus (Coleus scutellarioides) is a popular ornamental plant that exhibits a diverse array of foliar color patterns. New cultivars are currently hand selected by both amateur and experienced plant breeders. In this study, we reimagine breeding for color patterning using a quantitative color analysis framework. Despite impressive advances in high-throughput data collection and processing, complex color patterns remain challenging to extract from image datasets. Using a phenotyping approach called "ColourQuant," we extract and analyze pigmentation patterns from one of the largest coleus breeding populations in the world. Working with this massive dataset, we can analyze quantitative relationships between maternal plants and their progeny, identify features that underlie breeder-selections, and collect and compare public input on trait preferences. This study is one of the most comprehensive explorations into complex color patterning in plant biology and provides insights and tools for exploring the color pallet of the plant kingdom.

Research advance in gas detection of volatile organic compounds released in rice quality deterioration process

Rice quality deterioration will cause grievous waste of stored grain and various food safety problems. Gas detection of volatile organic compounds (VOCs) produced by deterioration is a nondestructive detection method to judge rice quality and alleviate rice spoilage. This review discussed the research advance of VOCs detection in terms of nondestructive detection methods of rice quality deterioration, applications of VOCs in grain detection, inspection of characteristic gas produced during rice spoilage, rice deterioration prevention and control, and detection of VOCs released by rice mildew and insect attack. According to the main causes of rice quality deterioration and major sources of VOCs with off-odor generated during rice storage, deterioration can be divided into mold and insect infection. The results of literature manifested that researches mainly focused on the infection of Aspergillus in the mildew process and the attack of certain pests in recent years, thus the research scope was limited. In this paper, the gas detection methods combined with the chemometrics to qualitatively analyze the VOCs, as well as the correlation with the number of colonies and insects were further studied based on the common dominant strains during rice mildew, that is, Aspergillus and Penicillium fungi, and the common pests during storage, that is, Sitophilus oryzae and Rhyzopertha dominica. Furthermore, this paper pointed out that the quantitative determination of characteristic VOCs, the numeration relationship between VOCs and the degree of mildew and insect infestation, the further expansion of detection range, and the application of degraded rice should be the spotlight of future research.

Classification of Apple Disease Based on Non-Linear Deep Features

Diseases in apple orchards (rot, scab, and blotch) worldwide cause a substantial loss in the agricultural industry. Traditional hand picking methods are subjective to human efforts. Conventional machine learning methods for apple disease classification depend on hand-crafted features that are not robust and are complex. Advanced artificial methods such as Convolutional Neural Networks (CNN’s) have become a promising way for achieving higher accuracy although they need a high volume of samples. This work investigates different Deep CNN (DCNN) applications to apple disease classification using deep generative images to obtain higher accuracy. In order to achieve this, our work progressively modifies a baseline model by using an end-to-end trained DCNN model that has fewer parameters, better recognition accuracy than existing models (i.e., ResNet, SqeezeNet, and MiniVGGNet). We have performed a comparative study with state-of-the-art CNN as well as conventional methods proposed in the literature, and comparative results confirm the superiority of our proposed model.

AI-Driven Framework for Recognition of Guava Plant Diseases through Machine Learning from DSLR Camera Sensor Based High Resolution Imagery

Plant diseases can cause a considerable reduction in the quality and number of agricultural products. Guava, well known to be the tropics' apple, is one significant fruit cultivated in tropical regions. It is attacked by 177 pathogens, including 167 fungal and others such as bacterial, algal, and nematodes. In addition, postharvest diseases may cause crucial production loss. Due to minor variations in various guava disease symptoms, an expert opinion is required for disease analysis. Improper diagnosis may cause economic losses to farmers' improper use of pesticides. Automatic detection of diseases in plants once they emerge on the plants' leaves and fruit is required to maintain high crop fields. In this paper, an artificial intelligence (AI) driven framework is presented to detect and classify the most common guava plant diseases. The proposed framework employs the ΔE color difference image segmentation to segregate the areas infected by the disease. Furthermore, color (RGB, HSV) histogram and textural (LBP) features are applied to extract rich, informative feature vectors. The combination of color and textural features are used to identify and attain similar outcomes compared to individual channels, while disease recognition is performed by employing advanced machine-learning classifiers (Fine KNN, Complex Tree, Boosted Tree, Bagged Tree, Cubic SVM). The proposed framework is evaluated on a high-resolution (18 MP) image dataset of guava leaves and fruit. The best recognition results were obtained by Bagged Tree classifier on a set of RGB, HSV, and LBP features (99% accuracy in recognizing four guava fruit diseases (Canker, Mummification, Dot, and Rust) against healthy fruit). The proposed framework may help the farmers to avoid possible production loss by taking early precautions.

