VPSLAM: VISUAL PROBABILITY MASS FUNCTION SLAM

Clebson Joel Mendes de Oliveira¹; Raquel Frizera Vassallo¹; Evandro Ottoni Teatini Salles¹

¹ Universidade Federal do Espírito Santo - UFES

Baixar PDF

Resumo

We present a novel Visual Simultaneous Localization and Mapping (VSLAM) algorithm, the VPSLAM algorithm (Visual Probability Mass Function SLAM), and tested it using some image sets from the state of the art in SLAM methods. The obtained results show that the proposed VPSLAM algorithm can perform place recognition in outdoor environments and is able to detect loop closures, even in low resolution imagery. Also, we can conclude that the characterization model C-IGL (de Oliveira et al., 2015) can be applied in appearance-based SLAM algorithms over dynamic environments.

Palavras-chave: Visual SLAM; Probabilistic Model; Outdoor Environment