Automated tomato leaf disease classification using transfer learning-based deep convolution neural network

Journal of Plant Diseases and Protection - Early and accurate detection of plant diseases is necessary to maximize crop yield. The artificial intelligence based deep learning method plays a vital...     

基于卷积神经网络的农作物叶片病害诊断

农作物病害诊断对于及时发现并采取防控措施具有重要意义.本研究针对苹果、玉米、番茄、葡萄等典型农作物的常见叶片病害问题,使用了两种目前使用最广泛的卷积神经网络——VGG 16及Resnet50,对典型农作物叶片病害进行识别.使用AI Challenger比赛的农作物叶片病害数据集图像,并对这些图像进行预处理,构建了47 ...     

Identification of tea foliar diseases and pest damage under practical field conditions using a convolutional neural network

Lesions of tea (Camellia sinensis) leaves are detrimental to the growth of tea crops. Their adverse effects include further disease of tea leaves and a direct reduction in yield and profit. Therefore, early detection and on-site monitoring of tea leaf lesions are necessary for effective management to control infections and prevent further yield loss. In this study, 1,822 images of tea leaves with lesions caused by three diseases (brown blight, Colletotrichum camelliae; blister blight, Exobasidium vexans; and algal leaf spot, Cephaleuros virescens) and four pests (leaf miner, Tropicomyia theae; tea thrip, Scirtothrips dorsalis; tea leaf roller, Homona magnanima; and tea mosquito bug, Helopeltis fasciaticollis) were collected from northern and central Taiwan. A faster region-based convolutional neural network (Faster R-CNN) was then trained to detect the locations of the lesions on the leaves and to identify the causes of the lesions. The trained Faster R-CNN detector achieved a precision of 77.5%, recall of 70.6%, an F1 score of 73.91%, and a mean average precision of 66.02%. An overall accuracy of 89.4% was obtained for identification of the seven classes of tea diseases and pests. The developed detector could assist tea farmers in identifying the causes of lesions in real time.     

Weather-based prediction of anthracnose severity using artificial neural network models

Data were collected and analysed from seven field sites in Australia, Brazil and Colombia on weather conditions and the severity of anthracnose disease of the tropical pasture legume Stylosanthes scabra caused by Colletotrichum gloeosporioides . Disease severity and weather data were analysed using artificial neural network (ANN) models developed using data from some or all field sites in Australia and/or South America to predict severity at other sites. Three series of models were developed using different weather summaries. Of these, ANN models with weather for the day of disease assessment and the previous 24 h period had the highest prediction success, and models trained on data from all sites within one continent correctly predicted disease severity in the other continent on more than 75% of days; the overall prediction error was 21·9% for the Australian and 22·1% for the South American model. Of the six cross-continent ANN models trained on pooled data for five sites from two continents to predict severity for the remaining sixth site, the model developed without data from Planaltina in Brazil was the most accurate, with >85% prediction success, and the model without Carimagua in Colombia was the least accurate, with only 54% success. In common with multiple regression models, moisture-related variables such as rain, leaf surface wetness and variables that influence moisture availability such as radiation and wind on the day of disease severity assessment or the day before assessment were the most important weather variables in all ANN models. A set of weights from the ANN models was used to calculate the overall risk of anthracnose for the various sites. Sites with high and low anthracnose risk are present in both continents, and weather conditions at centres of diversity in Brazil and Colombia do not appear to be more conducive than conditions in Australia to serious anthracnose development.     