Plant pest surveillance: from satellites to molecules

Plant pests and diseases impact both food security and natural ecosystems, and the impact has been accelerated in recent years due to several confounding factors. The globalisation of trade has moved pests out of natural ranges, creating damaging epidemics in new regions. Climate change has extended the range of pests and the pathogens they vector. Resistance to agrochemicals has made pathogens, pests, and weeds more difficult to control. Early detection is critical to achieve effective control, both from a biosecurity as well as an endemic pest perspective. Molecular diagnostics has revolutionised our ability to identify pests and diseases over the past two decades, but more recent technological innovations are enabling us to achieve better pest surveillance. In this review, we will explore the different technologies that are enabling this advancing capability and discuss the drivers that will shape its future deployment.

Multiscale Parallel Algorithm for Early Detection of Tomato Gray Mold in a Complex Natural Environment

Plant disease detection technology is an important part of the intelligent agricultural Internet of Things monitoring system. The real natural environment requires the plant disease detection system to have extremely high real time detection and accuracy. The lightweight network MobileNetv2-YOLOv3 model can meet the real-time detection, but the accuracy is not enough to meet the actual needs. This study proposed a multiscale parallel algorithm MP-YOLOv3 based on the MobileNetv2-YOLOv3 model. The proposed method put forward a multiscale feature fusion method, and an efficient channel attention mechanism was introduced into the detection layer of the network to achieve feature enhancement. The parallel detection algorithm was used to effectively improve the detection performance of multiscale tomato gray mold lesions while ensuring the real-time performance of the algorithm. The experimental results show that the proposed algorithm can accurately and real-time detect multiscale tomato gray mold lesions in a real natural environment. The F1 score and the average precision reached 95.6 and 93.4% on the self-built tomato gray mold detection dataset. The model size was only 16.9 MB, and the detection time of each image was 0.022 s.

Abdul Majeed A. A real-time approach of diagnosing rice leaf disease using deep learning-based faster R-CNN framework

The rice leaves related diseases often pose threats to the sustainable production of rice affecting many farmers around the world. Early diagnosis and appropriate remedy of the rice leaf infection is crucial in facilitating healthy growth of the rice plants to ensure adequate supply and food security to the rapidly increasing population. Therefore, machine-driven disease diagnosis systems could mitigate the limitations of the conventional methods for leaf disease diagnosis techniques that is often time-consuming, inaccurate, and expensive. Nowadays, computer-assisted rice leaf disease diagnosis systems are becoming very popular. However, several limitations ranging from strong image backgrounds, vague symptoms' edge, dissimilarity in the image capturing weather, lack of real field rice leaf image data, variation in symptoms from the same infection, multiple infections producing similar symptoms, and lack of efficient real-time system mar the efficacy of the system and its usage. To mitigate the aforesaid problems, a faster region-based convolutional neural network (Faster R-CNN) was employed for the real-time detection of rice leaf diseases in the present research. The Faster R-CNN algorithm introduces advanced RPN architecture that addresses the object location very precisely to generate candidate regions. The robustness of the Faster R-CNN model is enhanced by training the model with publicly available online and own real-field rice leaf datasets. The proposed deep-learning-based approach was observed to be effective in the automatic diagnosis of three discriminative rice leaf diseases including rice blast, brown spot, and hispa with an accuracy of 98.09%, 98.85%, and 99.17% respectively. Moreover, the model was able to identify a healthy rice leaf with an accuracy of 99.25%. The results obtained herein demonstrated that the Faster R-CNN model offers a high-performing rice leaf infection identification system that could diagnose the most common rice diseases more precisely in real-time.

Rapid Collection and Aptamer-Based Sensitive Electrochemical Detection of Soybean Rust Fungi Airborne Urediniospores

Plants are the central source of food for humans around the world. Unfortunately, plants can be negatively affected by diverse kinds of diseases that are responsible for major economic losses worldwide. Thus, monitoring plant health and early detection of pathogens are essential to reduce disease spread and facilitate effective management practices. Various detection approaches are currently practiced. These methods mainly include visual inspection and laboratory tests. Nonetheless, these methods are labor-intensive, time-consuming, expensive, and inefficient in the early stages of infection. Thus, it is extremely important to detect diseases at the early stages of the epidemic. Here, we would like to present a fast, sensitive, and reliable electrochemical sensing platform for the detection of airborne soybean rust spores. The suspected airborne soybean rust spores are first collected and trapped inside a carbon 3D electrode matrix by high-capacity air-sampling means. Then, a specific biotinylated aptamer, suitable to target specific sites of soybean rust spores is applied. This aptamer agent binds to the surface of the collected spores on the electrode. Finally, spore-bound aptamer units are incubated with a streptavidin-alkaline phosphatase agent leading to the enzymatic formation of p-nitrophenol, which is characterized by its unique electrochemical properties. Our method allows for the rapid (ca. 2 min), selective, and sensitive collection and detection of soybean rust spores (down to the limit of 100-200 collected spores per cm2 of electrode area). This method could be further optimized for its sensitivity and applied to the future multiplex early detection of various airborne plant diseases.

