Research On Construction Method Of Tomato Disease Image Database

Because of the short harvested period and the considerable revenue, tomato is one of the most widely grown vegetables in the open air in China. Nowadays we are extraordinary aim to enhance the output and the quality. However,insect pets can cause amazed damage to the tomato annually so the prevention and cure of insect disease is plays an decisive role to not only the production of tomato but also the national economic development while all kinds of insects disasters, which occur through the whole growing process of tomato, can engender the decrease or even a total loss. Subjective-experiential judgment is the major way to diagnose the disease by visual. This inefficiently manual labor is objective and wasted energy. This thesis is written to discern the insects disease of tomato by computer image processing technology combine to image similarity retrieval method to split image of tomato-scab,optimize the scab-eigenvalue extraction and discern the tomato-disease image intelligently. We can use computer technology to build the database system of tomato-disease image and then develop a further program. What we are studying accords with the precision of modern agriculture development and the development direction of intelligence technology of tomato disease diagnosis, which has a promising application prospect in vegetable disease control and create views to further study of similarity comparison of image database.The major research contents are as follows:(1) On the base of the previous research, construct the main diseases of tomato(scab,Botrytis cinerea, grey leaf spot, late blight, early blight) image database, the database structure for the relational data structure, content consists mainly of tomato five common leaf disease of shape features, color features and texture features image information data.The database realizes the image disease characteristics of tomato diseases and other description information storage process, to provide support for the identification system,and to provide data base for image based retrieval.(2) In terms of characteristic parameters extraction and optimization, using median filter in addition to the noise, using HIS space of H component disease spot segmentation,which reduces the complexity of feature extraction. In the process of feature parameters optimization, by using principal component analysis of nuclear optimize the different parameters, it is concluded that analysis the color parameter identification of correlation is larger, the texture parameters identification minimum correlation. Eventually extracted from 28 characteristic parameters optimization into 16 characteristic parameters used to construct the basic information database tables, using parameters decreased by 43%,effectively eliminate correlation of smaller redundancy parameter, improve the efficiency of recognition algorithm.(3) In terms of disease identification of the four types of database query algorithm is designed, the development of the four algorithms compared the single feature vector and combined with multiple characteristic vector of the query results. It is concluded that the three kinds of characteristic parameters of plant disease image collection is more efficient and accurate, and the average recognition accuracy reaches 80%. The system is adopted to extract three characteristic parameters of vector set of query algorithm, to implement the intelligent of the tomato plant disease image recognition.(4) Tomato disease intelligent recognition program is developed to implement the tomato disease disease spot image segmentation, feature extraction, feature parameter optimization, the disease identification, etc. And provides reference for other crops diseases identification.