Study On Intelligent Tobacco Fertilization Expert System In Shaanxi

Tobacco is an important kind of economic crops. Fertilization has a great impact not only on tobacco yield, but also on tobacco quality. Conventional fertilization systems, which mainly based on field experimentation and expert experience, always take a long time and have strong subjectivity. In order to establish a more scientific and effective fertilization recommendation system guiding tobacco production, soil fertility is evaluated accurately. After that, a tobacco fertilization model is constructed based on Least Squares Support Vector Machine. Finally, an intelligent tobacco fertilization recommendation system is developed by using GIS.The main contents are listed as follows:(1) Evaluation of tobacco soil fertility and fabrication of distribution map. Based on the previous study, the set of evaluation factors influencing the growth of tobacco is built according to some rules. The method of AHP- Fuzzy Comprehensive Evaluation is adopted to evaluate soil fertility of forty counties in Shaanxi Province, and the evaluation result is verified by referring to the list of tobacco plantation division with high quality in China. Based on this, the suitability level for tobacco plantation is graded under the guidance of experts and described on the distribution map about the soil fertility for tobacco plantation in Shaanxi Province by using GIS technology.(2) Construction of an intelligent tobacco fertilization recommendation model based on LS-SVM. By summarizing the defects of existing fertilization models and analyzing the advantages of LS-SVM, an intelligent fertilization recommendation model is constructed after describing its working principle in detail. By analyzing the features and combining the advantages of different types of kernel functions, Mix-kernel based on the Polynomial Function and Radial Basis Function is constructed as the kernel function depending on the method of constructing hybrid kernel by Smits. Considering the soil nutrient conditions in model training and prediction, the content of main soil factors and amount of Nitrogen, Phosphorus and Potassium are selected as model input, tobacco yield is selected as model output. The model is demonstrated by the field experimentation data and the actual data from tobacco farmers. Experimental results show that the mean accuracy of yield prediction is 87.43%, consistent with the practice. Especially, the small amount of data is no exception.(3) Design and Development of an intelligent tobacco fertilization recommendation system based on GIS. By building the database about soil fertility and fertilizer recommendation results, the distribution map about the suitability level for tobacco plantation in Shaanxi Province is associated with the intelligent tobacco fertilization recommendation model. On the basis of them, an intelligent tobacco fertilization recommendation system is designed and developed. The fertilization recommendation results are predict on the model and written to the database at the same time, is published, combining with the soil fertility conditions to the distribution map finally. A series of tests prove that the system can not only predict the amount of fertilizer scientifically and accurately, but provide the fertilization recommendation scheme for the local farmers quickly and easily.