On the 3-D Reconstruction of Seabed Using Multiple Sidescan Sonar Images

This paper proposes an original method for the 3-D reconstruction of seabed from multiple sidescan sonar images. The proposed method adopts a stereo-like vision approach based on a combination of the GPS positioning and the image matching technology to correspond among multiple sidescan sonar images taken from different viewmints. and then estract the 3-D structure information of 'seabed from the images according to the relation of the spatial position. This method consists of three stages of processing. First, the sidescan sonar is tracked in 3-D space. Second, based on maximum slant range, the corresponding sub-image is extracted from each sidescan sonar image. Third, the method of multi-step gray-level prqjective matching is applied to find the corresponding points among the sub-images. Finally, according to the spatial relation in 3-D space, the 3-D structure of seabed is estimated and reconstructed. The experiment with EdgeTech's DFIOOO sidescan sonar and JRC's DGPSZOO GPS is reported.