Title: Multispectral Imaging, Detection and Active Reflectance Instrument (MiDAR)-fused Machine Learning for Augmenting Earth Science Datasets
Presenting Author: Ved Chirayath
Organization: NASA Ames Research Center
Co-Author(s): Ron Instrella

Abstract:
We share developments from a MiDAR-fused supervised machine learning algorithm and parallelized open-source toolkit to augment past, current and future Earth Science Datasets from NASA's Earth Observing System (EOS). MiDAR is a new active remote sensing technology, based on ESTO-funded FluidCam instruments, selected under a NASA 2016 CIF grant for multispectral imaging, detection and active reflectance that offers two orders of magnitude higher Signal to Noise Ratio (SNR) and spatial resolution than NASA's current or future EOS satellites. Using airborne cm-scale MiDAR and FluidCam datasets, we developed a supervised machine learning algorithm for efficient augmentation of EOS datasets with multidimensional MiDAR multispectral imagery, providing quantifiable improvements in habitat classification and assessment accuracy. Multidimensional supervised machine learning and airborne MiDAR and FluidCam data-fusion present accurate assessment tools for large-scale 3D surveys of terrestrial and aquatic habitats with cm-scale spatial resolution, fine temporal sampling and quantifiable morphological and biological characterization. Ultimately, such data-fusion, coupled with supervised machine learning, can be used to significantly augment existing NASA EOS datasets.