Graph representation for room layout matching using spectral embedding

Abstract

Graph matching is efficient to search similar layout when the architectural floor plan data size is increasing. Because the computational time of floor plan matching using spectral embedding is only in seconds, so it is one of the popular methods. However, the isomorphism of each floor plan leads to low accuracy in the matching process and it becomes the weakness of this method. Therefore, we propose a graph representation for room layout matching using spectral embedding. Normally, graph representations of the floor plan define nodes as rooms and edges as connections between rooms. Besides, the graph spectral embedding is to find the feature vector of each floor plan by ignoring the semantic of rooms. Our proposed method also considers both room semantic which is the connection between the area outside and inside the room, and the structure of each layout. Furthermore, we show that by adding an extra node, our method can handle the isomorphism of a graph based on a mathematical idea called eigenvalue testing. There are three main processes in the proposed method: floor plan extracting, appended topology graph and floor plan matching. The performance from our experiment shows that our proposed method improve the matched accuracy from the conventional method by about 27.81 percent.