Abstract: Integrated multiple lens cameras have been adopted for aerial photogrammetry in the industries in recent years thanks to its high data-acquisition efficiency. However, the cameras yield redundant photos for non-target areas which causes extra computation power. This issue becomes serious when the shape of the target area gets complicated. In this paper, we propose a novel screening method for the redundant aerial photos based on optical geometry and building recognition algorithms. The method first project the photos onto target area, and the overlaps are estimated. Combining the building recognition results with the estimated overlap ratios, the redundant photos can be identified and removed. The results suggest that the proposed method is able to reduce 70% of the captured photos but keep almost the same accuracy for reconstruction of the target area.

Key words: oblique photogrammetry, redundant photo removal, overlapping area estimation, pattern and feature recognition