No description available.

In the United States alone, common diseases spread among plants account for billions of dollars lost in crop yield each year. This issue is exacerbated in countries with less infrastructure to defend against crop epidemics, and can lead to famine and forced migration. Farmers can seek the help of plant pathology experts to defend against diseases and detect crop irregularities early on. However, access to experts can be difficult, and even those trained in the field may miss symptoms before it is too late. To assist in early disease detection, a number of papers have been released on the potential for machine learning image classifiers to identify healthy plants from infected ones using convolutional neural networks. While these papers are promising, they often fail to implement a set of standardized practices in their model implementation or make use of realistic data sets. This thesis outlines a set of best practices to use when creating a convolutional neural network for plant disease detection. These principles were selected through a combination of related work analysis and generalized best practices on machine learning. A selection of 11 research articles that discuss their own plant disease image classifiers are analyzed on the grounds of these principles to assess their validity. Then, to demonstrate these principles in practice, we trained six models that each follow our set of guidelines to distinguish healthy strawberry plant images from diseased ones. While the focus of our paper centers on the need to use these practices to create field-realistic models, we achieved the best results on our strawberry image classifier using a VGG16 model architecture. We hope that this work will inspire a set of standardized practices to follow when developing a plant disease image classifier, and allow for more accurate model comparisons in the future.

In the most recent years, many deep learning models have been used to identify and classify diseases of plant leaves by inputting plant leaf images as input to the model. However, there is still a gap in research on how to improve the accuracy of the deep learning models of plant leaf diseases. This thesis is about investigating various techniques for improving the MobileNetV2 model's accuracy for plant disease detection in leaves and classification. These techniques involved adjusting the learning rate, adding additional layers, and various data-augmented operations. The results of this thesis have shown that these techniques can significantly improve the accuracy of the model, and the best results can be achieved by using random rotation and crop data augmentation. After adding random rotation and crop data augmentation to the model, it achieved an accuracy of 94%, a precision of 91%, a recall of 96%, and an F1-score of 95%. This shows that the proposed techniques can be used to improve the accuracy of plant leaf disease detection and classification models, which can help farmers identify and treat plant diseases.

No description available.

Plant pathogens are one of the biggest threats to plant health and food security worldwide. To effectively contain plant disease outbreaks, classification and precise identification of pathogens is crucial to determine treatment and preventive measurements. Conventional methods of detection such as PCR may not be sufficient when the pathogen in question is unknown. Advances in sequencing technology have made it possible to sequence entire genomes and metagenomes in real-time and at a relatively low cost, opening an opportunity for the development of alternative methods for detection of novel and unknown plant pathogens. Within this dissertation, an integrated approach is used to reclassify a high-impact group of plant pathogens. Additionally, the application of metagenomics and nanopore sequencing using the Oxford Nanopore Technologies (ONT) MinION for fungal and bacterial plant pathogen detection and precise identification are demonstrated. To improve the classification of the strains belonging to the Ralstonia solanacearum species complex (RSSC), we performed a meta-analysis using a comparative genomics and a reverse ecology approach to accurately portray and refine the understanding of the diversity and evolution of the RSSC. The groups identified by these approaches were circumscribed and made publicly available through the LINbase web server so future isolates can be properly classified. To develop a culture-free detection method of plant pathogens, we used metagenomes of various plants and long-read nanopore sequencing to precisely identify plant pathogens to the strain-level and performed phylogenetic analysis with SNP resolution. In the first paper, we used tomato plants to demonstrate the detection power of bacterial plant pathogens. We compared bioinformatics tools for detection at the strain-level using reads and assemblies. In the second paper, we used a read-based approach to test the feasibility of the methodology to precisely detect the fungal pathogen causing boxwood blight. Lastly, with the improvement in nanopore sequencing, we used grapevine petioles to investigate whether we can go beyond detection and identification and do a phylogenetic analysis. We assembled a metagenome-assembled genome (MAG) of almost the same quality as the genomes obtained from cultured isolates and did a phylogenetic analysis with SNP resolution. Finally, for the cases where there may be no related genome in the database like the pathogen in question, we used machine learning and metagenomics to develop a reference-free approach to detection of plant diseases. We trained eight different machine learning models with reads from healthy and infected plant metagenomes and compared the classification accuracy of reads as belonging to a healthy or infected plant. From the comparison, random forest was the best model in terms of computational resources needed while maintaining a high accuracy (> 0.90). / Doctor of Philosophy / Microbes are present in every environment on the planet and have been on Earth for billions of years. While some microbes are beneficial, others can cause diseases. To differentiate the ones causing diseases from those who do not, looking into the evolutionary forces making them different is crucial to classify and identify them correctly. Although microorganisms cause diseases in humans and animals, the ones causing diseases in plants are one of the biggest threats to plant health and food security worldwide. In a perfect world, plant diseases would be diagnosed by eye or simple procedures. However, when a plant disease is present, it is not always obvious which organism, if any, is causing the disease making it hard for outbreaks to be detected and contained promptly. With technological advances, it is now possible to obtain all the genetic information of not only one organism but all the organisms living in an environment at a time. This genetic information can then be used to precisely identify what organism is causing a disease in a plant for faster disease diagnosis and, consequently, more efficient disease prevention and control. In this dissertation, we used the bacterial group, called Ralstonia solanacearum species complex, which can cause different diseases in more than 200 crops, to investigate and understand the evolution and diversity of the members of this group. We also used newly developed technologies to obtain the genetic material of all the organisms living in multiple important plants including tomato, grapevine, and the ornamental bush, boxwood. Using this genetic material, we developed a methodology for the detection of bacteria and a fungus causing plant diseases. While this works well when the suspected organism or a similar one is available for comparison, the detection of plant diseases in cases where this information is not available is challenging. Machine learning models, where computers can learn complex patterns from data, have the potential to detect pathogens without the need to compare the sequences to sequences of other pathogens. Here we also used the genetic material to train and compare different machine learning models to classify plants as either being infected or healthy.