Biotic Yield Losses in the Southern Amazon, Brazil: Making Use of Smartphone-Assisted Plant Disease Diagnosis Data

Pathogens and animal pests (P&A) are a major threat to global food security as they directly affect the quantity and quality of food. The Southern Amazon, Brazil's largest domestic region for soybean, maize and cotton production, is particularly vulnerable to the outbreak of P&A due to its (sub)tropical climate and intensive farming systems. However, little is known about the spatial distribution of P&A and the related yield losses. Machine learning approaches for the automated recognition of plant diseases can help to overcome this research gap. The main objectives of this study are to (1) evaluate the performance of Convolutional Neural Networks (ConvNets) in classifying P&A, (2) map the spatial distribution of P&A in the Southern Amazon, and (3) quantify perceived yield and economic losses for the main soybean and maize P&A. The objectives were addressed by making use of data collected with the smartphone application Plantix. The core of the app's functioning is the automated recognition of plant diseases via ConvNets. Data on expected yield losses were gathered through a short survey included in an "expert" version of the application, which was distributed among agronomists. Between 2016 and 2020, Plantix users collected approximately 78,000 georeferenced P&A images in the Southern Amazon. The study results indicate a high performance of the trained ConvNets in classifying 420 different crop-disease combinations. Spatial distribution maps and expert-based yield loss estimates indicate that maize rust, bacterial stalk rot and the fall armyworm are among the most severe maize P&A, whereas soybean is mainly affected by P&A like anthracnose, downy mildew, frogeye leaf spot, stink bugs and brown spot. Perceived soybean and maize yield losses amount to 12 and 16%, respectively, resulting in annual yield losses of approximately 3.75 million tonnes for each crop and economic losses of US$2 billion for both crops together. The high level of accuracy of the trained ConvNets, when paired with widespread use from following a citizen-science approach, results in a data source that will shed new light on yield loss estimates, e.g., for the analysis of yield gaps and the development of measures to minimise them.

An embedded system for the automated generation of labeled plant images to enable machine learning applications in agriculture

A lack of sufficient training data, both in terms of variety and quantity, is often the bottleneck in the development of machine learning (ML) applications in any domain. For agricultural applications, ML-based models designed to perform tasks such as autonomous plant classification will typically be coupled to just one or perhaps a few plant species. As a consequence, each crop-specific task is very likely to require its own specialized training data, and the question of how to serve this need for data now often overshadows the more routine exercise of actually training such models. To tackle this problem, we have developed an embedded robotic system to automatically generate and label large datasets of plant images for ML applications in agriculture. The system can image plants from virtually any angle, thereby ensuring a wide variety of data; and with an imaging rate of up to one image per second, it can produce lableled datasets on the scale of thousands to tens of thousands of images per day. As such, this system offers an important alternative to time- and cost-intensive methods of manual generation and labeling. Furthermore, the use of a uniform background made of blue keying fabric enables additional image processing techniques such as background replacement and image segementation. It also helps in the training process, essentially forcing the model to focus on the plant features and eliminating random correlations. To demonstrate the capabilities of our system, we generated a dataset of over 34,000 labeled images, with which we trained an ML-model to distinguish grasses from non-grasses in test data from a variety of sources. We now plan to generate much larger datasets of Canadian crop plants and weeds that will be made publicly available in the hope of further enabling ML applications in the agriculture sector.

A Review of Advanced Technologies and Development for Hyperspectral-Based Plant Disease Detection in the Past Three Decades

The detection, quantification, diagnosis, and identification of plant diseases is particularly crucial for precision agriculture. Recently, traditional visual assessment technology has not been able to meet the needs of precision agricultural informatization development, and hyperspectral technology, as a typical type of non-invasive technology, has received increasing attention. On the basis of simply describing the types of pathogens and host–pathogen interaction processes, this review expounds the great advantages of hyperspectral technologies in plant disease detection. Then, in the process of describing the hyperspectral disease analysis steps, the articles, algorithms, and methods from disease detection to qualitative and quantitative evaluation are mainly summarizing. Additionally, according to the discussion of the current major problems in plant disease detection with hyperspectral technologies, we propose that different pathogens’ identification, biotic and abiotic stresses discrimination, plant disease early warning, and satellite-based hyperspectral technology are the primary challenges and pave the way for a targeted response.

