Progress of Intelligent Monitoring Technology for Wheat Fusarium Head Blight

2021-11-11QixunSUN

Asian Agricultural Research 2021年3期

Qixun SUN

University of Washington, Bothell 98011, USA

Abstract Fusarium head blight is one of the most important diseases affecting wheat yield and quality. It is of great significance to carry out intelligent monitoring of wheat Fusarium head blight for high yield, high quality and sustainable development of wheat. On the basis of identifying the harms of wheat Fusarium head blight, this paper analyzed the monitoring technology of wheat Fusarium head blight based on satellite remote sensing, hyperspectral, near-infrared, Internet of things and photoelectric system, to provide a reference for the intelligent monitoring of wheat Fusarium head blight.

Key words Wheat, Fusarium head blight, Hazard, Intelligent monitoring

1 Introduction

Fusarium

head blight (FHB) is a worldwide disease caused by

Fusarium

asiaticum

and

graminearium

. It mainly occurs in warm and humid areas. The wheat producing areas in the middle and lower reaches of the Yangtze River and spring wheat areas in eastern Northeast China are the main epidemic areas of FHB. China is the country with the greatest wheat FHB damage in the world, with an annual affected area of 7.5 million ha, accounting for 1/4 of the total wheat area in the country. The annual yield loss caused by FHB is 2 million to 3 million t. The wheat areas in the middle and lower reaches of the Yangtze River are the most frequently affected areas of wheat FHB. In ordinary years, the FHB damage can cause a yield loss of 10%-15%, and in epidemic years, the yield loss can be up to 50% or even no harvest.

2 Hazard of wheat FHB

Wheat FHB is also called ear blight rot, bad wheat head, wheat ear rot or red wheat head. The disease can harm every growth stage of wheat and cause spike rot, stalk rot, seedling rot, stalk rot, among which spike rot brings the greatest harm and loss in China. It is an important basis for comprehensive control of wheat FHB to find out the yield loss of wheat FHB and work out reasonable control indexes. In the past, the method of selecting individual plants with different disease grades to measure the yield was used to study the damage loss of wheat scab, but there was a great difference between individual plants. In addition, it was difficult to find out the single plant of different grades with the same panicle size in the same field at the same time, and the single plant of different fields was difficult to compare with each other due to the difference of fertility density and variety. According to the study of Song Fengxian

., in general, when the disease index increased by l%, the yield loss rate increased by about 0.5%; when the head rate of disease increased by 1.3%, the rate of disease grain increased by about 0.5%. Xiang Zepan

.believed that with the global warming and the change of farming systems such as straw returning to the field and no-tillage cultivation, the occurrence area of wheat FHB in China rapidly expanded from the Yangtze River Basin to the northwest and north China. In recent years, there had been serious outbreaks of barley and wheat FHB in Hebei, Henan, Shandong, Shanxi and Shaanxi provinces and some other regions. Especially in 2008 and 2010, the scab was seriously prevalent in China, and the northern provinces also suffered serious losses.

2 Intelligent monitoring technology of wheat FHB

2.1 Satellite remote sensing monitoring technology

At present, there were few studies on the application of satellite remote sensing in wheat scab monitoring. Yin Wen

.combined HJ satellite remote sensing images with field experimental data and used different mathematical modeling methods to construct an estimation model for winter wheat FHB, aiming to explore a comprehensive meteorological factor and remote sensing factor estimation model for winter wheat scab, which could directly reflect the incidence of winter wheat scab from space. The results showed that there was an exponential correlation between winter wheat scab index and NDVI relative humidity and biomass, a linear correlation with LAI, and a power exponential correlation with temperature. The fitting results of the multiple linear regression model constructed were relatively ideal. Jin Zhengting

.used multiple linear regression between wheat growth parameters and climatic factors and the disease index of FHB to establish an estimation model of FHB. The model was tested with data outside the modeling, and a comparison diagram of 1∶1 between the predicted value and the measured value was generated. RMSE was 1.412. The model had high accuracy and could be applied to the estimation of winter wheat FHB in the heading to flowering stage. Li Weiguo

.established the remote sensing estimation model of winter wheat scab disease index based on interactions between spectral information and climatic factors, combining 5 sensitive factors

. NDVI, RVI, DVI, mean daily temperature of 5 d and average daily relative humidity of 5 d. The estimated value of the model was consistent with the measured value, root mean square error (RMSE) is 5.3%, and the estimation accuracy was 90.46%. It showed that the estimation model in this study could effectively estimate winter wheat FHB.

