Sensor Fusion for Video Surveillance

In this paper, a multisensor data fusion system for object tracking is presented. It is able to track in real-time multiple targets in outdoor environments. The system can take advantage of the redundant information coming from different sensors monitoring the same scene. The measurements (positions of the targets) obtained from the available sources are fused together to obtain a more accurate estimate. Data fusion is performed considering sensor reliability at every time instant. A confidence measure has been employed to weight sensor data in the fusion process. Compared to single camera systems, the adopted approach has produced more accurate and continuous trajectories, reducing calibration and segmentation errors.