Abstract of Thesis presented at COPPE/UFRJ as a partial fulfillment of the requirements for the degree of Master of Science (M.Sc.)

Multiple Target Tracking Strategies: A Comparasion Between the Data Association Approaches MHT and Neural Networks

Manuel Ramón Vargas Avila

April/2015

Advisors:

Luiz Pereira Calôba Beatriz de Souza Leite Pires de Lima

Department:

Eletrical Engineering

      This work presents a multi-target tracking methodology based on radar sensor information in marine environments using artificial neural networks as data association method. The performance of this methodology is compared with the classical approach based on the data association algorithm Multi Hypothesys tracking (MHT). With the purpose of performing this comparison and validating the proposed methodology was used an artificial data generator and created a simulation tool for assessing different performance metrics.
      In general, the proposed methodology allowed a reduction of the execution time of the algorithm by comparing with the MHT approach, in addition, in some scenarios where leak detection are introduced, the classical methodology presents difficulties in maintaining tracking of some targets, which may be considered one of its weaknesses. Regarding the number of false tracks created, the classical methodology presents a greater number in relation to the proposed methodology, probably because of the complexity in the search strategy (combinatorial) and the initialization process. This demonstrates the feasibility of using the methodology proposed in this dissertation.