Remote Sensing Monitoring Of Wheat Diseases And Pests Based On Improved Least Squares Support Vector Machine

Crop diseases and insect pests are one of the main factors affecting agricultural production.China,as a large agricultural country,has many kinds of crop diseases and insect pests,which has caused a huge loss of grain production.Accurate and timely monitoring of the occurrence of crop diseases and insect pests at a regional scale can effectively guide the prevention and control work and reduce the impact of disease.It has become a hot research topic to use remote sensing technology to extract information of crop diseases and insect pests and monitor crop diseases and insect pests at the regional scale.However,how to select the appropriate and effective methods and maximize the effective information in remote sensing image data are the main problem for the researchers.The common diseases and insect pests of wheat:wheat powdery mildew and wheat aphids are studied,moreover,the mainline of this thesis is monitoring of wheat powdery mildew and wheat aphids at a regional scale.Using Landsat-8 remote sensing image data and Chinese Environment and Disaster Monitoring Small Satellite remote sensing image data to carry out the monitoring model and method of wheat diseases and pests,the specific research contents are as follows:(1)The least squares support vector machine model which optimized by the particle swarm optimization algorithm(PSO-LSSVM)is proposed to monitor wheat powdery mildew.Thesis study on wheat powdery mildew in parts of Guanzhong Plain,Shanxi Province in 2014.By using Landsat-8 satellite OLI and TIRS data,thesis extracts wheat growth factors and habitat factors of affecting wheat powdery mildew,including normalized difference vegetation index(NDVI),ratio vegetation index(RVI),greenness,wetness and land surface temperature(LST),the least square support vector machine(LSSVM)algorithm is used to monitor the powdery mildew.And the parameters of LSSVM model are optimized by particle swarm optimization(PSO)algorithm.The monitoring results are compared with the traditional LSSVM algorithm and support vector machine(SVM)algorithm.The results showed that:The overall monitoring accuracy of the least squares support vector machine model which optimized by the particle swarm optimization algorithm(PSO-LSSVM)is 92.8%,have better performance over the traditional LSSVM model(85.7%)and the SVM model(71.4%).The proposed method obtains good monitoring results.(2)The least square twin support vector machine(LSTSVM)is proposed to monitor wheat aphids.Thesis study on wheat aphids occurred in Tongzhou and Shunyi district of Beijing in 2010.By using the data from HJ-CCD and HJ-IRS,the growth factors and the environmental factors of wheat are extracted during grain filling period.Through the independent t-test method combined with the ground survey data to extract the feature factor screening,the characteristic factor of 0.999 confidence level is selected,including Normalized Difference Vegetation Index(NDVI),Green Normalized Difference Vegetation index(GNDVI),Reflectance of Red Band,Land Surface Temperature(LST)and Perpendicular Drought Index(PDI).The monitoring model of wheat aphids is established by using the least square twin support vector machine(LSTSVM).Compared with the traditional support vector machine(SVM),fisher linear discriminant analysis(FLDA)and LVQ neural network,the LSTSVM monitoring result show that:the overall monitoring accuracy of the least square twin support vector machine model is 86.4%and the kappa coefficient is 0.71,which is better performed than the traditional SVM model(77.3%,0.52),FLDA model(77.3%,0.54)and LVQ neural network model(72.7%,0.39).