NEURAL CELLS INSIGHTS ON PEDESTRIAN DETECTION
Douglas Almonfrey1; Raquel Frizera Vassallo2; Evandro Ottoni Teatini Salles2; Mylène Christine Queiroz de Farias3
1 Instituto Federal do Espírito Santo; 2 Universidade Federal do Espírito Santo; 3 Universidade de Brasília
Resumo
Pedestrian detection has become central in recent years mainly due to the emerging of driver assistance systems. One of the main challenges in this area is how to design a good detector, since there is no general theory that guarantees even regular results in detection and most solutions are developed based on empirical experiments by trial and error. Because of this, some works have proposed bio-inspired or data-driven techniques to solve the problem. In this sense, one of the most important part of a detector is the feature extraction. It is from the feature extraction step that is expected the great improvement in detection quality. This work uses theories related to neural cells to support the generation of features to a classifier that is part of a pedestrian detection system. As baseline, the Filtered Channel Features framework is used to analyse, apply and evaluate the method proposed.
Palavras-chave: Pedestrian Detection; Feature Extraction; Neural Cells; Natural Image Statistics; Filtered Channel Features