Detecção e Rastreamento de Pessoas em Sequências Dinâmicas com Baixa Amostragem Temporal (People Detection and Tracking in Low Frame-rate Dynamic Scenes)

Diego Luiz Siqueira (, Alexei Manso Correa Machado (

1Pontifícia Universidade Católica de Minas Gerais

This paper appears in: Revista IEEE América Latina

Publication Date: April 2016
Volume: 14,   Issue: 4 
ISSN: 1548-0992

People detection and tracking in video sequences are a crucial step for many applications such as security systems and entertainment. Although humans can easily perform these tasks, detecting and tracking people in dynamic background scenes are not trivial for computer vision systems. Furthermore, the amount of data generated by these applications has become overwhelming. Reducing the video frame rate can be an alternative, mainly in security systems, to reduce the amount of generated data. This paper aims to analyze how much a video frame rate can be reduced without affecting the performance of detection and tracking when people move in scenes with dynamic background. A supervised cascade classifier is used for detection and tracking is performed using the Kalman filter. The analysis is based on video sequences recorded from a vehicle. Results show that tracking is very dependent on the frame rate while detection is much more robust in this scenario.

Index Terms:
People Tracking, People Detection, Filtro de Kalman, Adaboost   

Documents that cite this document
This function is not implemented yet.

[PDF Full-Text (496)]