A rapid matching approach for large-scale urban imagery

摘要

Image matching is a common technique in series of remote sensing imagery applications. The proposed method conducts image matching rapidly in complex urban based on widely existing rectangular buildings. The optimized framework combines the external contour and internal structure to construct regional feature representing building roof, in which perceptual grouping and classification is introduced simultaneously. Then a newry proposed Euler Number matrix, combined with conventional area index and histogram, describes the regional feature, and relevant matching metrics measure the similarity. Specific experiments performed on urban areas with repetitive patterns and poor texture justify the effectiveness and robustness of the proposed image matching approach. The method can require the demand of image retrieve, image mosaic, and serve as primary match of geometric measurement applications.