Dissertation > Excellent graduate degree dissertation topics show

Study on Recognition Method of Rice Disease Based on Image

Author: GuanZeZuo
Tutor: YaoQing
School: Zhejiang University of Technology
Course: Applied Computer Technology
Keywords: Rice Diseases Computer Vision Image processing Feature Extraction Pattern Recognition Remote Indoor lesion evaluation
CLC: TP391.41
Type: Master's thesis
Year: 2010
Downloads: 155
Quote: 0
Read: Download Dissertation

Abstract

Rice is one of the most important food crops in China, an important goal to improve rice yield and quality of rice production today. However, the rice each year due to losses caused by the pest are quite amazing rice disease control in rice production and the development of the national economy occupies an extremely important position. Of agricultural rice disease identification still remain in the subjective judgment combined with the existing experience of of artificial visual inspection and comparison stage, objectivity, inefficient, labor-intensive and other issues, the paper uses computer image processing technology combined with pattern recognition methods the rice disease identification method. Build a rice disease database, and the development of a rice disease intelligent recognition system, realize The rice lesion segmentation, lesion eigenvalue extraction, automatic identification of the Rice indoor the lesion evaluation as well as rice disease; combined with network technology to develop the remote identification of rice diseases diagnostic system. This study conform to the requirements of the development of modern precision agriculture and modern agriculture automated detection diagnostic technology development direction, has a good prospect, is also a good idea for the follow-up study, its far-reaching. Main contents and results include: (1) the establishment of a rice disease database, rice disease picture, disease characteristics, and other descriptive information storage reappearance network for the system to provide a support system for identifying and, also for the future based on the content of to retrieve the data base; (2) in the image segmentation, a combination of the color characteristic lesions and spots outer contour extraction method. By comparison and analysis, the paper presents the segmentation method with the traditional method based on morphological features based on color features comparison, can effectively avoid the empty spot areas in the segmentation process, has strong anti-noise; ( 3) In the feature parameter extraction, the use of non-uniform quantization of the histogram extraction of the color characteristics, to improve the recognition robustness. GLCM texture feature extraction, compression of the gray scale level of the lesion image, reducing the 3/4 of the amount of calculation and reduce the complexity feature extraction; (4) in the characteristic parameter optimization, using the method of stepwise discriminant the set of parameters to optimize screening analyzed chrominance texture parameters identifying the relevant larger color parameters to identify the lowest correlation; and does not affect the basis of the recognition rate can make up parameter is reduced to 57.2% of the original, effectively removed correlation redundant parameters to reduce the burden of computer storage and computing to improve recognition rate of the disease; (5) in the area of ??disease recognition, namely the use of Bayesian neural network and support vector machine classifier 6 kinds of rice Diseases in four different sets of parameters for identifying and sorting. Results show that the use of these three classifiers morphological parameters set the lesion recognition rate less the color set of parameters, for texture parameter sets, neural network classifier showed weak generalization ability and stability weaknesses and correct rate less than 50%, but up to 98% recognition rate of the set of three parameters, the three classifiers. Support vector machine to solve the small sample, nonlinear and high dimensional pattern recognition problems in the performance of many unique advantages, and the algorithm fixed to avoid the output does not converge with the randomness of a larger system uses extracted disease three collection of species characteristic parameters, support vector machine as a classifier to identify rice disease; (6) embedded in a Web client using ActiveX development, combined Sockets socket implementation based on B / S structure of the rice disease remote identify, browse and query system, breaking the geographical restrictions diseases to identify real-time, the popularity of easy to advanced technology and promotion, strengthen technological exchanges between the technical departments and users; (7) in the rice-based image indoor lesion evaluation aspects, the secondary segmentation strategy for dividing the lesion healthy leaves, to improve the accuracy; raised rectangular box to remove sclerotia, operability and does not affect the segmentation accuracy; (8) the establishment of a rice diseases of intelligent recognition system, to achieve rice disease image acquisition, lesion segmentation, feature extraction, evaluation of indoor lesion, disease recognition and remote query browsing functionality. The research method can be extended to other crop diseases identification.

Related Dissertations

CLC: > Industrial Technology > Automation technology,computer technology > Computing technology,computer technology > Computer applications > Information processing (information processing) > Pattern Recognition and devices > Image recognition device
© 2012 www.DissertationTopic.Net Mobile