A Research About Wheat Stripe Rust Monitoring Model Based On Multi Source Information

One of the main factors affecting grain yield and quality are crop diseases and insect pests.Wheat stripe rust(Puccinia striiformis West.f.sp.tritici Eriks et Henn)is a kind of serious long-term of biological disaster which effects production safety of wheat in China,can repeat during the growing process of wheat and its spore propagated long distance with wind load.The incidence of wheat stripe rust is wide,strong,and has a high incidence,resulting in a substantial reduction in wheat yield and quality.It can effectively reduce the influence of disease on wheat yield and quality by prediction and real-time monitoring.At present,the remote sensing technology has the characteristics of non-contact and high throughput,which provides a scientific and effective way for the rapid,efficient,economic and non-destructive monitoring of wheat stripe rust disease.In this paper,on the basis of previous studies,combined with the existing wheat stripe rust occurrence characteristics,and the basic ideas of monitoring and early warning,combined with remote sensing technology,statistics,machine learning algorithm for multi-level,multi angle of wheat stripe rust disease precise monitoring,the main contents around of three aspects:(1)The canopy spectral reflectance of wheat infected stripe rust was measured and the disease indices(DI)were investigated in the field respectively during different growth stages.Smoothing the canopy spectra and continuum removal,extracting spectral absorption feature parameters such as absorption depth(D),absorption peak overall area(A),absorption peak at the left end of the area(A1)and symmetry(S),and bring the spectral sensitivity(spectral sensitivity)concept,comparison of different of wheat growth situation.Analyzing and comparing the prediction spectral absorption feature parameters and spectral sensitivity for estimated the whole wheat fertility period of disease index.The results indicated that,there are not much differences of canopy spectral reflectance and spectral absorption feature parameters A,D,S and A1 between health and disease wheat in the flagging stage because of the mild rust symptoms.Wheat had been infected with stripe of rust serious at early grain filling stage and grain filling stage as a result of canopy spectral curve with typical sensitivity curve feature.Theregression models were established by using the spectral sensitivity and the absorption characteristic parameters as independent variables,and the disease index of the whole growing period of wheat was accurately retrieved.Spectral sensitivity of each stage as variable has the better estimation precision for DI than the characteristic parameters of the spectrum absorption.The model root mean square error is 0.051,and relative error is 0.032,predictive fitting degree is 0.97.(2)We analysed the wheat stripe rust disease protection survey data and wheat growth period average temperature,average humidity,precipitation,sunshine duration and average wind speed and daily meteorological data,at the five Southeast cities of Gansu,as Longnan,Tianshui,Dingxi Pingliang and Qingyang from 2010 to 2012.We obtained that stripe rust disease state and incidence area change with the climate characteristics during wheat growth periods.And we analyzed the connection between the wind direction and speed of epidemic diseases spread through the calculation of air volume around the impact value in Gansu area.Based on the geographical spatial analysis technology and mathematical statistics,combined with the adaptive fuzzy logic vector method,accurately predicted the disease occurrence state of the study area.The results showed that there was a significant relationship between the diffusion direction of disease and the direction of wind speed and direction in the southeast of Gansu.There was a positive correlation between the temperature and the disease condition from the jointing stage to the ripe stage.(3)In order to explore the Gaofen satellite multi spectral data of wheat stripe rust monitoring precision,with GF-1 satellite WFV sensor band response function based on simulated measured canopy hyperspectral data for GF-1 multi spectral data,building NDVI,RVI,TVI,DVI,RDVI and other vegetation index based on partial least squares,Akaike information criterion and grey correlation algorithm from the selection of the best wheat disease sensitivity index of vegetation index,combined with the meteorological data,an adaptive fuzzy logic inference model of wheat stripe rust disease occurrence inversion.On this basis,using the wheat disease index obtained in the field to carry out the inversion precision verification.The results show that the DVI and the occurrence of wheat disease sensitive,combined with the average temperature of farmland,the use of adaptive fuzzy inference model for inversion of wheat stripe rust disease index to obtain the approximate accuracy and predictive value of fitting degree of R2 reached 0.99 and the measured value is 0.0012 RMSE.Therefore,the use of GF-1 multi spectral satellite data,combined with meteorological data to achieve accurate monitoring of wheat stripe rust disease,and laid a solid foundation for the fine management of farmland.