딥러닝 기반 육상기인 부유쓰레기 탐지 모델 성능 비교 및 현장 적용성 평가 = A Performance Comparison of Land-Based Floating Debris Detection Based on Deep Learning and Its Field Applications

다국어 초록 (Multilingual Abstract)

A large amount of floating debris from land-based sources during heavy rainfall has negativesocial, economic, and environmental impacts, but there is a lack of monitoring systems for floating debrisaccumulation areas and amounts. With the recent development of artificial intelligence technology, thereis a need to quickly and efficiently study large areas of water systems using drone imagery and deeplearning-based object detection models. In this study, we acquired various images as well as drone imagesand trained with You Only Look Once (YOLO)v5s and the recently developed YOLO7 and YOLOv8sto compare the performance of each model to propose an efficient detection technique for land-basedfloating debris. The qualitative performance evaluation of each model showed that all three models aregood at detecting floating debris under normal circumstances, but the YOLOv8s model missed orduplicated objects when the image was overexposed or the water surface was highly reflective of sunlight.
The quantitative performance evaluation showed that YOLOv7 had the best performance with a meanAverage Precision (intersection over union, IoU 0.5) of 0.940, which was better than YOLOv5s (0.922)and YOLOv8s (0.922). As a result of generating distortion in the color and high-frequency componentsto compare the performance of models according to data quality, the performance degradation of theYOLOv8s model was the most obvious, and the YOLOv7 model showed the lowest performancedegradation. This study confirms that the YOLOv7 model is more robust than the YOLOv5s andYOLOv8s models in detecting land-based floating debris. The deep learning-based floating debrisdetection technique proposed in this study can identify the spatial distribution of floating debris bycategory, which can contribute to the planning of future cleanup work.