構造物被害状況の自動抽出のための画像解析手法の開発

抄録

Development of an image analysis method for automated identification of post-seismic structural damages is presented in this paper. Identification of structural damages is carried out by i) making candidates of damaged structure model through numerical simulation, ii) comparison of the photo-image of damaged structure itself and templates generated from damaged structure model, and iii) identification of corresponding damaged structure model. The biggest technical difficulty to make this system work is the development of effective image analysis method for the phase ii), i. e., comaparison of picture-image and visualized numerical model. To overcome this difficulty, we developed image analysis method with a priori knowledge on the objects to be analyzed. This image analysis method focuses on topological characteristics of the objects, specifically, relative position of sharp corners of the object, which is a good indicator for the change in the shape of the buildings.