基于深度学习人工神经网络的青椒调亏灌溉水量预测

在2014—2018年,采用垄沟集雨覆盖种植滴灌技术与调亏灌溉技术相结合（MFR-RDI）对青椒进行试验研究,选取灌溉水利用效率最高的试验处理（即青椒结果后期重度亏水）进行灌水量预测。根据试验期间搜集的各项资料,在MFR-RDI种植方式下,以作物需水量、青椒生育期天数、作物生育期内的降水量、土壤含水率、前一天的灌水量作为模型输入因子,构建青椒作物灌水量的深度学习人工神经网络（DNN）预测模型。通过模型试验得到最佳DNN预测模型,该模型的隐含层包括4层,各隐含层神经元个数分别为：32、16、8、4。模型的激活函数采用“ReLU”,优化函数为“adam”,迭代次数为300。模型使用2018年的数据进行了测试。测试结果表明DNN模型的RMSE为0.898 mm,MAE为0.257 mm,NS为0.758,R²为0.7635,说明该预测模型具有较高的精度性能。通过预测结果可以得到此种植方式下青椒的灌溉制度,为实现高效智能节水灌溉提供参考。     

Image recognition and empirical application of desert plant species based on convolutional neural network

In recent years, deep convolution neural network has exhibited excellent performance in computer vision and has a far-reaching impact. Traditional plant taxonomic identification requires high expertise, which is time-consuming. Most nature reserves have problems such as incomplete species surveys, inaccurate taxonomic identification, and untimely updating of status data. Simple and accurate recognition of plant images can be achieved by applying convolutional neural network technology to explore the best network model. Taking 24 typical desert plant species that are widely distributed in the nature reserves in Xinjiang Uygur Autonomous Region of China as the research objects, this study established an image database and select the optimal network model for the image recognition of desert plant species to provide decision support for fine management in the nature reserves in Xinjiang, such as species investigation and monitoring, by using deep learning. Since desert plant species were not included in the public dataset, the images used in this study were mainly obtained through field shooting and downloaded from the Plant Photo Bank of China (PPBC). After the sorting process and statistical analysis, a total of 2331 plant images were finally collected (2071 images from field collection and 260 images from the PPBC), including 24 plant species belonging to 14 families and 22 genera. A large number of numerical experiments were also carried out to compare a series of 37 convolutional neural network models with good performance, from different perspectives, to find the optimal network model that is most suitable for the image recognition of desert plant species in Xinjiang. The results revealed 24 models with a recognition Accuracy, of greater than 70.000%. Among which, Residual Network X_8GF (RegNetX_8GF) performs the best, with Accuracy, Precision, Recall, and F1 (which refers to the harmonic mean of the Precision and Recall values) values of 78.33%, 77.65%, 69.55%, and 71.26%, respectively. Considering the demand factors of hardware equipment and inference time, Mobile NetworkV2 achieves the best balance among the Accuracy, the number of parameters and the number of floating-point operations. The number of parameters for Mobile Network V2 (MobileNetV2) is 1/16 of RegNetX_8GF, and the number of floating-point operations is 1/24. Our findings can facilitate efficient decision-making for the management of species survey, cataloging, inspection, and monitoring in the nature reserves in Xinjiang, providing a scientific basis for the protection and utilization of natural plant resources.     

基于小波神经网络和BP神经网络的麦蚜发生期预测对比

为建立更准确、稳定的病虫害预测预报模型,减少农作物病虫害损失、提高农作物产量与质量,运用主成分分析法从42个基础气象因子中整合形成8个新的自变量输入模型,采用试凑法对网络关键参数进行筛选,用2002-2011年数据进行网络训练,建立了以Morlet小波函数为传递函数的小波神经网络模型,并与以Sigmoid函数为传递函数的BP神经网络模型进行了比较.在小波神经网络训练过程中,有6年拟合精度在90％以上,平均拟合精度为89％,预测结果MAPE值为4.1939,MSE值为5.9764;在BP神经网络的训练过程中,有4年拟合精度超过90％,平均拟合精度仅为81.07％,预测结果中MAPE值为6.4694,MSE值为8.2457.从训练结果看,小波神经网络更能准确描述麦蚜发生期的变化规律,其拟合能力较BP神经网络好;从预测精度和模型的稳定性来看,小波神经网络好于BP神经网络.     