2.2 Hyperspectral monitoring technology

Hyperspectral data can be divided into imaging hyperspectral data and non-imaging hyperspectral data. The imaging hyperspectral technology effectively combines the spectral information representing the internal attributes of the target with the image information reflecting the external attributes of the target, which not only greatly improves the information richness, but also provides the possibility for more effective and reasonable analysis and processing of the spectral data in terms of theoretical research. Non-imaging hyperspectral data can be obtained by the portable ground object spectrometer (ASD), which mainly measures the spectral reflectance of the target in the band range of 350-2 500 nm. Compared with the imaging high spectral spectrometer, the number of bands obtained is more, and it is suitable for remote sensing monitoring of regional scale diseases and insect pests.

On the spike scale, imaging hyperspectral spikes were used to collect hyperspectral image data of wheat spikes of different disease severity, and the classification model of wheat FHB severity was constructed by combining spectral features and image features. On the canopy scale, a non-imaging ground feature spectrometer was used to collect healthy and diseased wheat canopy hyperspectral data as research objects. Using random frog jumping (RF), competitive adaptive weighted resampling (CARS), and Variable combination cluster analysis (VCPA) three-variable screening method. Based on particle swarm algorithm optimized support vector machine, different wheat canopy FHB disease surveillance models were constructed and compared.

In the aspect of wheat FHB symptom recognition, Bauriegel Elke

.used principal component analysis (PCA) and spectral Angle mapping (SAM) to model hyperspectral processing, and the diagnostic accuracy of FHB could reach 87%. Jin Xiu

.based on two typical structures of deep convolutional neural networks, convolutional neural networks with different depths were constructed to compare the training and testing results of hyperspectral data point sets of wheat FHB. The results showed that the deep neural network based on VGG could effectively extract the hyperspectral characteristics of wheat FHB. Jayme Garcia

.presented an algorithm for automatic detection of FHB in wheat kernels using HIS(hyperspectral imaging). The goal was to develop a simple and accurate algorithm which gave as output an index that can be interpreted as the likelihood of the kernel being infected by FHB. With the classification accuracy above 91%, the developed algorithm was robust to factors such as shape, orientation, shadowing and clustering of kernels. Zhang

. proposed a specific FHB classification index (FCI) for detection of this disease in wheat. The final FCI was FCI=0.25×[2(R668-R417)-R539], with an overall classification accuracy of 89.80%. The FCI was tested for its ability to detect and classify the healthy and diseased areas of wheat spikelets through comparison with six commonly used SVIs (simple spectral vegetation indices), and its disease identification accuracy is almost 30% higher than that of the best-performing SVI.In terms of wheat FHB grain identification, standard normal variable transform (SNV) and multiple scatter correction (MSC) methods were used for spectral data pretreatment，and continuous projection algorithm (CARS) and the positive adaptive weighted (SPA) algorithm were used to select wavelength. The results showed that the determination coefficients (R2) of MSC-SPA and SNV-SPA were 0.901 9 and 0.900 6, respectively, the root mean square errors were 0.223 8 and 0.223 2, respectively, and the numbers of selected wavelength were 7 and 5, respectively. Support vector machine (SVM) and BP neural network algorithms were used for modeling. The results showed that the accuracy of the four models were above 90%. Liu Shuang

.using hyperspectral imaging system combined with machine learning, proposed an algorithm for rapid visual recognition of a large number of wheat FHB grain samples. 400 healthy wheat samples and 400 infected wheat samples were collected from the position of the mask image, of which 75% were used for the modeling set and 25% for the test set. The classification model was established by cross validation method, linear discriminant analysis (LDA), k-nearest neighbor algorithm (KNN) and support vector machine (SVM), and the accuracy of the test set reached more than 90%.

