| In recent years, remote sensing has been more and more used in monitoring wheat growth and perdicting grain yield and quality of wheat due to its rapidity, non-destruction and big range.In order to improve the reliability and accuracy of the remote sensing monitoring model. The research was conducted in some cities in middle Jiangsu, we picked SPAD, LAI, leaf nitrogen content and biomass as main parameters of growth at jointing, booting and flowering stage. At the same time, HJ-1A/1B remote sensing data was downloaded and analyzed. We built several combinations based on the remote sensing variables and analyzed the correlation with main parameters of seedlings for establishment of a monitoring model. After the model was established, we evaluated it and compared it with the model of single variable to research the feasibility of monitoring main parameters of seedlings by models based on variable combinations of remote sensing.(1) By analyzing the correlation between main parameters of seedlings and difference combinations of the remote sensing variables, it shows that there are obvious correlations between LAI, SPAD, leaf nitrogen content, biomass and difference combinations of remote sensing variables during main during growth season. And at jointing stage, there’s a big increase in R2 and a marked decrease in RMSE from 5.56,0.37,425.78 to 4.78,0.38,347.14 due to the monitoring model changing from single-variable model to the models combined of NDVI-GNDVI, SIPI-EVI and NDVI-SIPI to monitor SPAD, leaf nitrogen content and biomass. There’s an accuracy’s increase of 14%,18.9%and 18.5%. It is feasible and even better than single variable. At booting stage, it is better than single variable to monitor SPAD, LAI, leaf nitrogen content and biomass by the models combined of NDVI-GNDVI, NRI-RVI, PSRI-DVI and NDVI-PSRI. There’s a marked decrease in RMSE from 2.45,0.55,0.22 to 2.07,0.49 and 0.18, there’s an accuracy’s increase of 15.5%,10.9% and 18.2%. But RMSE increases a little while R2 increases obviously when monitoring biomass. At flowering stage, there’s a big increase in R2 and a marked decrease in RMSE from 4.70,1.36,2062.11 to 4.09,1.19,1884.11 due to the monitoring model changing from single-variable model to the models combined of NDVI-GNDVI, NDVI-GNDVI and NDVI-PSRI to monitor SPAD, LAI and biomass. There’s an accuracy’s increase of 13%,12.5% and 8.6%. It is better than single variable.(2) By analyzing the relationship between main parameters of seedlings and ratio combinations of the remote sensing variables, it shows that there are obvious correlations between SPAD, LAI, leaf nitrogen content, biomass and ratio combinations of remote sensing variables during main during growth season. And at jointing stage, there’s a marked decrease in RMSE from 5.56,0.84,0.37,425.78 to 4.40,0.73,0.38 and 416.21 due to the monitoring model changing from single-variable model to the models combined of NDVI/GNDVI, NDVI/GNDVI, SIPI/EVI and NDWSIPI to monitor SPAD, LAI, leaf nitrogen content and biomass. It is feasible and even better than single variable, there’s an accuracy’s increase of 20.9%,13.1%and 2.2%. At booting stage, it is also better than single variable to monitor SPAD, LAI, leaf nitrogen content and biomass by the models combined of NDVI/GNDVI, SIPI/RVI, GNDVI/DVI and NDVI/GNDVI. There’s a marked decrease in RMSE from 2.45,0.55,0.22 to 2.08,0.48 and 0.16, there’s an accuracy’s increase of 15.1%,12.7%and 27.3%, but RMSE increases a little while R2 increases obviously when monitoring biomass. At flowering stage, there’s a big increase in R2 and a marked decrease in RMSE from 4.70,1.36,2062.11 to 3.85,1.08,1847.85 due to the monitoring model changing from single-variable model to the models combined of NDVI/DVI, NDVI/SIPI and GNDVI/DVI to monitor SPAD, LAI and biomass. It is better than single variable. There’s an accuracy’s increase of 18.1%,20.6%and 10.4%.(3) By analyzing the relationship between main parameters of seedlings and normalized combinations of the remote sensing variables, it shows that there are obvious correlations between LAI, SPAD, leaf nitrogen content, biomass and normalized combinations of remote sensing variables during main during growth season. And at jointing stage, there’s a big increase in R2 and a marked decrease in RMSE from 5.56,0.84,0.37,425.78 to 5.10,0.71,0.34 and 399.95 due to the monitoring model changing from single-variable model to the models combined of (NDVI-GNDVI)/(NDVI+GNDVI),(NDVI-GNDVI)/(NDVI+GNDVI), (SIPI-EVI)/(SIPI+EVI),(NDVI-SIPI)/(NDVI+SIPI) to monitor SPAD, LAI, leaf nitrogen content and biomass. There’s an accuracy’s increase of 8.3%,15.5%,8.1%and 6.1%. It is feasible and even better than single variable. At booting stage, it is also better than using single variable to monitor SPAD, LAI and biomass by the models combined of (NDVI-GNDVI)/(NDVI+GNDVI) and (SIPI- RVI)/(SIPI+RVI) and (NDVI-GNDVI)/(NDVI+GNDVI). There’s a marked decrease in RMSE from 2.45,0.55 to 2.27 and 0.44, and an accuracy’s increase of 7.3% and 20%, but RMSE increases a little while R2 increases obviously when monitoring biomass. At blooming stage, there’s a marked decrease in RMSE from 4.70,1.36,2062.11 to 4.35,1.30,1730.07 because of the monitoring model changing from single-variable model to the models combined of (NDVI-GNDVI)/(NDVI+GNDVI), (SIPI-RVI)/(SIPI+RVI) and (NDVI-GNDVI)/(NDVI+GNDVI) to monitor SPAD, LAI and biomass, there’s an accuracy’s increase of 7.4%,4.4%and 16.1%. It is better than single variable. |
PDF Full Text Request