A UWB Radar and Machine Learning-Based Tool for Detecting Victims Through Foliage in Search and Rescue Operations

This research explores the application of radar sensors and Machine Learning (ML) methods for human detection, aiming at helping First Responders (FRs) in their duties. Use cases of interest include scenarios where individuals, possibly victims of a blizzard or extreme temperatures, are obscured by vegetation and foliage - a common challenge during Search and Rescue (SAR) operations in forests and the wilderness. In such cases, the transmitted radar signal scattered from the unpredictably moving foliage is added to the one scattered by the victim. This adds extra layers of complexity to the problem at hand and specifically to the task of discriminating between the two signals while classifying the latter. The contribution of this work lies in: (a) the integration of a proposed tool for victim detection through foliage using predominantly Commercial-Off- The-Shelf (COTS) components and a radar sensor, (b) the development of a first-of-its-kind, openly accessible dataset of different human subjects in diverse types of vegetation and wind conditions during a 3 month period, and (c) the development of an ML-based tool, exploiting transfer learning, to accurately detect the presence of subjects concealed by foliage. The research presents promising results in terms of detection success rates, underscoring the potential of this technology in enhancing FRs' capabilities.