No description available.

Melon (Cucumis melo L.) and watermelon (Citrullus lanatus (Thunb.) Matsum and Nakai) belong to the family Cucurbitaceae. Bacterial fruit blotch (BFB) disease of cucurbits is an economically devastating plant disease that has caused an estimated loss of up to $450M on watermelon crops and $75M (worldwide) to the seed and transplant industries since 1996. Disease symptoms include water-soaked cotyledons, leaf necrosis, and internal fruit rot. Current commercial management strategies are very limited and include: seed production field sanitation, greenhouse transplant sanitation, copper-based bactericide sprays, crop rotation, disease-free healthy seeds, isolating diseased plants, and peroxyacetic acid seed treatments. The seedborne disease is usually spread by contaminated seeds, and there is a zero-tolerance policy in the seed industry for infected seeds. No nondestructive assays are commercially available to detect BFB in seeds. This research investigated several different aspects of BFB disease such as non-destructive seed detection, green tea seed treatment, candidate NB-LRR genes for disease resistance, and optimization of virus induced gene silencing for melon and watermelon crops. The potential application of attenuated total reflectance (ATR) Fourier transform infrared spectroscopy (ATR-FTIR) and high-resolution X-ray analysis methods for detection of BFB on seeds were evaluated. It was possible to detect BFB in seeds that were pistil inoculated via x-ray imaging and pericarp inoculated via ATR FT-IR. In vitro and in vivo experiments evaluated the potential of tea (Camellia sinensis) and tea polyphenols as seed treatments to sanitize seeds infected with A. citrulli. Green tea unlike black tea inhibited growth of A. citrulli because of polyphenols. Eighty one melon and forty four watermelon NB-LRR genes were reidentified, and genes that have potential resistance against A. citrulli on melon plants were screened based on host selectivity of the pathogen. Finally, the virus-induced, gene-silencing method was optimized for melon and watermelon for further analysis of potential disease resistance genes. BFB can be nondestructively identified in seeds and green tea may be an effective seed treatment with further development. Promising candidate R genes were identified that might confer stable resistance in the right genetic background. / Doctor of Philosophy / Melon and watermelon crops both belong to the gourd family. Bacterial fruit blotch (BFB) disease of cucurbits is an economically devastating plant disease that has caused an estimated loss of up to $450M on watermelon crops and $75M (worldwide) to the seed and transplant industries since 1996. Disease symptoms include water-soaked cotyledons, leaf necrosis, and internal fruit rot. Current commercial management strategies and detection methods are very limited. The seedborne disease is usually spread by contaminated seeds, and there is a zero-tolerance policy in the seed industry for infected seeds. This research investigated several different aspects of BFB disease such as non-destructive seed detection, green tea seed treatment, candidate disease resistance genes, and optimization of virus induced gene silencing methodology for melon and watermelon crops. There are currently no nondestructive assays available to detect BFB in seeds. We evaluated the potential application of attenuated total reflectance (ATR) Fourier transform infrared spectroscopy (ATR-FTIR) and high-resolution X-ray analysis methods for detection of BFB on seeds. It was possible to detect BFB inside layers of seeds that were naturally inoculated through the flowers via x-ray imaging and seedcoat inoculated via ATR FT-IR. In vitro and in vivo experiments evaluated the potential of tea and tea constituents as seed treatments to sanitize seeds infected with BFB. Green tea unlike black tea inhibited growth of BFB. Eighty one melon and forty four watermelon disease resistance genes were reidentified and genes that have potential resistance against BFB on melon plants were screened based on host selectivity of the pathogen. Finally, the virus induced gene silencing method was optimized for melon and watermelon plants for further analysis of potential disease resistance genes. BFB can be nondestructively identified in seeds and green tea may be an effective seed treatment with further development. Promising candidate resistance genes were identified that might confer stable resistance in the right genetic background.