2.3 Near-infrared monitoring technology

The near-infrared spectral characteristics of an organism are the essential reflection of its surface optical characteristics and intrinsic component chemical properties. A comparison between two VIS-NIR spectral based systems performed in laboratory vs. infield for the early detection of

Fusarium

head blight infection in two cultivars of durum wheat (Creso and Simeto) was carried out. Using the Euclidean distance matrix cladogram results for the laboratory, three models were used considering spectral data from GS70, GS71+GS73, GS75, while for in-field data from GS70+GS71 and GS73+GS75. In the laboratory good performance of classification (86%) was observed at GS71+GS73

., only 8-10 d after the infection. The in-field measurement showed a lower percentage of correct classification at the same growth stages. Guan Erqi

.based on near-infrared spectroscopy, grain samples of wheat varieties were analyzed, and stoichiometric methods such as Ward method cluster analysis and principal component analysis were adopted to construct SIMCA identification model for wheat grains infected with FHB and uninfected wheat grains. The results of model diagnosis and validation showed that the correct recognition rate of SIMCA model for wheat grains infected with FHB and uninfected wheat grains was 100%, and the recognition effect was good.Short-wave Infrared (SWIR) spectroscopy is similar to near-infrared spectroscopy, and it is also an important direction of modern spectroscopic analysis technique. Zhang Jian

.obtained spectral images of wheat scab samples by short-wave infrared spectrometer, and analyzed and processed the data. The experimental data showed that there were significant differences in the spectral line data between wheat seeds with scab disease and healthy wheat seeds in the spectral range of 1 350-1 600 nm, which proved that the application of short-wave infrared imaging spectroscopy technology in the detection of wheat scab has a certain degree of application.

2.4 Internet of Things monitoring technology

The wheat FHB monitoring based on the Internet of Things(IoT) is to monitor and recognize the symptoms of wheat FHB in the field by using image acquisition, data transmission, image recognition and other technologies, which is real-time and convenient. The study of Yuan Dongzhen

.showed that the monitoring and early warning system of wheat scab could issue early warning information one week before wheat flowering, with a prediction accuracy of 94.4%. The system operated stably and had a high degree of automation. The study of Huang Chong

.showed that the average accuracy of forecasting the diseased ears rate and the epidemic level of

Fusarium

head blight were 79.9% and 74.5%, respectively. The average forecast accuracy of the epidemic level in the Huanghuai wheat planted area was 84.3%, which was higher than that of 67.1% in the Yangtze River valley. The average forecast accuracy of the diseased ears rate was 86.8% in the Huanghuai wheat planted area, which was also higher than that of 73.0% in the Yangtze River valley. The model had a good short-term early warning effect on

Fusarium

head blight in the Huanghuai wheat planted area. The IoT monitoring technology of wheat FHB cannot be separated from the prediction model. Zhang Pingping

. established a prediction model of wheat FHB disease head rate in Guanzhong based on the density of husk-producing straw, and based on this, developed a forecaster and an automatic monitoring and early warning system for wheat FHB based on the IoT. Song Rui

.installed wheat scab predictors in Jiangsu, Shaanxi, Henan, Hubei, and Anhui, 18 counties (cities), and uncontrolled wheat fields were set in the surrounding areas to investigate the scab, and the accuracy of the automatic monitoring and early warning system for wheat scab was evaluated by comparing with the prediction results of the early warning software platform. The evaluation results showed that the prediction accuracy of the system reached 71.8% in 2018.

2.5 Other monitoring technologies

He Liuqin

., through the comparison of domestic and foreign technologies in the grain detection industry, found the superiority of the photoelectric detection system in the field of detection, and combined with the theoretical study of photoelectric detection, put forward the design scheme of the photoelectric detection system for wheat grain FHB. The prototype test results showed that the system was not only fast and efficient, but also had high stability and good real-time performance, which provided a reliable theoretical basis and technical support for the development and application of photoelectric detection technology in wheat FHB.

3 Conclusions

Once wheat FHB occurs, it will exert a great influence on the wheat yield and quality. In order to reduce the loss caused by scab to wheat production, effective monitoring and prevention should be carried out. With the continuous development of modern information technology, the intelligent monitoring technology is becoming more and more mature, and the monitoring method of wheat FHB will be gradually improved, and the accuracy will also be continuously improved, so as to ensure the high yield, high quality, ecology and sustainable development of wheat.

Asian Agricultural Research

2021年3期