Multiple Alternaria species groups are associated with leaf blotch and fruit spot diseases of apple in Australia

Leaf blotch and fruit spot of apple caused by Alternaria species occur in apple orchards in Australia. However, there is no information on the identity of the pathogens and whether one or more Alternaria species cause both diseases in Australia. Using DNA sequencing and morphological and cultural characteristics of 51 isolates obtained from apple leaves and fruit with symptoms in Australia, Alternaria species groups associated with leaf blotch and fruit spot of apples were identified. Sequences of Alternaria allergen a1 and endopolygalacturonase gene regions revealed that multiple Alternaria species groups are associated with both diseases. Phylogenetic analysis of concatenated sequences of the two genes resulted in four clades representing A. arborescens and A. arborescens‐like isolates in clade 1, A. tenuissima/A. mali isolates in clade 2, A. alternata/A. tenuissima intermediate isolates in clade 3 and A. longipes and A. longipes‐like isolates in clade 4. The clades formed using sequence information were supported by colony characteristics and sporulation patterns. The source of the isolates in each clade included both the leaf blotch variant and the fruit spot variant of the disease. Alternaria arborescens‐like isolates were the most prevalent (47%) and occurred in all six states of Australia, while A. alternata/A. tenuissima intermediate isolates (14%) and A. tenuissima/A. mali isolates (6%) occurred mostly in Queensland and New South Wales, respectively. Implications of multiple Alternaria species groups on apples in Australia are discussed.     

Regression and artificial neural network modeling for the prediction of gray leaf spot of maize

ABSTRACT Regression and artificial neural network (ANN) modeling approaches were combined to develop models to predict the severity of gray leaf spot of maize, caused by Cercospora zeae-maydis. In all, 329 cases consisting of environmental, cultural, and location-specific variables were collected for field plots in Iowa between 1998 and 2002. Disease severity on the ear leaf at the dough to dent plant growth stage was used as the response variable. Correlation and regression analyses were performed to select potentially useful predictor variables. Predictors from the best 9 of 80 regression models were used to develop ANN models. A random sample of 60% of the cases was used to train the networks, and 20% each for testing and validation. Model performance was evaluated based on coefficient of determination (R(2)) and mean square error (MSE) for the validation data set. The best models had R(2) ranging from 0.70 to 0.75 and MSE ranging from 174.7 to 202.8. The most useful predictor variables were hours of daily temperatures between 22 and 30 degrees C (85.50 to 230.50 h) and hours of nightly relative humidity >/=90% (122 to 330 h) for the period between growth stages V4 and V12, mean nightly temperature (65.26 to 76.56 degrees C) for the period between growth stages V12 and R2, longitude (90.08 to 95.14 degrees W), maize residue on the soil surface (0 to 100%), planting date (in day of the year; 112 to 182), and gray leaf spot resistance rating (2 to 7; based on a 1-to-9 scale, where 1 = most susceptible to 9 = most resistant).     

Comparison of sampling procedures for two rice diseases: leaf blast and tungro

The precision and accuracy of eight random and systemic sampling methods, along with various sample sizes, were compared by means of a sampling simulation program with actual field data for two rice diseases, leaf blast and tungro. Three severity levels of leaf blast and two incidence levels of tungro were used. Precision depended primarily on disease intensity, followed by the sample size and the sampling method. Relative accuracy did not prove to discriminate sampling methods adequately, but simulated absolute accuracy is able to identify biases of systematic sampling paths. The results emphasize the necessity of pilot sampling at various stages of epidemics. The usefulness of simulated sample sizes and sampling methods based on real data is also demonstrated. With this approach a more practical combination of sample size and method may be found for different levels of disease intensity using precision and absolute accuracy as criteria.     

Volatile metabolite profiling to detect and discriminate stem-end rot and anthracnose diseases of mango fruits

The volatile metabolites from the headspace gas of containerised mango ( Mangifera indica ) cv. Tommy Atkins fruits, surface wounded and inoculated with the two fungal anamorphic pathogens Colletotrichum gloeosporioides and Lasiodiplodia theobromae , or non-inoculated (controls), were profiled using a portable gas chromatograph/mass spectrometer to discriminate diseases of mango. Thirty-four compounds were detected relatively consistently among replicates. Several of these were disease/inoculation-discriminatory and were classified into three groups: (i) compounds unique to only one treatment; (ii) compounds common to two or more treatments, but not to all; and (iii) compounds common to all treatments, but varying in their abundance. Two compounds, 1-pentanol and ethyl boronate, were detected in L. theobromae- inoculated mangoes alone, while thujol was observed only in C. gloeosporioides- inoculated mangoes. Discriminant analysis models based on the abundance of significant mass ions and consistent compounds correctly classified diseases/inoculations in up to 100% of cases. The disease-discriminatory compounds and discriminant analysis models developed here have the potential to be used in the early detection of postharvest diseases of mango fruits after validation under commercial conditions.     