Stem rot of peanut (Arachis hypogaea L.), caused by a soilborne fungus Athelia rolfsii (Curzi) C. C. Tu and Kimbr. (anamorph: Sclerotium rolfsii Sacc.), is one of the most important diseases in peanut production worldwide. Though new varieties with increased partial resistance to this disease have been developed, there is still a need to utilize fungicides for disease control during the growing season. Fungicides with activity against A. rolfsii are available, and several new products have been recently registered for control of stem rot in peanut. However, fungicides are most effective when applied before or during the early stages of infection. Current scouting methods can detect disease once signs or symptoms are present, but to optimize the timing of fungicide applications and protect crop yield, a method for early detection of soilborne diseases is needed. Previous studies have utilized optical sensors combined with machine learning analysis for the early detection of plant diseases, but these studies mainly focused on foliar diseases. Few studies have applied these technologies for the early detection of soilborne diseases in field crops, including peanut. Thus, the overall goal of this research was to integrate sensor technologies, modern data analytic tools, and properties of standard and newly registered fungicides to develop improved management strategies for stem rot control in peanuts. The specific objectives of this work were to 1) characterize the spectral and thermal responses of peanut to infection with A. rolfsii under controlled conditions, 2) identify optimal wavelengths to detect stem rot of peanut using hyperspectral sensor and machine learning, and 3) evaluate the standard and newly registered peanut fungicides with different modes of action for stem rot control in peanuts using a laboratory bioassay. For Objective 1, spectral reflectance and leaf temperature of peanut plants were measured by spectral and thermal sensors in controlled greenhouse experiments. Differences in sensor-based responses between A. rolfsii-infected and non-infected plants were detected 0 to 1 day after observation of foliar disease symptoms. In addition, spectral responses of peanut to the infection of A. rolfsii were more pronounced and consistent than thermal changes as the disease progressed. Objective 2 aimed to identify specific signatures of stem rot from reflectance data collected in Objective 1 utilizing a machine learning approach. Wavelengths around 505, 690, and 884 nm were repeatedly selected by different methods. The top 10 wavelengths identified by the recursive feature selection methods performed as well as all bands for the classification of healthy peanut plants and plants at different stages of disease development. Whereas the first two objectives focused on disease detection, Objective 3 focused on disease control and compared the properties of different fungicides that are labeled for stem rot control in peanut using a laboratory bioassay of detached peanut tissues. All of the foliar-applied fungicides evaluated provided inhibition of A. rolfsii for up to two weeks on plant tissues that received a direct application. Succinate dehydrogenase inhibitors provided less basipetal protection of stem tissues than quinone outside inhibitor or demethylation inhibitor fungicides. Overall, results of this research provide a foundation for developing sensor/drone-based methods that use disease-specific spectral indices for scouting in the field and for making fungicide application recommendations to manage stem rot of peanut and other soilborne diseases. / Doctor of Philosophy / Plant diseases are a major constraint to crop production worldwide. Developing effective and economical management strategies for these diseases, including selection of proper fungicide chemistries and making timely fungicide application, is dependent on the ability to accurately detect and diagnose their signs and/or symptoms prior to widespread development in a crop. Optical sensors combined with machine learning analysis are promising tools for automated crop disease detection, but research is still needed to optimize and validate methods for the detection of specific plant diseases. The overarching goal of this research was to use the peanut-stem rot plant disease system to identify and evaluate sensor-based technologies and different fungicide chemistries that can be utilized for the management of soilborne plant diseases. The specific objectives of this work were to 1) characterize the temporal progress of spectral and thermal responses of peanut to infection and colonization with Athelia rolfsii, the causal agent of peanut stem rot 2) identify optimal wavelengths to detect stem rot of peanut using hyperspectral sensor and machine learning, and 3) evaluate standard and newly registered peanut fungicides with different modes of action for stem rot control in peanuts using a laboratory bioassay. Results of this work demonstrate that spectral reflectance measurements are able to distinguish between diseased and healthy plants more consistently than thermal measurements. Several wavelengths were identified using machine learning approaches that can accurately differentiate between peanut plants with symptoms of stem rot and non-symptomatic plants. In addition, a new method was developed to select the top-ranked, non-redundant wavelengths with a custom distance. These selected wavelengths performed better than using all wavelengths, providing a basis for designing low-cost optical filters to specifically detect this disease. In the laboratory bioassay evaluation of fungicides, all of the foliar-applied fungicides provided inhibition of A. rolfsii for up to two weeks on leaf tissues that received a direct application. Percent inhibition of A. rolfsii decreased over time, and the activity of all fungicides decreased at a similar rate. Overall, the findings of this research provide a foundation for developing sensor-based methods for disease scouting and making fungicide application recommendations to manage stem rot of peanut and other soilborne diseases.

