Multi-site Airborne Hyperspectral Datasets for Precision Crop Mapping Applications

Technology infusion in agriculture has been progressing steadily, touching up on various spheres of agriculture such as crop identification, soil classification, yield prediction, disease detection, and weed-crop discrimination. On-demand detection of crop type, often realized as crop mapping, is a primary requirement in agriculture. Hyperspectral remote sensing has emerged as the most versatile technique for the mapping and prediction of various parameters of interest in agriculture. Thanks to the readily availability of a host of machine leaning models and frameworks, the prospect of automatic identification of crop is perceived a realistic goal achievable in the near-future. The ongoing developments in the methods and algorithms of remote sensing data analyses for crop mapping requires the availability of curated, multi-site hyperspectral datasets, under various cases of variations by type and number of crops, geographic site and verifiable ground truth data. The availability of high-resolution airborne hyperspectral datasets across multiple sites with compositions of different types of crops is scare in remote sensing literature. Aimed at enabling the development of knowledge-transfer and potentially automatic approaches for multi-crop mapping using hyperspectral remote sensing, we present a comprehensive high-resolution airborne hyperspectral datasets acquired using AVIRIS-NG for five different sites encompassing different types of crops and agro-climatic regions with associated growth truth data. Though the acquired over sites in India, the diversity in terms of types and number of crops and geographic and agro-climatic diversity maintained by the hyperspectral imaging settings enable the study, implementation and validation of a host of methods and algorithms for agriculture studies using hyperspectral imaging datasets. This is the first of its kind datasets acquisition in which the sensor-imaging and acquisition geometry are maintained consistently similar across different sites and the composting and diversity of crop types maintained for undertaking multi-pronged studies on precision crop mapping suing remote sensing.