Underwater object reconstruction method for target recognition in an Autonomous Underwater Vehicle

We consider the problem of reconstructing 3D models of underwater objects from multiview images and propose a low complexity algorithm to achieve this without loss in accuracy as compared to existing work. We utilize bicubic interpolation as a substitute for Gaussian decomposition to arrive at multiresolution images. We use Kolmogorov modeling to generate disparity maps from various views and hence obtain surface terrains which are combined to construct a complete 3D model. We demonstrate that the proposed algorithm entails significant time saving while retaining the structural integrity necessary for the recognition system in an Autonomous Underwater Vehicle (AUV).