In Field Detection of Downy Mildew Symptoms with Proximal Colour Imaging

This paper proposes to study the potentialities of on-board colour imaging for the in-field detection of a textbook case disease: the grapevine downy mildew. It introduces an algorithmic strategy for the detection of various forms of foliar symptoms on proximal high-resolution images. The proposed strategy is based on structure-colour representations and probabilistic models of grapevine tissues. It operates in three steps: (i) Formulating descriptors to extract the characteristic and discriminating properties of each class. They combine the Local Structure Tensors (LST) with colorimetric statistics calculated in pixel's neighbourhood. (ii) Modelling the statistical distributions of these descriptors in each class. To account for the specific nature of LSTs, the descriptors are mapped in the Log-Euclidean space. In this space, the classes of interest can be modelled with mixtures of multivariate Gaussian distributions. (iii) Assigning each pixel to one of the classes according to its suitability to their models. The decision method is based on a "seed growth segmentation" process. This step exploits statistical criteria derived from the probabilistic model. The resulting processing chain reliably detects downy mildew symptoms and estimates the area of the affected tissues. A leave-one-out cross-validation is conducted on a dataset constituted of a hundred independent images of grapevines affected only by downy mildew and/or abiotic stresses. The proposed method achieves an extensive and accurate recovery of foliar symptoms, with on average, a 83% pixel-wise precision and a 76% pixel-wise recall.

Deep Learning Application in Plant Stress Imaging: A Review

Plant stress is one of major issues that cause significant economic loss for growers. The labor-intensive conventional methods for identifying the stressed plants constrain their applications. To address this issue, rapid methods are in urgent needs. Developments of advanced sensing and machine learning techniques trigger revolutions for precision agriculture based on deep learning and big data. In this paper, we reviewed the latest deep learning approaches pertinent to the image analysis of crop stress diagnosis. We compiled the current sensor tools and deep learning principles involved in plant stress phenotyping. In addition, we reviewed a variety of deep learning applications/functions with plant stress imaging, including classification, object detection, and segmentation, of which are closely intertwined. Furthermore, we summarized and discussed the current challenges and future development avenues in plant phenotyping.

Research Progress on the Early Monitoring of Pine Wilt Disease Using Hyperspectral Techniques

Pine wilt disease (PWD) caused by pine wood nematode (PWN, Bursaphelenchus xylophilus) originated in North America and has since spread to Asia and Europe. PWN is currently a quarantine object in 52 countries. In recent years, pine wilt disease has caused considerable economic losses to the pine forest production industry in China, as it is difficult to control. Thus, one of the key strategies for controlling pine wilt disease is to identify epidemic points as early as possible. The use of hyperspectral cameras mounted on drones is expected to enable PWD monitoring over large areas of forest, and hyperspectral images can reflect different stages of PWD. The trend of applying hyperspectral techniques to the monitoring of pine wilt disease is analyzed, and the corresponding strategies to address the existing technical problems are proposed, such as data collection of early warning stages, needs of using unmanned aerial vehicles (UAVs), and establishment of models after preprocessing.

A Simple, Reliable, and Inexpensive Solution for Contact Color Measurement in Small Plant Samples

Correct color measurement by contact-type color measuring devices requires that the sample surface fully covers the head of the device, so their use on small samples remains a challenge. Here, we propose to use cardboard adaptors on the two aperture masks (3 and 8 mm diameter measuring area) of a broadly used portable spectrophotometer. Adaptors in black and white to reduce the measuring area by 50% and 70% were applied in this study. Representatives of the family Campanulaceae have been used to test the methodology, given the occurrence of small leaves. Our results show that, following colorimetric criteria, the only setting providing indistinguishable colors according to the perception of the human eye is the use of a 50%-reducing adaptor on the 3-mm aperture. In addition, statistical analysis suggests the use of the white adaptor. Our contribution offers a sound measurement technique to gather ecological information from the color of leaves, petals, and other small samples.

Determining Mancozeb Deposition Benchmark Values on Apple Leaves for the Management of Venturia inaequalis

Apple scab, caused by Venturia inaequalis, is the most common fruit and foliar disease in commercial apple production worldwide. Early in the production season, preventative contact fungicide sprays are essential for protecting highly susceptible continuously unfolding and expanding young leaves. In South Africa, mancozeb is a key contact fungicide used for controlling apple scab early in the season. The current study developed deposition benchmarks indicative of the biological efficacy of mancozeb against apple scab, using a laboratory-based apple seedling model system. The model system employed a yellow fluorescent pigment that is known to be an effective tracer of mancozeb deposition. A concentration range of mancozeb (0.15 to 1 times the registered dosage) and fluorescent pigment concentrations was sprayed onto seedling leaves, which yielded various fluorescent particle coverage (FPC%) levels. Modeling of the FPC% values versus percent disease control yielded different benchmark values when disease quantification was conducted using two different methods. Thermal infrared imaging (TIRI) disease quantification resulted in a benchmark model where 0.40%, 0.79%, and 1.35 FPC% yielded 50, 75, and 90% apple scab control, respectively. These FPC% values were higher than the benchmarks (0.10, 0.20, and 0.34 FPC%, respectively) obtained with quantitative real-time PCR (qPCR) disease quantification. The qPCR benchmark model is recommended as a guideline for evaluating the efficacy of mancozeb sprays on leaves in apple orchards since the TIRI benchmark model underestimated disease control. The TIRI benchmark model yielded 68% disease control at the lowest mancozeb dosage, yet no visible lesion developed at this dosage. Both benchmark models showed that mancozeb yielded high levels of disease control at very low concentrations; for the qPCR benchmark model the FPC% value of the FPC₉₀ (90% control) corresponded to 0.15 times that of the registered mancozeb concentration in South Africa, i.e., 85% lower than the registered dosage.

