Multiple facade images matching for 3D building modeling

Reliable facade image matching is an essential and difficult task for 3D building modeling. This paper presents a hybrid approach for multiple facade image matching method for 3D building modeling, and it employed multiple matching primitives for matching. The proposed method is a three-stage process: the first stage is feature matching, which matches DOG features combining high distinctive SIFT descriptor to constrain subsequent edge matching. The edge matching checked by the figural continuity constraint will be used to constrain the interest point matching. The second stage is to filter feature matching results in each segment from Mean-shift segmentation method in object space for ensuring the reliability. The third stage is to propagate the feature matching result to its neighborhood to obtain more matching results. Two sets of actual triple-overlapped facade images were used for the experiment, the results revealed that the proposed approach can produce dense and accurate matching results.