Research Of The Early Rice Blast Analysis Based On PCA And SVM

Rice is one of the staple food crops in China with a wide growing area, covering about a quarter of the total farmland. The annual rice production also makes up almost half of the total national grain output. The rice blast is one of the serious diseases affecting the high and stable yields of rice. In recent decades, the loss of planting areas has been more than three million hectares and the decrease of grain has also come to tens of millions ton every year. How to accurately forecast the rice blast in time and take measures to prevent pest have already been the basic work to control the rice blast disease and secure the safety of rice efficiently.Based on weather conditions, this paper studies on forecasting methods of the early rice blast. It makes experiments in Tongling Pu Jiwei farm. According to the biological properties and meteorology principle of rice blast, this paper chooses the principal component analysis (PCA) to analyze the main meteorological factors affecting the occurrence of rice blast. Through this method, the paper selects and analyzes the main meteorological factors about the degree of rice blast. It finds that the rate of contributions of the previous four main components has already reached 81 percent, conforming to the standard of extracting main components. By observing the previous four main components, this paper discovers that the meteorological factors, such as amounts and days of precipitation, temperature, heliometric index, relative humidity and so on, play a vital role in the degree of the occurrence of early rice blast. Especially the amount of precipitation as the decisive meteorological factor directly influences the forming, dyeing and spreading of the bacteria of the early rice blast. Generally speaking, the long and continuous rainy days are the peak periods of the happening of early rice blast.This paper starts from extracting the main components of meteorological factors from environment by the method of PCA. Then referring to the levels of early rice blast in each year, this paper uses SVM to establish the forecasting model of the degree of the early rice blast. The author takes the data of Tongling Pu Jiwei farm as the training samples from 1991 to2002 and from 2005to 2006, and uses the LIB SVM tool box of MATLAB to forecast and analyze. After that, the author applies the actual data from 2003 to 2004 and from 2007 to 2010 to check the forecasting efficiency of the model, finding that the rate of accuracy as high as 83 percent. As a result, on the basis of PCA and SVM, the forecasting model aiming at the meteorological factors is of high feasibility and quite potential.This paper at last adopts the language C#、JQuery、Json and bases on the VS.Net platform to realize the pre-warning system of the degree of early rice blast. Users can input the meteorological environment index of certain past May and June into the main interface, and then the system can make a warning about the level of the early rice blast that year. The interface of this system is concise, clear, and easy to operate. It employs the SQL Server 2005 as data base processing which is very convenient for managing data and information. The exploration of the system is of vital importance not only for the efficient forecasting of the degree of early rice blast but also for improving national grain output, bring about more social, economic and ecological benefits.