Submitted by GEORGE DEROCO MARTINS null (deroco87@hotmail.com) on 2017-02-08T14:17:06Z No. of bitstreams: 1 martins_gd_dr_prud.pdf: 4750620 bytes, checksum: 2b77564f0f8c206ce23e6d7ca4bac5d6 (MD5) / Approved for entry into archive by LUIZA DE MENEZES ROMANETTO (luizamenezes@reitoria.unesp.br) on 2017-02-13T18:01:23Z (GMT) No. of bitstreams: 1 martins_gd_dr_prud.pdf: 4750620 bytes, checksum: 2b77564f0f8c206ce23e6d7ca4bac5d6 (MD5) / Made available in DSpace on 2017-02-13T18:01:23Z (GMT). No. of bitstreams: 1 martins_gd_dr_prud.pdf: 4750620 bytes, checksum: 2b77564f0f8c206ce23e6d7ca4bac5d6 (MD5) Previous issue date: 2016-12-19 / Coordenação de Aperfeiçoamento de Pessoal de Nível Superior (CAPES) / Os nematoides são importantes fitoparasitas que se constituem em um problema sério para o cultivo do café no Brasil. Como a ocorrência de nematoides no sistema radicular do cafeeiro causa desequilíbrios nutricionais na planta que provocam variações na resposta espectral da folha e define uma configuração espacial característica às áreas infectadas, o objetivo desta pesquisa avaliar o potencial de dados de sensoriamento remoto adquiridos em multiescala para discriminar e mapear o café sadio, em estágio inicial de infecção e severamente infectado. A pesquisa foi desenvolvida em três áreas experimentais, localizadas no sul do estado de Minas Gerais, nas quais foi certificada a ocorrência de nematoides e realizadas medições de variáveis biofísicas e dados hiperespectrais na folha e sobre o dossel da planta. Os dados hiperespectrais também foram utilizados em simulação de bandas dos sensores do RapidEye e OLI/Landsat 8 para identificar as faixas espectrais mais sensíveis para a discriminação de patógenos em plantas de café. Nenhum dos parâmetros biofísicos avaliados discriminou eficientemente as folhas de plantas sadias e infectadas, mas a simulação de bandas indicou que os intervalos espectrais do vermelho, vermelho limítrofe e infravermelho próximos do RapidEye foram complementares para a discriminação de plantas de café sadio e dos dois níveis de infecção. Essas bandas, mais uma imagem NDVI, foram utilizadas na classificação das áreas infectadas por nematoides, a qual definiu a distribuição espacial de café sadio e dos dois níveis de infecção, com uma acurácia global de 78% e coeficiente kappa de 0,71. A classificação não supervisionada da imagem multiespectral OLI/Landsat 8 também definiu as três condições, porém com baixa confiabilidade (coeficiente kappa igual a 0,41). Por outro lado, uma inferência espacial quantitativa da concentração de nematoides/cm³ no solo, a partir de um modelo empírico baseado na imagem RapidEye, apresentou um erro consideravelmente alto (21,89%). / Nematodes are important phytoparasites that constitute a serious issue for coffee cultivation in Brazil. Because root infection by nematodes induces spectral variation in leaves and defines a unique spatial configuration in the cultivation field, the aim of this study is to evaluate the potential of remote sensing data acquired in multiscale to discriminate and map healthy, early infected and severely infected coffee plants. This study was carried out in three experimental areas, located in the in southern Minas Gerais State, in which the occurrence of nematodes was certified and biophysical and hyperspectral measurements of the leaves and on the canopy were made. Hyperspectral data were also used to simulate the bands of the RapidEye and OLI/Landsat 8 sensors to identify the most sensitive spectral ranges for pathogen discrimination in coffee plants. None of the biophysical parameters efficiently discriminated the leaves of healthy and infected plants, but the band simulations indicated that red, red edge and near infrared spectral ranges were complementary to the discrimination of healthy coffee plants and the two levels of infection. These bands, plus an (NDVI) image, were used for a multispectral classification of healthy and nematode-infected areas. The multispectral classification defined the spatial distribution of healthy, early infected and two levels of infection, with an overall accuracy of 78% and kappa coefficient of 0.71. The unsupervised classification of the multispectral image OLI/Landsat 8 also defined the three conditions, but with low reliability (kappa coefficient equal to 0.41). In contrast, a quantitative spatial inference of the soil nematode concentration/cm³, from an empirical model based on the RapidEye image, presented a considerably high error (21.89%).

