Detecção e acompanhamento automáticos para sonares

Abstract

In this work, the construction of autonomous contact detection and tracking system is addressed, analyzing each of the necessary steps. Starting with the presentation of the data from the sonar processing, both passive and active, going through the conversion of the data to a set of points, called detections. Moving on to the application of techniques to reduce dimensionality, such as clustering. In addition to observing the problems of direct application of clustering techniques to temporal space data, changes are proposed in the classic techniques to suit the application and improve the performance of the algorithms. Subsequently, each set of clusters is applied to the tracking system, which should prospect the presence of contacts, monitor the temporal evolution of the clusters of each contact and analyze the movement of the contacts for a future position and speed prediction.