Distributed Visual Sensor Network Fusion

Visual sensor, distributed network, metadata management, moving objects, spatio-temporal data, Kalman filter, sensor fusion, object tracking, large area surveillance system.

The poster deals with a framework for distributed visual sensor network metadata management system. It is assumed that data coming from many cameras is annotated using computer vision modules to produce metadata representing moving objects in their states. The data is supposed to be noisy, uncertain and some states might be missing. Firstly, here is described the spatio-temporal data cleaning using Kalman filter. Secondly, it copes with many visual sensors fusion and persistent object tracking within a large area. Thirdly, it describes the data and architecture model.