Data-Driven Artificial Intelligence for Calibration of Hyperspectral Big Data

Near-earth hyperspectral big data present both huge opportunities and challenges for spurring developments in agriculture high-throughput plant phenotyping breeding. In this article, we data-driven approaches to address the calibration utilizing near-earth agriculture. A data-driven, fully automated workflow that includes a suite of robust algorithms radiometric calibration, bidirectional reflectance distribution function (BRDF) correction normalization, soil shadow masking, image quality assessments was developed. An empirical method utilizes predetermined models between camera photon counts (digital numbers) downwelling irradiance measurements each spectral band established perform calibration. kernel-driven semiempirical BRDF based on Ross Thick-Li Sparse (RTLS) model used normalize changes solar elevation sensor view angle differences attributed pixel location within field view. Following rigorous corrections, novel rule-based methods were developed conduct automatic removal; newly proposed approach assessment; additionally, masking plot-level feature extraction carried out. Our results show processing, storage, analysis pipeline work can effectively handle massive amounts urgent related production sustainable bioenergy food crops, targeting accelerate breeding improving yield biomass traits.