黑龙江省水稻品种资源对水稻细菌性褐斑病的抗性鉴定

选育和应用抗病品种是防治水稻细菌性病害最经济有效的措施。本研究采用苗期喷雾法和针刺法对黑龙江省42个水稻主栽品种进行抗性鉴定。结果表明,利用针刺法鉴定的各水稻品种中表现中抗以上的品种有14个,其中高抗品种2个,对接种水稻品种病斑长度进行差异显著性分析,42个水稻主栽品种间抗性存在显著差异;利用喷雾法鉴定各品种,病级表现1级的品种有7个,占16.7%。对数据进行比较分析,针刺法和喷雾法对抗病品种鉴定的结果基本一致。本研究鉴定的2个高抗稻种资源对于抗细菌性褐斑病生产实践提供重要价值。因此,加大对水稻品种资源的深入研究,对实现水稻细菌性褐斑病的可持续控制有着重要意义。     

Correction to: Droplet-vitrification cryotherapy and thermotherapy as efficient tools for the eradication of apple chlorotic leaf spot virus and apple stem grooving virus from virus-infected quince in vitro cultures

European Journal of Plant Pathology -     

Adaptability of machine learning methods and hydrological models to discharge simulations in datasparse glaciated watersheds

The accurate simulation and prediction of runoff in alpine glaciated watersheds is of increasing importance for the comprehensive management and utilization of water resources. In this study, long shortterm memory(LSTM), a state-of-the-art artificial neural network algorithm, is applied to simulate the daily discharge of two data-sparse glaciated watersheds in the Tianshan Mountains in Central Asia. Two other classic machine learning methods, namely extreme gradient boosting(XGBoost) and support vector regression(SVR), along with a distributed hydrological model(Soil and Water Assessment Tool(SWAT) and an extended SWAT model(SWAT_Glacier) are also employed for comparison. This paper aims to provide an efficient and reliable method for simulating discharge in glaciated alpine regions that have insufficient observed meteorological data. The two typical basins in this study are the main tributaries(the Kumaric and Toxkan rivers) of the Aksu River in the south Tianshan Mountains, which are dominated by snow and glacier meltwater and precipitation. Our comparative analysis indicates that simulations from the LSTM shows the best agreement with the observations. The performance metrics Nash-Sutcliffe efficiency coefficient(NS) and correlation coefficient(R~2) of LSTM are higher than 0.90 in both the training and testing periods in the Kumaric River Basin, and NS and R~2 are also higher than 0.70 in the Toxkan River Basin. Compared to classic machine learning algorithms, LSTM shows significant advantages over most evaluating indices. XGBoost also has high NS value in the training period, but is prone to overfitting the discharge. Compared with the widely used hydrological models, LSTM has advantages in predicting accuracy, despite having fewer data inputs. Moreover, LSTM only requires meteorological data rather than physical characteristics of underlying data. As an extension of SWAT, the SWAT_Glacier model shows good adaptability in discharge simulation, outperforming the original SWAT model, but at the cost of increasing the complexity of the model. Compared with the oftentimes complex semi-distributed physical hydrological models, the LSTM method not only eliminates the tedious calibration process of hydrological parameters, but also significantly reduces the calculation time and costs. Overall, LSTM shows immense promise in dealing with scarce meteorological data in glaciated catchments.     

