People Tracking and Recognition using the Multi-Object Particle Filter Algorithm and Hierarchical PCA Method

This paper presents a method to detect, recognize and track people using mount cameras fixed on a building. The method consists of two independent stages. One is dedicated to detect and track any moving object within the image frame. The other one is in charge to discard any moving object that is not a human being. To perform the first task, a particle filter algorithm is used, in such way that it can perform the tracking of multiple objects. For the recognition stage a PCA (principal components analysis) method is applied to several body parts (head, arms, etc.) respecting their geometrical constraints. The performance of the system has been tested successfully. Some experimental results and conclusions are presented