RGB-D Sensor-Aided Radio Map Estimation Using Materials Classification

Abstract

This letter proposes a novel RGB-Depth (RGB-D) sensor-aided radio map estimation framework using materials classification. Ray-tracing simulators for radio map estimation require expensive and detailed vector data and materials information for use indoors. Advances in image sensor (e.g., RGB-D, LiDAR)-based methods can reduce this cost but often overlook material information. This letter shows that color information from RGB-D sensors provides more accurate estimations by coupling with materials classification. Experiments with Wi-Fi devices showed our method can distinguish materials even if the shapes of the objects are similar, achieving a 2.14[dB] gain in non-line-of-sight environments.