BP神经网络和时间序列模型在包头市空气质量预测中的应用

运用BP神经网络法和时间序列法,对包头市的二氧化硫、二氧化氮和可吸入颗粒物的年份浓度值和月份浓度值进行预测。预测结果为:包头市2010年的二氧化硫的年浓度值为0.0578mg/m3,二氧化氮的年浓度值为0.0402mg/m3,可吸入颗粒物的年浓度值为0.1024mg/m3。2011年3月的二氧化硫的浓度值为0.0592mg/m3,二氧化氮的值为0.0372mg/m3,可吸入颗粒物的月浓度均值是0.1210mg/m3,由预测结果可以发现污染物浓度降低,空气质量呈好转趋势。     

Comparison of the reaction of tomato lines to infection by tomato yellow leaf curl begomovirus in Lebanon

Tomato yellow leaf curl begomovirus (TYLCV), transmitted by the whitefly Bemisia tabaci , is epidemic in Africa, the Middle East and South-East Asia. It is also reported in some European countries and the American continent. In Lebanon, it is the major limiting factor for summer and autumn production of tomato. Comparison of the nucleotide sequence in the intergenic region with other reported leaf curl viruses showed the Lebanese TYLCV isolate to be closely related to Egyptian, Israeli and Jamaican isolates (94–96% identity). However, it is not closely related to isolates from Sardinia, Spain and Thailand, or to tomato leaf curl isolates from India, Taiwan and Australia. In field and greenhouse screening tests conducted for 5 years on 67 tomato lines, several were identified as promising. TY-Carla, PSR and RS lines were among the most promising with determinate growth, while S&G 143 and the DR lines were the most promising with semi-determinate and indeterminate growth, respectively. Virus concentrations in most, but not all, tolerant tomato lines were significantly lower than in the susceptible lines. None of the lines tested was immune to the virus. A survey of TYLCV alternative hosts on at least 58 plant species, using nucleic acid hybridization, showed that Amaranthus sp., Malva sp., Sonchus oleraceus , Plantago sp., Solanum melongena , Phaseolus vulgaris and Mercurialis annua may play an important role in the epidemiology of TYLCV in Lebanon. Mercurialis annua is a newly reported host for TYLCV.     

Studies on the mechanisms responsible for variable toxicity of azinphosmethyl to various larval instars of the tufted apple budmoth, Platynota idaeusalis

The susceptibility of tufted apple budmoth larvae, Platynota idaeusalis, to azinphosmethyl decreases with each successive instar. A comparison between fifth and third instars showed that the fifth instars have a higher level of glutathione S-transferase activity per milligram of protein, a lower content of cytochrome P-450 per milligram of protein, and absorb a lesser percentage of and LD₀₁ dose than the third instar. Results of metabolism were consistent with these findings. In addition to these factors, the fifth instar larvae may have a threshold mechanism for eliminating penetrated azinphosmethyl from the body, unchanged. This allows the fifth instar larval population to withstand increasingly higher doses of azinphosmethyl without a proportional increase in mortality.     

Comparison of empirical models depicting density of Amaranthus retroflexus L. and relative leaf area as predictors of yield loss in maize (Zea mays L.)

The outcome of crop-weed competition should be predicted as early as possible in order to allow time for weed control measures. Maize grain yield losses caused by interference from Amaranthus retroflexus L. (redroot pigweed) were determined in 1991 and 1992. The performance of three empirical models of crop-weed competition were evaluated. Damage functions were calculated based on the weed density or relative leaf area of the weed. In the yield loss-weed density model, values of I (percentage yield loss at low weed density) were relatively stable for similar emergence dates of A. retroflexus across years and locations. Estimated maximum yield loss (A) was more variable between locations and may reflect environmental variation and its effect on crop-weed competition, at least in 1991. The two-parameter yield loss-relative leaf area model, based on m (maximum yield loss caused by weeds) and q (the relative damage coefficient) gave a better fit than the single-parameter version of the model (which includes only q). In both relative leaf area models, the values of q varied between years and locations. Attempts to stabilize the value of q by using the relative growth rate of the leaves of the crop and weed were successful; however, the practical application of such relative leaf area models may still be limited owing to the lack of a method to estimate leaf area index quickly and accurately.