High-throughput method for detection and quantification of lesions on leaf scale based on trypan blue staining and digital image analysis

BACKGROUND

Field-grown leafy vegetables can be damaged by biotic and abiotic factors, or mechanically damaged by farming practices. Available methods to evaluate leaf tissue damage mainly rely on colour differentiation between healthy and damaged tissues. Alternatively, sophisticated equipment such as microscopy and hyperspectral cameras can be employed. Depending on the causal factor, colour change in the wounded area is not always induced and, by the time symptoms become visible, a plant can already be severely affected. To accurately detect and quantify damage on leaf scale, including microlesions, reliable differentiation between healthy and damaged tissue is essential. We stained whole leaves with trypan blue dye, which traverses compromised cell membranes but is not absorbed in viable cells, followed by automated quantification of damage on leaf scale.

RESULTS

We present a robust, fast and sensitive method for leaf-scale visualisation, accurate automated extraction and measurement of damaged area on leaves of leafy vegetables. The image analysis pipeline we developed automatically identifies leaf area and individual stained (lesion) areas down to cell level. As proof of principle, we tested the methodology for damage detection and quantification on two field-grown leafy vegetable species, spinach and Swiss chard.

CONCLUSIONS

Our novel lesion quantification method can be used for detection of large (macro) or single-cell (micro) lesions on leaf scale, enabling quantification of lesions at any stage and without requiring symptoms to be in the visible spectrum. Quantifying the wounded area on leaf scale is necessary for generating prediction models for economic losses and produce shelf-life. In addition, risk assessments are based on accurate prediction of the relationship between leaf damage and infection rates by opportunistic pathogens and our method helps determine the severity of leaf damage at fine resolution.

Detection of white head symptoms of panicle blast caused by Pyricularia oryzae using cut-flower dye

BACKGROUND

Breeding of rice with panicle resistance to rice blast disease caused by Pyricularia oryzae is a challenge towards sustainable rice production. Methods for accurate estimation of disease severity can support breeding. White head symptoms are a commonly used index of panicle blast in the field. As the development mechanism of this symptom remains unclear, we used cut-flower dye (CFD) solution to visualize the infected panicle tissues.

RESULTS

CFD delineated the edge of white head symptoms in rice panicles artificially infected with P. oryzae. Hyphae within the tissues were confirmed through staining with a fluorescent wheat germ agglutinin conjugate. Hyphal density was obviously diminished at the dye edge. Growing hyphae preferred to move along the vascular bundles; infected tissues lost the ability to transport water, leading to white head formation. By marking the edge of the white heads, this simple dyeing technique precisely reveals the extent of infection. Further, digital imaging allowed dried samples to be stored and reassessed later.

CONCLUSIONS

The CFD detection technique served as a powerful tool for estimating disease severity by color, as it clearly revealed lesions in both the panicles and leaves. Combined with reliable methods for artificial inoculation and observation of infecting hyphae, this technique will advance the research and breeding of panicle blast-resistant rice.

Object-Based Image Analysis Applied to Low Altitude Aerial Imagery for Potato Plant Trait Retrieval and Pathogen Detection

There is a growing demand in both food quality and quantity, but as of now, one-third of all food produced for human consumption is lost due to pests and other pathogens accounting for roughly 40% of pre-harvest loss in potatoes. Pathogens in potato plants, like the Erwinia bacteria and the PVY^NTN virus for example, exhibit symptoms of varying severity that are not easily captured by pixel-based classes (as these ignore shape, texture, and context in general). The aim of this research is to develop an object-based image analysis (OBIA) method for trait retrieval of individual potato plants that maximizes information output from Unmanned Aerial Vehicle (UAV) RGB very high resolution (VHR) imagery and its derivatives, to be used for disease detection of the Solanum tuberosum. The approach proposed can be split in two steps: (1) object-based mapping of potato plants using an optimized implementation of large scale mean-shift segmentation (LSMSS), and (2) classification of disease using a random forest (RF) model for a set of morphological traits computed from their associative objects. The approach was proven viable as the associative RF model detected presence of Erwinia and PVY pathogens with a maximum F1 score of 0.75 and an average Matthews Correlation Coefficient (MCC) score of 0.47. It also shows that low-altitude imagery acquired with a commercial UAV is a viable off-the-shelf tool for precision farming, and potato pathogen detection.

