Assessment of Multispectral Imaging System for UAS Navigation in a GPS-denied Environment

Abstract

Multispectral (MS) imaging systems have been used for the detection, identification, and quantification in numerous environmental and military applications already. It is proposed to analyze feasibility of utilizing this emerging technology on small unmanned aerial vehicles (sUAS) for the purpose of enhancing accuracy and precision of object detection (identification), classification and tracking (DCT) that may contribute to a variety of downstream applications including threat detection, forensics, battle damage-assessment, additional/alternative aid to navigation (ATON) in the GPS-degraded or GPS-denied environments. This study assesses applicability and benefits of using a MS sensor as opposed to standard infrared (IR) and/or electro-optical (EO) sensors for DCT applications. It also includes an assessment of the computer-vision (CV) and artificial intelligence (AI) algorithms to quickly and reliably process the sensor output data. It is envisioned that a MicaSense RedEdge-MX or Altum like high-resolution global-shutter 5-band MS sensor integrated with a commercial-of-the-shelf (COTS) Group 1 or Group2 sUAS will be used to collect data to train a deep-learning (DL) convolutional neural network (DCNN) capable to handle one or two specific DCT problems to address the following research questions: Whether using multiple spectral bands has any benefits compared to a standard EO sensor or EO sensor combined with IR sensor? That includes benefits of having a spectral profile of surrounding background area and objects from the standpoint of more reliable/precise DCT. What are the limitations of using MS sensors and CV/AI algorithms to process data from the standpoint of operating environment, terrain, altitudes, object size and material, time of the day, weather, number of spectral bands, resolution, narrow field of view, addition of a downwelling light sensor)? What computational resources would be required to enable DTS capability aboard COTS sUAS The study will look at the requirements to such a system and its CONOPS, followed by conducting numerical experiments and field testing to gather and analyze data coming out of a MS imaging sensor. It is expected to involve SE, OC and CS students, and summarize all the findings in the final report.