Inversion Of The Apple Tree Canopy Chlorophyll Contents In Hilly Region Based On Remote Sensing Data

Chlorophyll is the main carrier of photosynthesis,and it is an important index to detect the photosynthetic capacity and developmental status of of crops.The analysis of chlorophyll content by traditional laboratory stoichiometry methods is time-consuming,laborious,and unfavorable to monitor crop growth status in a large area.In recent years,remote sensing technology has the advantages of fast analysis speed,low cost and large area providing a new technical method for chlorophyll content determination.Therefore,it is of important theoretical and practical significance to use remote sensing technology to retrieve chlorophyll content in crops.In this study,Qixia City of Shandong Province known as the capital of apple,was taken as the study area.The chlorophyll content of apple tree canopy was retrieved by Sentinel-2A remote sensing images combined with the measured data in the near ground.The cosine correction and Minnaert model were used to correct the terrain radiation of the remote sensing image after atmospheric correction.On the basis of the mixed pixel decomposition combined with the near ground data,the apple tree canopy reflectance was inversed.From the previous vegetation indexs,the vegetation indexs were constructed with the visible light band,red edge band and near infrared band of Sentinel-2A.Based on the vegetation index,the chlorophyll content inversion models of apple trees' canopy were constructed,then the best inversion model was selected in contrast to the accuracy of the models.The main results were as follows:(1)Inversion and accuracy analysis of canopy reflectance of apple treesThe cosine correction and Minnaert model were used to correct the terrain radiation of the Sentinel-2A remote sensing image after atmospheric correction of the study area.The mean value and standard deviation of the image after the Minnaert model were smaller than those of the cosine corrected image,and the mean value of the image after the Minnaert model was close to the mean value of the atmospheric correction image.The Minnaert model could remove the shadow of the terrain,reduce the contrast ratio of the sunny and shade slope,eliminate the effect of terrain,and get the surface inversion reflectivity.Combined with the measured data in the near ground,the linear model was used to decompose the mixed pixels to obtain the reflectance of the apple tree canopy.The relative error between apparent reflectance,surface inversion reflectance,canopy inversion reflectance and canopy reflectance were gradually reduced by processing the images.The canopy inversion reflectance was closer to the real canopy reflectance.The relative error of band 2~8A was 14.4%,9.5%,10.1%,1.6%,0.4%,1.4% and 2%,respectively.It showed that more real canopy spectra had been obtained by various image processing,which provided the accuracy guarantee for the subsequent analysis.(2)The Canopy Chlorophyll Content and vegetation index of apple trees were constructed and screenedBy considering the spectral characteristics of green vegetation and the band characteristics of Sentinel-2A image,12 vegetation index were constructed using the blue band 2,green band 3,red band 4,red edge band 7,near infrared band 8 and near infrared band 8A of Sentinel-2A considering the construction principle and form of RVI,CI,NDVI.Through the correlation analysis between vegetation index and chlorophyll content,and the autocorrelation analysis of vegetation indexs,3 vegetation index series were selected.The three vegetation index series were series 1(RVIblue+RVIred+RVIre),series 2(CIblue+CIred+CIre)and series 3(NDVIgreen+NDVIred+NDVIre).(3)The chlorophyll content inversion model of apple tree canopy were established and validatedUsing vegetation index series 1,series 2 and series 3 as independent variables and the chlorophyll content of apple tree canopy as dependent variable,the BP neural network and support vector machine regression inversion model were established.The decision coefficients of modeling and testing of the BP neural network inversion model 3 based on NDVIgreen+NDVIred+NDVIre vegetation index series was larger than model 1 and model 2 of BP neural network,which were 0.674 and 0.601,respectively.The root mean square error of the BP neural network inversion model 3 was less than model 1 and model 2 of BP neural network,which were 0.169 and 0.185,respectively.The decision coefficients of modeling and testing of the support vector machine regression inversion model 3 based on NDVIgreen+NDVIred+NDVIre vegetation index series was larger than model 1 and model 2 of support vector machine regression,which were 0.729 and 0.667,respectively.The root mean square error of the support vector machine regression inversion model 3 was less than model 1 and model 2 of support vector machine regression,which were 0.159 and 0.178,respectively.In addition,the support vector machine regression model 3 was better than BP neural network inversion model 3.The support vector machine regression model 3 had the best effect and could be used to retrieve the canopy chlorophyll content of apple trees.It also showed the effectiveness of Sentinel-2A image in the canopy chlorophyll inversion.In summary,Sentinel-2A remote sensing image combined with the near ground data provides a new method for rapid monitoring and accurate diagnosis of chlorophyll content in apple tree canopy in hilly areas,which provides theoretical basis and technical support for the development of agricultural informatization.