Grape Diseases Intelligent Diagnosis Models And System

By analysing the domain knowledge of grape diseases, two kinds of diagnosis models which were based on the BP network and the rule similarity were established.Developing the grape diseases intelligent diagnosis systerm based on web and the diagnosis algorithms test system based on java is in order to diagnose exactly the grape disease and control effectively the incidence of diseases,which is very significant for farmers to guarantee the stabilization of grape production and increase their income.The main contributions in this dissertation are the followings:⑴By analysing diagnosis problems and discussing the diagnosis process of diseases with the expert,the expert's thought pattern was simulated.A research was organized with grape diseases focusing on pathogen, classification and symptom. Diseases diagnosis parameters were confirmed by the expert's helpness.⑵According to the different contents or modalities,the domain knowledge of grape diseases was divided into the description, the fact and the verdict. The methods of corresponding knowledge expression and memory were studied.Going on the relational database theory, the notional and thelogical structures of repository about grape diseases diagnosis were disigned.⑶On the basis of analysing the domain knowledge of grape diseases,by using the methods of the affixation monmentum and the self-adaptation learn rate to improve on the nomal BP algorithm, the diagnosis model based on the BP network was builted.Diagnosis parameters were viewed as input, and diseases were thought as output, using training samples which were selected from the diagnosis rules to train the network.The result of experiment indicated that the average diagnosis veracity of the model was 86.80%.In allusion to the chatacter of diagnosis parameters, a dynamic code method was put forward in order to realize the convert from the fact knowledge to the digital information.⑷The diagnosis model was established, which was based on the rule similarity.The diagnosis space was formed on the ground of parts and periods of diseases.Appling statistics and rough set theories to confirm the symptom weights of rules in the diagnosis space, and symptom weights of the diagnosis problem were also confirmed by average allocating. Computing similarity between diagnosis problem and rules in the diagnosis space, the results of experiment indicated that the mothod of confirming weights based on statistics theory was better than that based on rough set theory and it can reflect the similar extent between the fact and existed rules well.Using the same samples to test two kinds of diagnosis models, the result indicated that the average diagnosis veracity of the BP network model was higher than that of the rule similarity model. Therefore, the BP network model had the better effects on diagnosing.⑸By analysing the domain knowledge and studying diagnosis models, the grape diseases intelligent diagnosis systerm based on web and the diagnosis algorithms test system were developed.The algorithms test systerm may update diagnosis knowledge about grape diseases and show the process of analysing problems in two kinds of diagnosis models.The intelligent diagnosis systerm made users refer the detailed information about grape diseases and diagnose diseases so that people can take the available meassure to control diseases.