Identifying Verticillium dahliae Resistance in Strawberry Through Disease Screening of Multiple Populations and Image Based Phenotyping

Verticillium dahliae is a highly detrimental pathogen of soil cultivated strawberry (Fragaria x ananassa). Breeding of Verticillium wilt resistance into commercially viable strawberry cultivars can help mitigate the impact of the disease. In this study we describe novel sources of resistance identified in multiple strawberry populations, creating a wealth of data for breeders to exploit. Pathogen-informed experiments have allowed the differentiation of subclade-specific resistance responses, through studying V. dahliae subclade II-1 specific resistance in the cultivar "Redgauntlet" and subclade II-2 specific resistance in "Fenella" and "Chandler." A large-scale low-cost phenotyping platform was developed utilizing automated unmanned vehicles and near infrared imaging cameras to assess field-based disease trials. The images were used to calculate disease susceptibility for infected plants through the normalized difference vegetation index score. The automated disease scores showed a strong correlation with the manual scores. A co-dominant resistant QTL; FaRVd3D, present in both "Redgauntlet" and "Hapil" cultivars exhibited a major effect of 18.3% when the two resistance alleles were combined. Another allele, FaRVd5D, identified in the "Emily" cultivar was associated with an increase in Verticillium wilt susceptibility of 17.2%, though whether this allele truly represents a susceptibility factor requires further research, due to the nature of the F1 mapping population. Markers identified in populations were validated across a set of 92 accessions to determine whether they remained closely linked to resistance genes in the wider germplasm. The resistant markers FaRVd2B from "Redgauntlet" and FaRVd6D from "Chandler" were associated with resistance across the wider germplasm. Furthermore, comparison of imaging versus manual phenotyping revealed the automated platform could identify three out of four disease resistance markers. As such, this automated wilt disease phenotyping platform is considered to be a good, time saving, substitute for manual assessment.

Plant Disease Diagnosis for Smart Phone Applications with Extensible Set of Diseases

A plant disease diagnosis method that can be implemented with the resources of a mobile phone application, that does not have to be connected to a remote server, is presented and evaluated on citrus diseases. It can be used both by amateur gardeners and by professional agriculturists for early detection of diseases. The features used are extracted from photographs of plant parts like leaves or fruits and include the color, the relative area and the number of the lesion spots. These classification features, along with additional information like weather metadata, form disease signatures that can be easily defined by the end user (e.g., an agronomist). These signatures are based on the statistical processing of a small number of representative training photographs. The extracted features of a test photograph are compared against the disease signatures in order to select the most likely disease. An important advantage of the proposed approach is that the diagnosis does not depend on the orientation, the scale or the resolution of the photograph. The experiments have been conducted under several light exposure conditions. The accuracy was experimentally measured between 70% and 99%. An acceptable accuracy higher than 90% can be achieved in most of the cases since the lesion spots can recognized interactively with high precision.

Comparison of SIFT Encoded and Deep Learning Features for the Classification and Detection of Esca Disease in Bordeaux Vineyards

Grapevine wood fungal diseases such as esca are among the biggest threats in vineyards nowadays. The lack of very efficient preventive (best results using commercial products report 20% efficiency) and curative means induces huge economic losses. The study presented in this paper is centered around the in-field detection of foliar esca symptoms during summer, exhibiting a typical “striped” pattern. Indeed, in-field disease detection has shown great potential for commercial applications and has been successfully used for other agricultural needs such as yield estimation. Differentiation with foliar symptoms caused by other diseases or abiotic stresses was also considered. Two vineyards from the Bordeaux region (France, Aquitaine) were chosen as the basis for the experiment. Pictures of diseased and healthy vine plants were acquired during summer 2017 and labeled at the leaf scale, resulting in a patch database of around 6000 images (224 × 224 pixels) divided into red cultivar and white cultivar samples. Then, we tackled the classification part of the problem comparing state-of-the-art SIFT encoding and pre-trained deep learning feature extractors for the classification of database patches. In the best case, 91% overall accuracy was obtained using deep features extracted from MobileNet network trained on ImageNet database, demonstrating the efficiency of simple transfer learning approaches without the need to design an ad-hoc specific feature extractor. The third part aimed at disease detection (using bounding boxes) within full plant images. For this purpose, we integrated the deep learning base network within a “one-step” detection network (RetinaNet), allowing us to perform detection queries in real time (approximately six frames per second on GPU). Recall/Precision (RP) and Average Precision (AP) metrics then allowed us to evaluate the performance of the network on a 91-image (plants) validation database. Overall, 90% precision for a 40% recall was obtained while best esca AP was about 70%. Good correlation between annotated and detected symptomatic surface per plant was also obtained, meaning slightly symptomatic plants can be efficiently separated from severely attacked plants.

