Underwater object recognition in photo images

Abstract

The identification problem of artificial or natural objects in underwater images is considered in this paper. Any object (including a previously received image) with unequivocally determined contours can be used as a pattern. Recognition of the required object is made by an exhaustive search of key points or small pieces of borders with a subsequent increase of accumulator cells. Processing of real sea-bottom images has confirmed the algorithm's efficiency (including dead reckoning, marker recognition and line recognition tasks).