Autonomous Registration of LiDAR Data to Single Aerial Image

This paper proposes the use of phase correlation for the automatic registration of light detection and ranging (LiDAR) data and aerial imagery. First, buildings existent in the LiDAR data and aerial imagery are detected. Then the LiDAR data is interpolated to fixed point spacings, producing both a range image and a building binary mask. In the range image the pixel intensities correspond to the terrain´s elevation and in the building mask the bright pixels correspond to buildings and dark pixels to everything else. A building binary mask is also produced from buildings detected in a corresponding aerial image. The Fourier transforms and the log polar Fourier transforms of both building binary masks are computed. Phase components are correlated and their peaks reveal the translation, rotation and scaling geometric transformation parameters. Results with real data are presented.