Automatic shadow detection for precise matching points extraction

Using multi-sensor or multi-temporal high resolution satellite images is essential for efficient city analysis and modeling. Yet even when acquired from the same location, these multi-modal images have a geometric inconsistency because of their motion instability, orbital change, and geometrically corrected level. Matching points between images, therefore, must be extracted to co-register the images. With images of an urban area, however, it is difficult to extract matching points because buildings, trees, bridges, and other artificial objects cause shadows, which have different intensities and directions in multi-temporal images. In this study, we propose a shadow detection method to increase the correct-match rate of matching points. The shadow segments are extracted using spatial and spectral attributes derived from the image segmentation. Also, we consider information of shadow adjacency with the building edge buffer. Finally, matching points are extracted through the SIFT method, the representative matching points extraction method, and the points extracted from shadow segments are eliminated from matching point pairs. The results of our study show that we can raise the correct-match rate by about 11% using the proposed shadow detection method.