| Title |
UWB Radar?Based Classification of Marine Debris |
| Authors |
유지현(Jihyeon Yoo) ; 조성빈(Sungbin Cho) ; 고진환(Jinhwan Koh) |
| DOI |
https://doi.org/10.5573/ieie.2026.63.8.73 |
| Keywords |
UWB radar; Marine debris classification; Material identification; Artificial intelligence; Signal-based sensing |
| Abstract |
Marine debris causes serious environmental problems, including marine ecosystem degradation and maritime accidents. Therefore, a technology capable of accurately classifying such debris is required to establish effective management and collection strategies. However, the mainstream vision-based classification methods are vulnerable to external environmental factors, making it difficult to ensure stable performance in real marine environments. To overcome these limitations, this study proposes a marine-debris classification method using Ultra-Wideband (UWB) radar, along with a neural network architecture designed by considering the characteristics of UWB radar-based B-scan data. Various types of marine-debris data were collected using the UWB radar, and classification performance was compared using CNN models. Experimental results showed that conventional CNN models exhibited low accuracy in the 80% range and unstable performance in Out-of-Distribution (OOD) environments. To address this issue, a lightweight hybrid neural network was designed by combining an initial large-kernel-based feature extraction structure with residual learning. The proposed model achieved an accuracy of 97.17% in cross-validation. These results suggest that UWB radar can complement the limitations of image-based sensors and serve as a reliable alternative for marine-debris classification in real marine environments. |