OBSERVATION AND BEHAVOIR ANALYSIS OF FLOATING DEBRIS IN AN URBAN TIDAL RIVER BY USING A DEEP LEARNING MODEL

DOI Web Site

Bibliographic Information

Other Title
  • 都市感潮河川における深層学習モデルを用いた浮遊ごみの連続観測と挙動解析

Abstract

<p> Using the deep learning model, YOLOv5, we constructed a system to continuously detect floating debris in rivers from water surface images captured by fixed-point cameras. This system gave good detection in accuracy results. We conducted observations of floating debris using multiple fixed-point cameras and the developed system in the Hirano River, an urban tidal river in Osaka, and analyzed the spatiotemporal behavior characteristics of floating debris. The behavior of floating debris was mainly controlled by tidal currents, and the accumulation was observed in the upper reaches of the river, where the flow was stagnant at high tide. The density of floating debris tended to be larger in the upstream of the river and smaller in the downstream, due to the material diffusion caused by the fluctuating component of the tidal flow was larger in the downstream than in the upstream. In the Hirano River, a large amount of floating debris was carried downstream even during sunny days, and it was considered that the amount of debris directly dumped on the water surface increases due to easy human access along the river.</p>

Journal

Keywords

Details 詳細情報について

Report a problem

Back to top