Building facade detection via plane support points clustering

In this paper, a building facade detection algorithm is proposed. Proposed algorithm first detects the edges present in an image and connects them into line segments. Then a polar histogram mapping is used to find the vanishing points in the scene. An approach with low complexity is introduced to cluster the line segments according to their vanishing points. Line segments and their corresponding vanishing points help to cluster the edges according to the plane surfaces they support. Clustering of plane support points determines the planes present in the scene and thus enables detection of building facades. Our new plane support point clustering algorithm has low complexity and less computation time in comparison with other prior art algorithms. Results show that our proposed algorithm detects facades with 78% accuracy.