Study On The Estimation Model Of Apple Biochemical Parameters Based On Hyperspectral Remote Sensing

Shaanxi Province is located in the dominant apple production area of the Loess Plateau,with apple planting area and output ranking in the forefront of the country.It is a local characteristic industry to promote economic development and increase farmers'income.The development of apple industry is of great significance to optimize and adjust the regional industrial structure,promote the development of regional characteristic industries,increase the employment of farmers and realize the increase of local farmers'income.The Hyperspectral Estimation Method of biochemical parameters of apple leaves was discussed,which provided theoretical basis for field management and fertilization of fruit trees.Taking the standardized apple orchard in Zhaozhai village of Shaanxi Province as the research object,the original hyperspectral data,chlorophyll content(SPAD),nitrogen balance index(NBI)and water content of apple leaves at the end of flowering,young fruit,expanding fruit,fruit coloring and fruit ripening stages were measured,and the first-order differential spectrum of each growth stage was obtained,and the changes of biochemical parameters at different growth stages were analyzed Hyperspectral characteristics,and the change rule of hyperspectral under different biochemical parameters.On this basis,a variety of modeling methods were selected to build the estimation model of chlorophyll,nitrogen and water content in different growth stages of apple leaves,and the accuracy of the model was verified.The main conclusions are as follows:(1)In the Hyperspectral Estimation Model of apple leaf biochemical parameters,the best estimation accuracy and fitting degree is chlorophyll content,followed by nitrogen balance index,and the worst is water content.On the whole,the estimation accuracy and fitting degree of fruit maturity,Fruit Coloration and fruit expansion are better than that of final flowering and young fruit.Among them,the fruit coloring period is a sensitive period for the estimation and monitoring of Apple biochemical parameters.(2)The sensitive bands of chlorophyll content and nitrogen balance index in apple leaves were mainly around 700 nm in the red band,followed by 550 nm in the green band.The sensitive spectral bands of water content in apple leaves were mainly concentrated at 1250?1650 nm,1800?1900 nm and 2000?2100 nm in visible and near infrared bands.Compared with the original spectrum,the first-order differential spectrum can slightly improve the correlation between the sensitive bands of biochemical parameters,including the green band of chlorophyll content,nitrogen balance index and the near-infrared band of water content.The first-order differential estimation model is superior to the original spectral estimation model in accuracy and fit.(3)The sensitive vegetation index of apple leaf chlorophyll content is green ratio vegetation index(GRVI),MERIS ground chlorophyll index(MTCI),improved chlorophyll absorption reflectance index(MCARI).The sensitive"three side"parameters are red yellow normalized index,blue edge amplitude,red blue normalized index;the sensitive vegetation index of nitrogen balance index is MCARI,GRVI,MTCI,sensitive"three side"”The parameters are red yellow normalized index,yellow edge area and blue edge amplitude;the sensitive vegetation index of water content is GRVI and MCARI,and the sensitive"three sides"parameters are yellow edge amplitude and red yellow normalized index.It is not difficult to find that GRVI,MCARI and normalized index of red and yellow have good sensitivity and stability to biochemical parameters of apple leaves.When using single spectral parameters to estimate chlorophyll content and water content of apple leaves,the red yellow normalized index model should be selected first,and the MCARI model should be selected first for nitrogen balance index.(4)Compared with multiple linear regression,random forest algorithm can greatly improve the accuracy of the model,R~2 has been significantly improved,but the accuracy of RV2 has no significant change.(5)The main components of apple leaf biochemical parameters reduced dimension random forest estimation model and multispectral parameters random forest estimation model have achieved very ideal estimation results.Comprehensive comparison model R~2 and R_V~2 can be obtained.The multispectral parameters random forest estimation model of leaf chlorophyll content has better sensitivity and stability.The main components of leaf nitrogen balance index and water content are The fractal dimension reduction random forest estimation model has better sensitivity and stability.