Studies And Implementation Of Expert System For Control Of Disease And Pest Of Common Vegetables In Shaan Xi Province

The protection against diseases and pests is particularly important to improve crop yield and quality. The traditional way based on human observation and experiences to determinate the occurrence of a disease or a pest cannot meet the requirements of modern agricultural. This paper aims at the study of the intelligent diagnosis method, based on Back Propagate (BP) neural networks and expert systems. At the same time, the prototype of expert system for ecological control of common vegetables’high occurring diseases and pests such as Eggplant cucumber in Shaanxi Province, has been developed in order to assist agricultural workers.The main research contents of this dissertation are the followings:(1) Building a knowledge base. The construction of knowledge base has huge impact on the whole system. Through communication with experts and local farmers, we determine the requirements of users. According to the variant contents of disease and pest knowledge, the focus of the research consists of knowledge acquisition, selection, classification and display.(2) Establishing the diagnosis model. There is the possibility of failure expert system rule matching sheer,due to the incompleteness of the knowledge base.Aiming at this problem, the paper uses the expert system and the artificial intelligent (diagnosis model based on BP neural network) to establish vegetable diseases and insect pest diagnosis model.(3) Improving the traditional BP neural network algorithm. Since the traditional neural network has many weaknesses, such as high time complexity, low diagnostic accuracy, slow convergence and apparent concussion when dealing with high dimension data, two improvements have been proposed:one consists of selecting out the high discrimitive features for disease classification another is by modifying the learning rate dynamically according to the network’s output deviation. These improvements have been applied to build a disease detection model. The comparative test of two algorithms before and after improvement shows that after the improvement the detection accuracy is15%higher, the standard derivation respectively7.48%lower; Furthermore, the improved algorithm is superior to the traditional BP neural network in many aspects, such as, statistical indicators, convergence speed and stability. Thus, the reliability and efficiency of the proposed method are verified. (4) Implementation of the prototype of expert system for ecological control of common vegetables’ high occurring diseases and pests. The prototype expert system is accessible on the WEB. It offers the functions of common vegetables diseases diagnosis, display of images and texts on disease and pest, recommendation of prevention and protection solutions as well as online system helps.