Skin Detection With Small Unmanned Aerial Systems by Integration of Area Scan Multispectral Imagers and Factors Affecting Their Design and Operation

Abstract

Dismount skin detection from an aerial platform has posed challenges compared to ground-based platforms. A small, area scanning multispectral imager was constructed and tested on a Small Unmanned Aerial System (SUAS). Computer vision registration, stereo camera calibration, and geolocation from autopilot telemetry were utilized to design a dismount detection platform. The test expedient prototype was 2kg and exhibited skin detection performance similar to a larger line scan hyperspectral imager (HSI). Outdoor tests with a line scan HSI and the prototype resulted in an average 5.112% difference in Receiver Operating Characteristic (ROC) Area Under Curve (AUC). This research indicated that SUAS-based Spectral Imagers are capable tools in dismount detection protocols.