Studies On Visual Rice Growth Models And Cultivation Expert System Of Computational Intelligence

Visual Rice Growth Models (VRGM) and rice expert system of cultivation management for high yield were established by synthesizing the results of "National Rice Project" and combining the cultivation knowledge, experience of experts, while the techniques of artificial neural network and fuzzy logic were employed to improve the rice growth models and the expert system. The main results are as follows.1. Visual Rice Growth Models (VRGM) system represented an effort to design crop modeling software with object-oriented paradigm and to organize classes in system with model-document-view architecture. Firstly, the crop-environment system was abstracted as many subsystems and physiological processes with object-oriented paradigm while many classes were established to simulate the behaviors of these subsystems and physiological processes. In VRGM, the document-view architecture in Visual C++ was extended as model-document-view architecture for agricultural integrated system. In the model-document-view architecture, document gets the final simulation results by manipulating the models classes and acts as a communicating intermediacy between model and view. Views display the simulation results and translate users' operation and data entering to document. The architecture separates the domain models, data management and user interface. Modelers can adds classes in architectures to extend the system without having to change system structure, which is useful for construction and maintenance of agricultural models.2. Rice expert system of cultivation management for high yield was constructed by object-oriented paradigm. All factors are described as knowledge type梡rototype storage. Based on Visual Rice Growth Models (VRGM) and the techniques of 'Vigorous root-strong culm-heavy panicle", "once-for-all basal application" etc, rice expert system of cultivation management for high yield has functions of inquiry, optimum decision on agronomy measures, cultivation management decision, diagnosis and control of physiological problems, disease and insect pests. In a word, it is characterized of knowledge and data treatment, logical inference and decision ability, high precision and reliability, user-friendly interface, and easy operation.3. Based on Visual Rice Growth Models (VRGM) and with rice tillering as examples, the back-propagation artificial neural networks (BP) and radial basis function networks (RBF) were established to simulate the rice growth and to compare with statistical model and dynamic model. The study shows that these models accurately simulate rice tillering. The model established by BP network has a good general ability and a slow impending speed. On the other hand, the model established by RBF neural network has a bad general ability and a fast impending speed. The models established by artificial neural network have the similar simulation capability with statistical model and dynamic model. The artificial neural network provides convenient and fast tools to simulate rice growth and gets ride of the limitation of traditional models.4. Fuzzy discrimination system of photothermal traits was developed with fuzzy logic theory and MATLAB 5.1. The acquisition, expression and application of fuzzy knowledge were fulfilled through fuzzification of photothermal traits, establishment of fuzzy logic rules, fuzzy language and inference etc. The goodness of fit to practical situation is up to 92.4%.