Plant Disease Detection High-resolution Image Acquisition And Spectral Mixture Model-based Approach

Digitized noninvasive identification of plant diseases has become a trend-based diagnosis of crop disease,Aiming plant disease symptom complex,existing only through a single diagnostic techniques to identify the image contrast and low detection efficiency,utilization spectral imaging,spectroscopy,knowledge spectrum database technology and color science and many other areas,carried out on plant leaves major diseases fast,non-destructive testing methods research,and the establishment of a model for rapid detection and disease diagnosis system based on this disease.Through in-depth study of the spectrum of the cube structure,based on high-resolution spectral image format,proposes a universal especially suitable for high-resolution spectral image data fast processing of the data structure;in conjunction with SQL Server database,based on the high-resolution spectral imaging system,culture and other horticultural plants trees on one hundred samples were collected and analyzed to obtain a large amount of image and spectral data gardening diseases,crop diseases to find a variety of different concentration and time of onset of the spectrum law,and color differences provide a data basis.The results of this paper is to provide rapid diagnosis of plant diseases,new instruments,and computer technology,information technology and spectroscopy provides an example of application in agriculture has important theoretical and practical significance.Articles from subject background to start,on the basis of imaging spectroscopy,plant disease detection theory elaborated on,starting with the spectroscopic imaging cube data structure,this paper introduces the principle of spectral imaging technology and the common spectral imaging technology,and built based on the defect detection spectrum LCTF imaging and spectroscopic imaging hardware system;then discussed local and network data storage structure of spectral image data,and presents a model for database storage model and the localization of high-resolution spectral image data,spectral imaging system for LCTF data processing and related software design;and finally by comparing the principal component analysis,linear discriminant analysis of several common characteristics,neural networks and other methods to extract and summarize the various analyzes of its suitability for use in disease discriminant analysis We analyze,presents a stepwise discriminant model Fisher method of spectral dimensionality reduction,and the use of RBF neural network to classify the spectral dimensionality reduction capacity before and after the test based.