The 'PhenoBox', a flexible, automated, open-source plant phenotyping solution

There is a need for flexible and affordable plant phenotyping solutions for basic research and plant breeding. We demonstrate our open source plant imaging and processing solution ('PhenoBox'/'PhenoPipe') and provide construction plans, source code and documentation to rebuild the system. Use of the PhenoBox is exemplified by studying infection of the model grass Brachypodium distachyon by the head smut fungus Ustilago bromivora, comparing phenotypic responses of maize to infection with a solopathogenic Ustilago maydis (corn smut) strain and effector deletion strains, and studying salt stress response in Nicotiana benthamiana. In U. bromivora-infected grass, phenotypic differences between infected and uninfected plants were detectable weeks before qualitative head smut symptoms. Based on this, we could predict the infection outcome for individual plants with high accuracy. Using a PhenoPipe module for calculation of multi-dimensional distances from phenotyping data, we observe a time after infection-dependent impact of U. maydis effector deletion strains on phenotypic response in maize. The PhenoBox/PhenoPipe system is able to detect established salt stress responses in N. benthamiana. We have developed an affordable, automated, open source imaging and data processing solution that can be adapted to various phenotyping applications in plant biology and beyond.

Computer Vision for High-Throughput Quantitative Phenotyping: A Case Study of Grapevine Downy Mildew Sporulation and Leaf Trichomes

Quantitative phenotyping of downy mildew sporulation is frequently used in plant breeding and genetic studies, as well as in studies focused on pathogen biology such as chemical efficacy trials. In these scenarios, phenotyping a large number of genotypes or treatments can be advantageous but is often limited by time and cost. We present a novel computational pipeline dedicated to estimating the percent area of downy mildew sporulation from images of inoculated grapevine leaf discs in a manner that is time and cost efficient. The pipeline was tested on images from leaf disc assay experiments involving two F₁ grapevine families, one that had glabrous leaves (Vitis rupestris B38 × 'Horizon' [RH]) and another that had leaf trichomes (Horizon × V. cinerea B9 [HC]). Correlations between computer vision and manual visual ratings reached 0.89 in the RH family and 0.43 in the HC family. Additionally, we were able to use the computer vision system prior to sporulation to measure the percent leaf trichome area. We estimate that an experienced rater scoring sporulation would spend at least 90% less time using the computer vision system compared with the manual visual method. This will allow more treatments to be phenotyped in order to better understand the genetic architecture of downy mildew resistance and of leaf trichome density. We anticipate that this computer vision system will find applications in other pathosystems or traits where responses can be imaged with sufficient contrast from the background.

Cluster: A New Application for Spatial Analysis of Pixelated Data for Epiphytotics

Spatial analysis of epiphytotics is essential to develop and test hypotheses about pathogen ecology, disease dynamics, and to optimize plant disease management strategies. Data collection for spatial analysis requires substantial investment in time to depict patterns in various frames and hierarchies. We developed a new approach for spatial analysis of pixelated data in digital imagery and incorporated the method in a stand-alone desktop application called Cluster. The user isolates target entities (clusters) by designating up to 24 pixel colors as nontargets and moves a threshold slider to visualize the targets. The app calculates the percent area occupied by targeted pixels, identifies the centroids of targeted clusters, and computes the relative compass angle of orientation for each cluster. Users can deselect anomalous clusters manually and/or automatically by specifying a size threshold value to exclude smaller targets from the analysis. Up to 1,000 stochastic simulations randomly place the centroids of each cluster in ranked order of size (largest to smallest) within each matrix while preserving their calculated angles of orientation for the long axes. A two-tailed probability t test compares the mean inter-cluster distances for the observed versus the values derived from randomly simulated maps. This is the basis for statistical testing of the null hypothesis that the clusters are randomly distributed within the frame of interest. These frames can assume any shape, from natural (e.g., leaf) to arbitrary (e.g., a rectangular or polygonal field). Cluster summarizes normalized attributes of clusters, including pixel number, axis length, axis width, compass orientation, and the length/width ratio, available to the user as a downloadable spreadsheet. Each simulated map may be saved as an image and inspected. Provided examples demonstrate the utility of Cluster to analyze patterns at various spatial scales in plant pathology and ecology and highlight the limitations, trade-offs, and considerations for the sensitivities of variables and the biological interpretations of results. The Cluster app is available as a free download for Apple computers at iTunes, with a link to a user guide website.

Automated Identification of Northern Leaf Blight-Infected Maize Plants from Field Imagery Using Deep Learning

Northern leaf blight (NLB) can cause severe yield loss in maize; however, scouting large areas to accurately diagnose the disease is time consuming and difficult. We demonstrate a system capable of automatically identifying NLB lesions in field-acquired images of maize plants with high reliability. This approach uses a computational pipeline of convolutional neural networks (CNNs) that addresses the challenges of limited data and the myriad irregularities that appear in images of field-grown plants. Several CNNs were trained to classify small regions of images as containing NLB lesions or not; their predictions were combined into separate heat maps, then fed into a final CNN trained to classify the entire image as containing diseased plants or not. The system achieved 96.7% accuracy on test set images not used in training. We suggest that such systems mounted on aerial- or ground-based vehicles can help in automated high-throughput plant phenotyping, precision breeding for disease resistance, and reduced pesticide use through targeted application across a variety of plant and disease categories.

