Author(s): Jin Kashiwada; Kosuke Kawagishi; Riku Kubota; Yasuo Nihei
Linked Author(s): Jin Kashiwada, Yasuo Nihei
Keywords: Image analysis stabilization optical flow flooding
Abstract: In this study, we propose a robust image processing technique that consists of stabilization and velocity analysis for measuring the planar distribution of water surface velocity from videos in which the viewpoint is moved widely and shot urgently during floods. The video, characterized by shaking, panning, tilting, and rotation to capture surrounding flooding conditions, is divided into appropriate time segments and is then subjected to stabilization based on feature point matching, horizontal plane projection based on Ground Control Points, and then optical flow is applied to obtain planar velocity distribution. The proposed method is applied to the video that was taken by residents of the Tsukada River, where severe flooding and loss of buildings occurred, and the fundamental performance of the proposed method for measuring velocity distributions during flood damage was evaluated.
Year: 2025