空間群聚分析應用於流行病學已行之有年，但國內這方面的研究仍較缺乏，尤其在找出哪些地區有較高疾病發生率的群聚偵測。本文針對台灣鄉鎮市資料的特性，提出一套合適的群聚檢測方法，這個方法使用兩階段的電腦模擬，實證上更容易使用；這個方法除了可找出最大顯著群聚外，也能夠偵測出多個群聚的分佈。本文使用電腦模擬比較本文的方法與目前使用較為廣泛的方法(包括Kulldorff(1995)的spatial scan statistic和Tango(2005)的flexible scan statistic)，以型一誤差、型二誤差及錯誤率三種標準衡量方法的優劣。最後套用台灣癌症死亡率與健保就診次數資料，探討台灣癌症空間群聚與就診情形的變化。 / Spatial cluster analyses have applied in epidemiology for many years. In this topic there still are few researches in Taiwan, especially in detecting the areas which have higher disease intensity. In this paper, we proposed a new cluster detection method which is aimed at Taiwan counties’ data. This method which uses two-stage computer simulation procedures is useful in practice. This method can find the most likely cluster. Besides, it can find multiple clusters. We use computer simulations to compare our method with others (Kulldorff’s spatial scan statistic& Tango’s flexible scan statistic). Type-I error, Type-II error and error rate are criterions of measurement. At last, we use Taiwan cancer mortality data and all the people health insurance data to discuss Taiwan cancer spatial clusters and the change of diagnoses.