Regresar

Cascade classifiers based robust pedestrian detection

Abstract:

This paper proposes a dataset and algorithms for pedestrian detection in UAVs. The method proposed is a HAARLBP based cascade classifier combined with saliency maps for improving the performance of the detector. In addition we introduce a dataset with images from surveillance cameras at different angles and altitudes emulating a UAV. We validate our dataset by the implementation of HOG algorithm and compared it with other approaches from the literature. The results show that HAAR-LBP algorithm has better performance than HAAR like features; our dataset is better for pedestrian detection using UAVs and the use of saliency maps improves the performance of cascade classifiers.