While optical cameras have proven to be valuable assets in underwater visual survey, their application is limited by visibility conditions. In contrast, sonar imaging systems can penetrate silt and generate images at relatively good quality. Currently, the primary obstacle to their deployment is the high cost. However, once 2-D sonar video cameras can be produced at relatively inexpensive prices, they can become part of standard equipment, as an optical video camera currently is, on underwater remotely operated vehicles (ROVs) and other submersible platforms. To deploy such ROVs in a range of conditions, one expects that both cameras would be configured for an overlapping field of view (as the operator's remote eyes) over the targets of interest that are to be searched, inspected, recorded for visual documentation and (or) mapped. Under this scenario, the optical and sonar cameras would provide multi-modal "opti-acoustic" stereovision capability, which provides valuable visual cues about targets of interest where visibility allows.
In this talk, we explore "opti-acoustic stereo imaging" as a new paradigm for 3-D target imaging in underwater. We address the epipolar geometry of a calibrated system, and present methods for 3-D reconstruction and motion estimation. The fundamental problem yet to be solved is how to establish "opti-acoustic correspondences". We propose a particular approach the exploits the motion and stereo epipolar geometries. We present preliminary experimental results to demonstrate the potential application of the new technology.