Research On Methods Of Diagnosing Wheat Drought Status Based On Image Processing

Based on the analysis of the background of wheat phenotype research both at home and abroad,this paper combs and draws lessons from previous research ideas and achievements.Combining with the purpose and requirements of this experiment,we have developed a detailed research program to manage the wheat drought phenotypic characteristics and to establish a set of models in order to identify and characterize the drought phenotypic characteristics of wheat and to use the method of digital image processing to judge the drought status of wheat.Finally,three purposes are realized: 1.Quantitative method of wheat drought phenotypic index;2.Image analysis and phenotypic eigenvalue calculation;3.Establishment of wheat drought phenotypic classification model.In this paper,three kinds of potted wheat with different irrigation amount were cultivated in the greenhouse,and the 18-megapixel SLR camera with Canon EOS 700 D was used to carry out three-dimensional shooting from the jointing stage to the heading date without damage.The image data were obtained and processed.This includes: a.Gray processing,b.Image enhancement,c.Image segmentation,d.Morphological processing,e.Wheat color assignment,and then use Matlab to extract 34 eigenvalues,including color eigenvalues,texture eigenvalues and the energy eigenvalues,they are as follows: R(red),G(green),B(blue)and their linear combinations: R / G,R / B,R /(G + B),G / R,G / B,G /(R + B),B / R,B / G,B /(R + G),2G-R,2G-B,2G-RB,and H(hue)I A(hue)I(saturation),S(brightness)and H(color),S(purity),V(lightness).The energy eigenvalues include energy R、energy G、energy B、energy H、energy S、energy I、energy H、energy S、energy V.The texture eigenvalues include the contrast,uniformity,energy,and correlation in the gray covariance matrixEventually,I selected 18 effective eigenvalues of the three treatments(appropriate,semi-drought,excessive drought).They are as follows: R,G,B,I,V,R / B,2G-R,2G-B,2G-RB,energy R,energy G,energy B,energy I,energy V,contrast,energy value,uniformity,correlation.Then the trend of these 18 eigenvalues and the relationship between the eigenvalues and the control group under different water stress were obtained.Furthermore,the time slot suitable for studying the drought status of wheat is 7: 30-10:30,15: 30-16: 30,according to the time trend.By utilizing the training model sample to establish the prediction model and taking use of SVM,I get two SVM functions,and then the SVM is used to train the SVM parameters.Finally,the vector training parameters are verified by the test training samples,and the accuracy rate is 90%,which can be employed to determine the drought of the wheat state efficiently and accurately,without any sort of destruction.