An online multi-object tracking by detection approach based on particle filtering framework

In this paper, we address the problem of automatically detecting and tracking a variable number of persons in complex scenes using a monocular, potentially moving, uncalibrated camera. We propose a novel approach for multi-object tracking by detection in a particle filtering framework. First, in order to improve detection performance, moving objects are automatically extracted using boosting factor and labeling method. Then, multiple objects are tracked in particle filtering framework by adaptively selecting block template according to human model and adaptively updating template using current detection. Finally, we resolve the data association using Hungarian assignment algorithm and compute the observation likelihood function of each particle filter using the associated detection. We present a new method to deal with the variety of the number of objects at random times. The resulting algorithm robustly tracks a variable number of dynamically moving objects in complex scenes with occlusions. The approach relies only on information from the past and is suitable for online application, and does not require any camera or ground plane calibration. We evaluate the performance on a variety of datasets and show that it improves upon state-of-the-art methods.