A stereo imaging framework in 3-D mapping of benthic habitats and seafloor structures

We address the deployment of stereovision imaging for underwater 3D mapping. A key component in system performance is the ability to determine the vehicle's position during data acquisition, ensuring that the images are acquired at desired positions along the pre-planned trajectory. We investigate the use of stereo images from the integration of incremental motions between consecutive frames. This is achieved within a complete framework, comprising (1) suitable trajectories to be executed for data collection, (2) data processing for mapping as well as for trajectory following and recursive alignment of images, and finally (3) 3D mapping by the fusion of various visual cues, including motion and stereo within a Kalman filter. The computational requirements of the system are evaluated, formalizing how online processing performance may be achieved. Experiments with underwater images are presented to demonstrate how the trajectory estimation is improved by the proposed alignment scheme.