Inversion Of Cotton Main Growth Parameters Based On Unmanned Aerial Vehicle(UAV)sensing Image

Rapid and non-destructive monitoring of crop growth is the premise of timely and accurate prediction of crop yield,and it is also the core link of modern precision agriculture.The survey of crop growth parameters is to monitor the growth status and changes of crops at all growth stages,It is an important basis for reflecting agricultural information,guiding production and macro-management decision-making.Cotton is one of the most important cash crops in China,The Leaf Area Index and Above-ground Dry Biomass are important parameters to measure cotton growth.The actual monitoring procedures of above-ground biomass and Leaf Area Index is complex,if UAV platform can be used to achieve rapid monitoring,it will play an important role in guiding the realization of precise farmland management.Xinjiang is an important cotton production base in China,Its main planting mode is field planting.This study was conducted from June to September 2018,Taking Xinjiang Academy of Agricultural Reclamation Sciences Test Site as Test Base,the research object is Xinluzao 60 cotton variety.Design of Water-Nitrogen Coupling Test for Cotton,The multispectral images of cotton canopy during the main growth period were obtained by using the Rededge-M multispectral sensor on the 3DR SOLO,meanwhile,synchronous destructive sampling was carried out on the ground to obtain agronomic parameters such as cotton Leaf Area Index and Above-ground Dry Biomass.Based on the above data,extracting multispectral reflectance of cotton canopy from UAV images.Based on spectral index method and BP(Back Propagation)artificial neural network method,the inversion models of cotton Leaf Area Index and Above-ground Dry Biomass were constructed.By comparing the two inversion models,selecting the best models for estimating Leaf Area Index and Above-ground Dry Biomass,The spatiotemporal distribution of growth parameters was also plotted.The results show that:(1)There were significant differences in leaf area index and aboveground dry biomass of cotton under different water treatments.There was no significant correlation between LAI and reflectance at blue,green,red and red edges,but the correlation coefficient between LAI and reflectance at near infrared band was 0.817.Establishing the inversion model of cotton leaf area index based on nine common vegetation indices shows that the quantitative relationship between TVI and cotton leaf area index is the best,the decision coefficient of modeling is 0.75,the Root Mean Square Error(RMSE)is 0.60,and the relative standard error(RMSE%)is20.73%;the R2 of test is 0.80,RMSE is 0.52,RMSE%is 18.06%.The model of cotton leaf area index inversion based on BP artificial neural network is 0.80,RMSE is 0.52,RMSE%is18.06%,and the test R2 is 0.78,RMSE is 0.60 and RMSE%is 20.67%.(2)In the visible to red-edge band(480-720 nm),the above-ground dry biomass was negatively correlated with spectral reflectance.Except for the 720-nm reflectance correlation coefficient of 0.488,the other band reflectance correlation coefficients were all above 0.57;in the near infrared band(840 nm),the above-ground dry biomass of cotton was positively correlated with spectral reflectance,and the correlation coefficient was 0.59.Establishing the inversion model of cotton dry biomass from bud stage to florescence stage by vegetation index shows that MTVI2 and cotton dry biomass inversion model have the best estimation effect.The model determination coefficient R~2 is 0.61,RMSE and RMSE%are 1.17Mg/ha and 21.10%,the model validation R~2 is 0.34,the prediction standard error and relative standard error are1.41Mg/ha and 24.72%,respectively.Estimation model of cotton aboveground dry biomass based on neural network was constructed.The model R~2 was 0.88,RMSE was 0.62 Mg/ha and RMSE%was 11.19%.The external test results were R~2 0.64,RMSE was 1.06 Mg/ha and RMSE%was 18.74%.(3)Regardless of the inversion of cotton leaf area index or aboveground dry biomass,the precision of BP artificial neural network model is obviously better than that of the inversion model based on spectral index.The inversion parameters and the actual parameters are evenly distributed on both sides of the 1:1 line,which shows that better estimation results of leaf area index and aboveground dry biomass have been obtained.(4)In time,with the development of cotton,the leaf area index of cotton increased first and then decreased rapidly in boll stage,while the above-ground dry biomass increased continuously.Spatially,the leaf area index and above-ground dry biomass of cotton increased with the increase of irrigation water content in the same growth period.