Real-Time Continuous Identification of Greenhouse Plant Pathogens Based on Recyclable Microfluidic Bioassay System

The development of a real-time continuous analytical platform for the pathogen detection is of great scientific importance for achieving better disease control and prevention. In this work, we report a rapid and recyclable microfluidic bioassay system constructed from oligonucleotide arrays for selective and sensitive continuous identification of DNA targets of fungal pathogens. We employ the thermal denaturation method to effectively regenerate the oligonucleotide arrays for multiple sample detection, which could considerably reduce the screening effort and costs. The combination of thermal denaturation and laser-induced fluorescence detection technique enables real-time continuous identification of multiple samples (<10 min per sample). As a proof of concept, we have demonstrated that two DNA targets of fungal pathogens (Botrytis cinerea and Didymella bryoniae) can be sequentially analyzed using our rapid microfluidic bioassay system, which provides a new paradigm in the design of microfluidic bioassay system and will be valuable for chemical and biomedical analysis.

Direct quantitative evaluation of disease symptoms on living plant leaves growing under natural light

Leaf color is an important indicator when evaluating plant growth and responses to biotic/abiotic stress. Acquisition of images by digital cameras allows analysis and long-term storage of the acquired images. However, under field conditions, where light intensity can fluctuate and other factors (shade, reflection, and background, etc.) vary, stable and reproducible measurement and quantification of leaf color are hard to achieve. Digital scanners provide fixed conditions for obtaining image data, allowing stable and reliable comparison among samples, but require detached plant materials to capture images, and the destructive processes involved often induce deformation of plant materials (curled leaves and faded colors, etc.). In this study, by using a lightweight digital scanner connected to a mobile computer, we obtained digital image data from intact plant leaves grown in natural-light greenhouses without detaching the targets. We took images of soybean leaves infected by Xanthomonas campestris pv. glycines, and distinctively quantified two disease symptoms (brown lesions and yellow halos) using freely available image processing software. The image data were amenable to quantitative and statistical analyses, allowing precise and objective evaluation of disease resistance.

Characterization of Botrytis-plant interactions using PathTrack© -an automated system for dynamic analysis of disease development

The measurement of disease development is integral in studies on plant-microbe interactions. To address the need for a dynamic and quantitative disease evaluation, we developed PathTrack© , and used it to analyse the interaction of plants with Botrytis cinerea. PathTrack© is composed of an infection chamber, a photography unit and software that produces video files and numerical values of disease progression. We identified a previously unrecognized infection stage and determined numerical parameters of pathogenic development. Using these parameters, we identified differences in disease dynamics between seemingly similar B. cinerea pathogenicity mutants, and revealed new details on plant susceptibility to the fungus. We showed that the difference between the lesion expansion rate on leaves and colony spreading rate on artificial medium reflects the levels of the plant immune system, suggesting that this parameter can be used to quantify plant defence. Our results shed new light and reveal new details of the interaction between the model necrotrophic pathogen B. cinerea and plants. The concept that we present is universal and may be applied to facilitate the study of various types of plant-pathogen association.

Effect of Spectral Quality of Monochromatic LED Lights on the Growth of Artichoke Seedlings

Indoor farming is becoming a popular alternative approach in food production to meet the demand of a growing world population. Under this production system, artificial light provides the main source of illumination in sustaining plant growth and development. The use of light-emitting diodes (LEDs) is a popular source of artificial light for indoor farms due to its narrow light spectra, modular design and energy efficiency. This study purposely assessed the effect of monochromatic LED light quality on the growth of three varieties of artichoke seedlings compared to greenhouse condition. Spectral quality assessment showed that photosynthetic photon flux density (PPFD) was highest under red LED light, but only a third of the total PPFD under natural light. Seedlings grown under red light showed 60-100% more shoot dry weight and were 67-115% taller than seedlings grown in the greenhouse. However, seedlings under blue or white light conditions showed 67-76% less in biomass compared to greenhouse-grown seedlings. Overall, plant response of seedlings under red light condition was much better compared to greenhouse-grown seedlings emphasizing the importance of red light spectral quality in plant growth and development.

Thermal imaging of soybean response to drought stress: the effect of Ascophyllum nodosum seaweed extract

Previous experiments have demonstrated positive effect of Acadian(®) extract of Ascophyllum nodosum on plant stress-resistance, however the mode of action is not fully understood. The aim of this study was to understand the physiological effect of Acadian(®) seaweed extract on the plant response to drought stress. Leaf temperature and leaf angle were measured as early-stage indicators of plant stress with thermal imaging "in situ" over a 5-day stress-recovery trial. The early stress-response of control became visible on the third day as a rapid wilting of leaves, accompanied with the asymptotic increase of leaf temperature on 4-5 °C to the thermal equilibrium with ambient air temperature. At the same time Acadian(®) treated plants still maintained turgor, accompanied with the linear increase in leaf temperature, which indicated better control of stomatal closure. Re-watering on the fifth day showed better survival of treated plants compared to control. This study demonstrated the ability of Acadian(®) seaweed extract to improve resistance of soybean plants to water stress.