Strategy To Automatically Detect Agricultural Parcels With Abnormal Agronomic Development

Abstract

Using an unsupervised outlier identification method, this research investigates the detection of abnormal crop growth at the parcel level. Rapeseed and wheat fields in India are used for the experimental validation. The suggested approach may be broken down into four distinct phases: Using data from the Sentinel-1 and Sentinel-2 satellites, we perform four steps: (1) preprocessing; (2) extracting pixel-level features from the SAR and multispectral data; (3) computing parcel-level features using zonal statistics; and (4) detecting outliers.