Fault-tolerant tracking of multiple targets in collaborative Camera Networks

Abstract

Abstract-Emerging Camera Sensor Networks (CSN) leverage the collaboration, processing and communication capabilities of modern cameras to handle a wide variety of security and safety-critical tasks, including target tracking. However, the performance of CSN in terms of tracking accuracy can be severely degraded when faults occur. Faults may be caused by unpredictable software errors (e.g., in the image processing, feature extraction and data association modules), hardware malfunctions (e.g., in the camera mechanical parts or lens), or as the result of a malicious attack. We propose a decentralized CSN-based system for tracking multiple targets that mitigates the effect of faulty cameras. According to our findings, the proposed solution can well handle faulty camera observations and is able to reliably track a number of targets that may change over time.