Underwater localization using imaging sonars in 3D environments

Abstract

This work proposes a localization method using a mechanically scanned imaging sonar (MSIS), which stands out by its low cost and weight. The proposed method implements a Particle Filter, a Bayesian Estimator, and introduces a measurement model based on sonar simulation theory. To the best of author’s knowledge, there is no similar approach in the literature, as sonar simulation current methods target in syntethic data generation, mostly for object recognition . This stands as the major contribution of the thesis as allows the introduction of the computation of intensity values provided by imaging sonars, while maitaining compability with the already used methods, such as range extraction. Simulations shows the efficiency of the method as well its viability to the utilization of imaging sonar in underwater localization. The new approach make possible, under certain constraints, the extraction of 3D information from a sensor considered, in the literature, as 2D and also in situations where there is no reference at the same horizontal plane of the sensor transducer scanning axis. The localization in complex 3D environment is also an advantage provided by the proposed method.