This project focuses on developing new algorithms for the automatic generation and analysis of 2D and 3D mosaics from acoustic underwater images acquired with high-resolution sonar systems. Current methods for generating 2D acoustic mosaics employ a manual procedure of stitching acoustic images based on the manual identification of landmark points. The manual nature of this process prohibits any additional real-time processing that may be required for ROV applications. We propose the development of an automatic technique for 2D and 3D image stitching based on the automatic or semi-automatic identification of landmark points for the computation of image transformations. This automatic technique will provide the advantage of creating real-time underwater mosaics, which can be used for the above mentioned applications. Technical challenges include handling of shadows, the high variability in the appearance of landmarks and the low resolution of acoustic images. Team:
Sponsor: Kongsberg Mesotech LTD |
Research >