Study On Spectral Reflection Characteristics Of Chestnut Leaves Under Stress Of Red Mite

Chestnut is an important agricultural product and economic crop,which has been listed in the national medium and long term strategic grain reserve plan.However,the diseases and insect pests of chestnut trees,especially the red mite pests,have seriously affected the yield and quality of chestnut.The traditional methods of monitoring and controlling red mite diseases,are not only high cost and low efficiency,but also can cause the problem of pesticide abuse.Hyperspectral remote sensing has the characteristics of non-destructive,fast,accurate and wide range,and has significant advantages in the monitoring of crop diseases and insect pests in a large area,and has become an important research hotspot in the field of remote sensing.However,in order to use hyperspectral technology to monitor and forecast the diseases and insect pests of Chinese chestnut red mite,first of all,it is necessary to determine the law of the influence of red mite diseases and insect pests on the spectral characteristics of chestnut leaves.This aspect is still rarely reported.Therefore,this dissertation takes the leaf of chestnut tree as the research object,collects the leaf samples of different time and class of insect pests in the field,determines the spectral curve of the samples,extracts the spectral characteristics,analyzes the corresponding relationship between the class of insect pests and the spectral characteristics,explores the diagnostic spectral characteristics of the response of chestnut leaves to red mite pests and diseases,and establishes a spectral feature set which can effectively identify red mite pests and diseases,which lays a solid foundation for hyperspectral monitoring of large area chestnut forest red mite pests and diseases.The main work of this dissertation is as follows:(1)Field data collection of chestnut leaves and others subsidiary informationsIn this dissertation,20 chestnut trees were selected as experimental samples in the northwest mountainous area of Xingtai County,Hebei Province.From mid-April to the end of August 2018,field samples collection were carried out 18 times,and 3436 blades were collected as alternative samples.For the subsequent spatial analysis,we have collected the aerial photos based on UAVs in research area and built the DEM and DOM to determine the spatial location of each sample plant and the topographic characteristics of the experimental area.(2)Establishment of spectral curve sample set of chestnut leaves under the stress of red miteThe leaves collected in the field were selected and sorted to form leaf samples.Under the illumination of standard light source in dark room,the hyperspectral image and spectral curve of leaf samples were collected by UHD185 hyperspectral camera and ASD non-imaging spectrometer respectively.After smoothing,denoising and so on,the spectral curve sample set of chestnut leaves under red mite stress was formed by building database with field acquisition information and RGB image of leaf samples.The sample set has a hyperspectral image of 2218 chestnut leaves and 1180 spectral curves,which provides a complete basic data for the study of hyperspectral monitoring of pests and diseases of chestnut red mite.(3)Classification of the severity of chestnut leaf stress by red miteBy studying and analyzing the classification methods and standards of pest and disease severity proposed from different scholars and different industries,combined with the characteristics of this dissertation to study the spectral characteristics of leaf grade,a classification standard of pest and disease severity of Chinese chestnut leaf red mite was put forward,and the classification index of pest grade was defined with the ratio of spot area to leaf area;Then,by simplifying the maximum likelihood method into the second class problem,the index value of each leaf sample is extracted efficiently and accurately,so as to realize the automatic classification of each leaf pest and disease grade;Finally,the classification results are associated with the previous spectral sample set to make the spectral set more perfect.(4)Study on the response law of spectral characteristic to red mite pestThe stability analysis of the existing four spectral characteristics,such as green peak,red valley,red edge and high shoulder,shows that the four characteristic points have nonunique phenomenon in different degrees,and the stability from high to low is green peak,red edge,red valley and high shoulder,which causes some trouble to analyze the relationship between spectral characteristics and red mite diseases and insect pests.Therefore,based on the analyzing on variation law of the first derivative of the spectral curve in the sample set,the dissertation puts forward two new spectral features of low variable speed point(low position)and high variable speed point(high position),and while its definition and extraction method has been put.Experiments show that the low position has excellent uniqueness and the stability is the best among the six spectral features.Pearson and Spearman correlation tests were conducted on the relationship between the six characteristics and their combinations and the severity of the harm of red mite.The results showed that red edge and low position had strong negative correlation with red mite pest,and the correlation coefficient reached 0.909 and 0.870,respectively;Red valley and green peak have strong positive correlation with pest severity,the correlation coefficient is 0.767 and 0.648;The correlation coefficients of high position and high shoulder were only 0.498 and 0.008.It can be seen that the most sensitive spectral characteristics of red mite diseases and insect pests are red edge,low position,red valley and green peak in turn.If its stability is considered,the best spectral characteristics of identifying red mite diseases and insect pests are low position,red edge,green peak and red valley in turn.(5)Analysis of temporal and spatial changes of annual cycle of red mite diseases and insect pestsBased on the analyzing on the time sequence and spatial distribution of red mite pests in the research area.In sequential,it was found that red mite appeared from the beginning of May and it spread rapidly around a week,and became more serious with time,but Rain Water can effectively kill red mite and enter the next development cycle after rain;In spatial distribution,with the altitude from low to high,red mite’s damage severity has the law of change from light to heavy.92 graphs,30 tables,119 references.