Multiple visual-targets tracking in decentralized wireless camera sensor networks

To track 3-dimensional multiple visual-targets states, we develop the decentralized Wireless Camera Sensor Network (WCSN) and the corresponding data fusion scheme. In our proposed scheme, each activated camera sensor-node obtains the local encoded 2-dimensional observation of the visual-targets´ states and sends it to the global fusion center. This global fusion center processes the local information and provide the 3-dimensional estimation of the multiple visual-target states. We also propose the distributed visual-PHD (Probability Hypothesis Density) filtering algorithm which can be used to detect the mobile visual-targets´ random appearance and disappearance in the clutter environments with high accuracy and low computational complexity. The performance analyses validates and evaluates our proposed distributed visual-PHD filtering algorithm and the data fusion scheme for the decentralized WCSN in terms of the multiple visual-target states estimation accuracy and the robustness to the interference/noise.