NDVI Data Correction of Alpine Grassland Remote Sensing Image

Remote sensing data of satellite is not accurate enough caused by the atmosphere factor, etc. In this paper we propose to correct the remote sensing data with the UAV image for alpine grassland. Firstly, for each 30x30m center area of the drone multispectral image, the related pixel in landsat8 remote sensing image with the same longitude and latitude coordinate is retrieved. Then three fitting function models, construct linear, quadratic polynomial and logarithmic are tried on samplings. The accuracy of the models are evaluated through the determinable coefficient (R2), mean absolute error (MAE) and root mean square error (RMSE). The experiment results show that the construct linear prediction model has the highest prediction accuracy for the NDVI value of alpine grassland with determine factor R2=0.2949, prediction accuracy MAE=0.01867, RMSE=0.01181. The prediction curve of the construct linear prediction model has the perfect matching performance with the image data captured by the drone.