EarthEncode: A Vision-Language Framework for Balanced SoilProperty Mapping
147
by Vishvam Porwal
Published onMay 27, 2024
EarthEncode: A Vision-Language Framework for Balanced SoilProperty Mapping
·
Abstract
Digital Soil Mapping (DSM) is a technique for creating soil property maps and their geographical distribution, crucial for soil health monitoring and sustainable land use [1]. Despite advancements in DSM, mapping soil characteristics remains challenging due to high variance in soil properties. Existing strategies for optimizing the count and positioning of soil samples draw on environmental data, such as elevation and LiDAR data [2,3]. However, these techniques do not account for the accessibility challenges of a location, such as the existence of buildings, barns, or other structures. This oversight leads to a bias towards easily accessible locations, resulting in unbalanced soil property mapping. The use of satellite imagery to evaluate location accessibility can address this issue, thereby facilitating more balanced soil property mapping.
Creative Commons Attribution 4.0 International License (CC-BY 4.0)