A new method for calibration of an opti-acoustic stereo imaging system

Range measurements offer a key advantage in deploying a sonar and an optical camera as a multi-modal stereo imaging system for 3-D object shape recovery in underwater. Establishing the relative pose of these cameras by a priori calibration enables exploiting the opti-acoustic epipolar geometry, and confine the complex multi-modal correspondence problem to a 1-D search. By imaging points on a known planar grid with distinct visual and acoustic reflectance properties, we previously proposed an iterative optimization scheme for opti-acoustic stereo calibration, requiring a search in 6+3N space based on N opti-acoustic stereo pairs of the grid at different orientations. In this paper, we present a new calibration method, where most of the computations are carried out in closed form, with a final search in 3-D space. Although extendable to make use of data from several views, the current implementation applies to only one view of the target grid. Thus, the results can be less accurate than the solution from the iterative method, for sonar cameras with a small field of view. We assess the performance of this method and discuss its merits based on experimental results with synthetic and real data.