Landsat 8: UDT-CWT based Denoising and Yield Estimation

This paper proposes the regression model for the estimation of the yield for the two crops (Rice and Groundnut) by applying image processing techniques on Landsat 8 imagery. The Undecimated Dual-Tree Complex Wavelet Transform (UDT-CWT) is used to denoise the Landsat 8 Imagery. After that Remote Sensing (RS) parameters like Leaf Area Index (LAI), and 6 Vegetation Indices (VI) including Normalized Difference of Vegetation Index (NDVI), Soil and Atmospherically Resistant Vegetation Index (SARVI), and Modified Soil Vegetation Index (MSAVI) etc., were retrieved. This paper, reveals the impact of noise on these derived RS parameters and the denoising technique results PSNR= 49 dB; SSIM=1 for SARVI. Then designed a regression model with this SARVI approach. This regression model also involves Meteorological Parameters like Rainfall, Temperature, Evapotranspiration which are collected from Automatic Weather Stations (AWS). The outcome of the proposed estimation model is precisely validated for a period of 2014 to 2017 with ground truth data and outperforms good correlation R-2=0.78 for Kharif and R-2=0.74 for Rabi seasons.