Automatic tag detection in multimedia survey condition

Recently, aquatic ecosystems have captured the interest of the international scientific community, spurring the development of new instruments and methodologies to explore, monitor, and preserve these environmental systems. In this context, the detection of shapes in shallow waters (either in submerged or semi-submerged scenarios) constitutes an intriguing research topic. Our investigation is framed within the multimedia photogrammetry domain, which aims to retrieve geometric information of static objects submerged or semi-submerged in a liquid (usually water), with one or more cameras placed outside the liquid itself. Performing a photogrammetric survey of objects in shallow waters with above-water cameras remains an open research field in the domain of Optics, presenting several theoretical problems and technical bottlenecks, starting with the study of refraction behaviour. Our goal is to develop an automatically applicable methodology to estimate a priori (and correct a posteriori) the effects of refraction on camera behaviour when capturing images of submerged or semi-submerged objects under certain conditions. We therefore tested the behaviour of the cameras in a controlled environment, through different depth levels and water motion conditions, and then elaborated a mathematical model of the optical distortion phenomenon encountered. Feature extraction presents many bottlenecks, mainly due to the particular optical conditions defined by multimedia acquisition and the eventual perturbations present if the liquid is in a condition of turbulence. This paper focuses on such technical problems, presenting part of the qualitative and quantitative results obtained at this stage of our research and the approach used to resolve some detection limitations.