Detection Of Sclerotinia Stem Rot Of Oilseed Rape By Using Image-based Low-altitude Remote Sensing Technology

Oilseed rape is one of the most important oil crops in China,but its production and quality often decline sharply due to disease.So monitoring the oilseed rape disease timely and effectively and developping reasonable control measures is very significant for improving the yield and quality.The traditional method of disease detection is time-consuming,which can damage the integrity of the sample.In addition,it can only be carried out in the laboratory micro-detection and can not achieve a large area of field monitoring.Therefore,it is important to evaluate the disease in a large area by using Low-altitude remote sensing imaging technology,and to develop the preventive measures to prevent and control the spread of the disease.Oilseed rape as the research object in this study,Sclerotinia sclerotiorum was infected on it.the UAV model platform equipped with hyperspectral imager and thermal infrared imager were used for the detection of Sclerotinia sclerotiorum.The main research contents are as follows:(1)In this paper,the hyperspectral images of the canopy data were acquired and analyzed.The hyperspectral images were pretreated to obtain the spectral reflectance values of the whole plant.The canopy data was pretreated by MAS,SG,MSC,Detrending and SNV respectively.Then,the models of PLS-DA,SVM,ELM,KNN were built based on the full band,characteristic wavelength.It can be concluded that different pretreatment methods and model methods will be different in the analysis of Sclerotinia sclerotiorum.The results of the pretreatment method MSC and ELM model are the best,the classification accuracy reached at 100%.The feature wavelength was extracted by SPA,2nd Der and GA-PLS.Then SVM,PLS-DA,ELM and KNN models were used to analyze the characteristic wavelength.The results show that the model based on SPA selected feature wavelength is the best.(2)The vegetation index was calculated by spectral data and analyzed by correlation analysis and single factor analysis of variance.According to the results,the indexes such as DVI,TVI,RVSI,RDVI,CARI and OS AVI were correlated with the health and disease status.Then,the single vegetation index and the combined vegetation index were used to build the model.The results showed that the single vegetation index was not significant in the classification and the combined vegetation index can better classify healthy and oilseed rape samples inoculated with sclerotia.(3)The hyperspectral images of the leaf data was acquired and analyzed.The hyperspectral images were pretreated to obtain the spectral reflectance values of the surrounding area of leaf lesion.The leaf spectral data were pretreated by MAS,SG,MSC,Detrending and SNV respectively.Then,the models of PLS-DA,SVM,ELM,KNN were built based on the full band,characteristic wavelength.It can be concluded that different pretreatment methods and model methods will be different in the analysis of Brassicanapus disease.The results of the pretreatment method MSC and ELM model are the best,the classification accuracy reached at 100%.Feature extraction methods of SPA,2nd Der and GA-PLS were used to extract the characteristic wavelengths.Then,the SVM,PLS-DA,ELM and KNN models were established based on the characteristic wavelengths.The results show that the model based on SPA selected feature wavelength is the best.(4)The thermal infrared camera was used to analyze the samples,and the temperature information of canopy were obtained.The average temperature,maximum temperature difference(MTD)were used to predict the healthy and diseased samples.The temperature of healthy samples and the diseased samples was compared and analyzed by ANOVA.The results showed that the difference of the maximum temperature difference between healthy and diseased plants was obvious.The average temperature difference between healthy and diseased plants was obvious with the number of days.The single factor analysis of variance showed that the maximum temperature difference had significant difference(P<0.01)on the first day after oilseed rape infection.Furthermore,the physiological indexes of Brassica napus L.were analyzed with the number of days.The results showed that there was a significant correlation between photosynthetic rate,carbon dioxide concentration and transpiration rate and temperature.(5)The thermal infrared data of leaf scale were used to analyze the samples.The temperature information of healthy and diseased area was obtaineds in diseased leaves.Thermal infrared image can visually identify the process of disease infection and use the pixel value to determine the temperature difference between the health and diseased area.The maximum temperature,minimum temperature,average temperature and maximum temperature difference were used to predict the healthy and diseased plants.The results showed that the maximum temperature,the minimum temperature,the maximum temperature difference and the average temperature in the healthy and infected areas were significantly different,and the lesion area temperature was higher than the healthy area.The single factor analysis of variance showed that there was significant difference(P<0.01)at the maximum temperature difference on the first day,and the early identification of sclerotinia sclerotiorum could be realized.