Real-time multiple object tracking in particle filtering framework using codebook model and adaptive labeling

In this paper, we present an algorithm for multi-object tracking in a particle filtering framework. This algorithm incorporates foreground segmentation, adaptive labeling detections and data association into particle filtering framework. Objects are extracted relying on background modeling by codebook construction. In order to reduce fragment and noise, detections are labeled by adaptive labeling method. Each detection is assigned an independent particle sets. Meanwhile, data association is implemented by Hungarian algorithm. The detection guides a particle filter of one tracker associated detection. Experimental results show that our system is able to automatically and robustly track a variable number of targets and correctly maintain their identities regardless of background clutter and frequent mutual occlusion between targets.