Analysis Of High Spectral Characteristics Of Ramie Leaves And Identification Of Brown Spot Disease

Hyperspectral telemetry is called imaging spectrometry,and it's a lot of very narrow and continuous image data from the imaging spectrometers,and the current high spectral range is mainly applied to geology and mineral exploration,vegetation recognition,classification,atmospheric monitoring and so on.With the continuous development of hyperspectral technology,its application field will be more and more widely.Since the 1990 s,high spectral technology has gradually started to be applied to the nondestructive monitoring of agricultural products,and has gained stable and continuous development.At home and abroad,we have tried to use the high spectral technique to study agricultural products,crop physiological and biochemical parameter prediction,disease and insect identification,etc.,and have obtained good results.Ramie is an important economic crop in China,and has large planting area in China,and plays an important role in the fields of medicine and textiles.However,there is little research on the hyperspectral of ramie at home and abroad.In this paper,ramie leaves were used as the research object,and FieldSpec 3 portable ground object spectrometer and supporting blade grillers produced by ASD company were used to collect the high spectral data of ramie healthy leaves and brown spot leaves through field experiments.The high spectral response characteristics of ramie healthy leaves and leaf spots were compared from the original high spectral reflectance peak valley parameters and vegetation index.The characteristic variables were selected and extracted using the original high spectral reflectance peak valley parameters,vegetation index and principal component analysis(PCA).Respectively,the original high spectral peak value,vegetation index and principal component factors as characteristic parameter,using the Fisher linear discriminant method of ramie,leaf spot and healthy leaves of classification and recognition based on hyperspectral ramie leaf recognition model.Main work and research results of this paper:(1)study on the spectral characteristics of leaf spot and healthy leaves of ramie.It is found that in the original reflectance,the regions with obvious differences include: the green peak area and the 670-970 nm region,the healthy leaves of the region are significantly higher than the leaf of the brown spot,and the red valley and valley 1 are all located in the region.In the original spectral position parameters,valley 2 was the most different,with the mean value of the center of the leaf of the brown spot was 1176.7nm and the healthy leaf was 1190.5nm.(2)using the original high spectral peak valley value,vegetation index and principal component analysis,three methods were used to reduce the high spectral data of ramie leaves and extract the characteristic variables.In the method of the original high spectral peak valley,we've got the green peaks,red valley,valley 1,peak 1,the valley 2,the peak 2,the valley 3,the peak 3,the valley 4,the peak 4 and the peak of 10 of the high spectral peaks of the valley of the high spectral peaks of the valley of the high and the wavelength of the mountain,In the vegetation index method,the normalization,difference and ratio are selected as the characteristic variables.The principal component analysis method was carried out for the original high spectral peak valley value,and the principal component was extracted as the characteristic variable.In order to better explore the effect of the modeling of different principal components,6,11 and 15 principal components were used as the characteristic variables respectively.(3)to apply the above characteristic variable Fisher linear discriminant leaf recognition based on high spectral model is set up,the results show that if only considering the accuracy and to identify the best effect is based on the original high spectral peak valley value model,prediction set recognition rate is 94.38%;If the accuracy and number of variables are considered comprehensively,the model based on principal component analysis is better,and the accuracy rate of the model prediction set based on 11 principal components is 92.5%.The recognition model based on vegetation index has the worst effect.At present,the identification of ramie brown spot is mainly based on visual and experimental chemical detection,and there is little research on the identification and classification based on high spectral characteristics.In this study,a method based on high spectra for the identification of brown leaf spot in ramie leaves was proposed,which filled the gap in the current research field.