Underwater image mosaicing using maximum a posteriori image registration

Abstract

An underwater remotely operated vehicle (ROV) was developed to allow inspection process without the need for human divers to enter the water. The ROV has a video camera, a Doppler navigation sonar and gyroscope-based orientation sensor. Each image of the underwater scene is saved along with the video camera's instantaneous position and orientation. The images are then patched together into a large composite picture of the structure. The method, which is based on the maximum a posteriori estimation technique, combines the least-mean-squared-error estimator and Kalman Filter. It provides smooth and robust image shift estimation. This method has been tested and shown as a practical and potentially useful underwater inspection tool.