Inversion Of Canopy Leaf Area Index In Different Growth Periods Of Broccoli Based On Multi-source Remote Sensing Data

The leaf area index(LAI)of crops determines the growth and yield of crops.Therefore,accurate and rapid estimation of the LAI of crops can help better monitoring.In this study,broccoli was used as the research object,and three different growth stages of broccoli were obtained from the slow seedling stage,rosette stage and heading stage.Combined with Sentinel-2 and UAV remote sensing images during the same period,different growth stages and different LAI values were analyzed.Spectral characteristics of broccoli canopy at time;Pearson correlation between measured LAI and band reflectance and vegetation index was calculated;3 different data input modes were constructed,and 4 different machine learning algorithms were used to establish a model for inversion of broccoli LAI at different growth periods,To test the stability of the model,verify the accuracy of the model,find the best inversion model,and compare and analyze the inversion results of the two types of remote sensing data.The main conclusions are as follows:(1)The spectral reflectance characteristics of broccoli canopy with different growth periods and LAI values have similar changing trends.The canopy reflectance in the visible light region between 400nm and 700nm shows a trend of first rising and then falling,at the blue wavelength of 450nm and the red wavelength of 650nm There is an absorption valley on the left and right,and a reflection peak at about 550 nm of the green wavelength.In the near-infrared range from 700nm to 850nm,the reflectance increases with the increase in wavelength,and the maximum reflectance value appears around 850nm in the red band.After 850nm,the reflectivity decreases as the wavelength increases.(2)Based on the band reflectivity extracted from Sentinel-2 remote sensing images and the calculated vegetation index,the two red-side parameter bands B6 and B7,and the two near-infrared bands B8 and B8a have a very significant correlation with the measured LAI.Among them,the B8 band Has the best performance and the IRECI index has the best correlation.Correlation between Band 85(wavelength 738nm),Band 96(wavelength 783nm),Band 111(wavelength 843nm)and Band 116(wavelength 863nm)and the measured LAI based on the band reflectivity extracted from the UAV hyperspectral image and the calculated vegetation index The performance reached a very significant level of 0.001.Among them,Band 111(wavelength 843nm)performed the best,and the IRECI index had the best correlation.The correlation coefficient of the overall sample is greater than that of a single growth period,and the correlation is better.(3)Based on the band reflectivity extracted from Sentinel-2 remote sensing images and UAV hyperspectral images and the calculated vegetation index,the machine learning algorithm is used to invert the broccoli LAI.It is found that in the training set,the GPR algorithm is equal to the three growth periods of broccoli.Shows strong predictive ability and best stability.In the test set,when using the GPR algorithm,it is found that the mode 1 with the band reflectance data input mode has the best inversion accuracy,and as the growth period advances,the LAI value increases,and the inversion accuracy also increases.(4)Comparing Sentinel-2 remote sensing images and UAV hyperspectral images using machine learning algorithm inversion results,it is found that under the same data input mode and machine learning algorithm,UAV hyperspectral images have higher inversion accuracy.