航空写真からの屋根種類判別と稜線認識に関する研究 : 明度分布の曲線近似と平面回帰による2段階判別法の構築

概要

論文の詳細を見る
This paper proposes a method for discriminating roof types and determining roof ridges from aerial images. Observing that according to the slope change of a roof surface its brightness changes, our method first approximates the histogram of brightness distribution as a smooth curve. We then extract effective characteristic variables to develop a discrimination model for classifying roof types into deck roof and inclined roof. For the roofs which are classified as inclined ones, we develop a method for classifying them into hip roofs and gables after extracting additional explanatory variables obtained by applying multiple regression to 3-dimensional brightness distribution. We finally applied the model to aerial image of residential area in Kyoto city, and observed that the classification ability of our proposed method was quite high.
社団法人日本建築学会の論文
2007-01-30