A methodology for true orthorectification of large-scale urban aerial images and automatic detection of building occlusions using digital surface model

In urban area with high buildings, conventional orthorectification using digital terrain model (DTM) cannot meet the requirements of true orthoimage generation. Using digital surface model (DSM) integrating from DTM and digital building model (DBM), we propose a methodology in the study to automatically orthorectify large-scale urban aerial image and detect building occlusions for true orthoimage generation. Principle behind the methodology is the optical characteristic of buildings in photogrammetry, in which roof and root sharing the same coordinate are featured with different distances to imagery center. Therefore, in practical operation, we first compute the distances of input image pixels to imagery center and their coordinates is output orthoimage. A data matrix is used to remember the distances and coordinated. Then we compare the coordinates in the matrix. When pixels have the same coordinate, the one with the longest distance represents building root. For true orthoimage generation, only gray value of the roof need to convert into the output image and others are treated as occluded area. By creating an index image, the occluded area can be recorded for next step processing, such as refilling form neighboring orthoimages. Therefore, using a data, matrix and an index image, we are able to automatically orthorectify large-scale urban aerial images and detect building occlusions for true orthoimage generation, provided that the digital surface model is available and the overlap of neighboring image is large enough to ensure 100% visibility of the occluded area.