Analysis of Discrete Wavelet Transforms for Different Crops using Sentinel-1 SAR data

The agricultural sector plays an important part in the Indian economy. Crop diversification benefits both the economy and the environment. Crop growth patterns and their practices increase soil fertility and ensure market diversity. Farmers working in small croplands rely heavily on seasonal rainfall for crop irrigation. Synthetic Aperture Radar (SAR) imagery is widely used in crop monitoring and classification. SAR backscattering is sensitive to crop dielectric properties. The study deals with the assessment of crop growth temporal patterns using multi-temporal Sentinel-IA SAR data. The Sentinel-IA Ground Range Detected (GRD) data products (VH and VV polarization) are acquired from 21st August 2019 to 01st January 2022 with a 12-day time interval. Ginger, tobacco, paddy, cabbage, and pumpkin were the crops cultivated during the considered period. The study focused on the applicability of multi-temporal SAR imagery to crop classification and assessment of crop growth. Crop classification using Random Forest (RF) is carried out on SAR images and Discrete Wavelet Transform (DWT) with ‘Haar’ wavelet filter is applied to SAR images to study the effectiveness of wavelet transform on backscattering values. From the study, it is concluded that DWT analysis of VH polarized data can distinguish crops better than VV polarized data with a distinct range of crop pixel values. Hence by using the DWT techniques for crops alone, misclassification can be reduced which helps to improve the accuracy of crop monitoring.