Sensor Data Fusion: Trends, Solutions, Applications at INFORMATIK 2011 - Informatik schafft Communities
Berlin 2011

Abstract: Multi sensor systems have performances that depend on the type of multi sensor data fusion. Multi sensor data fusion of raw detections (plots) has high performances but is seldom used in operational applications. A less efficient data fusion is preferred in the majority of operational multi sensor systems: data fusion of sensor tracks. Data fusion of sensor tracks is simpler and requires limited resources of communication and central computing but performances are limited because data fusion is unable to create system tracks without sensor tracks. Multi sensor plot data fusion has not this limitation but the huge number of plots imposes larger bandwidth sensor networks, more powerful central computers and more sophisticated data fusion algorithms. The risk and the cost of this data fusion are considered too important compared with the gain of performances. The purpose of paper is to present a modular architecture that makes possible to control the complexity of multi sensor plot data fusion. The architecture is a multi algorithm data fusion controlled by anomaly mapping and expert system. Expert system coherently manages sensors and data fusion to fulfil the missions whatever the environment and its evolution. This approach has been applied with success to the surveillance system of a French navy ship. This system has to detect, track and identify Air and Surface objects with on board active and passive sensors. Some gains are presented in comparison with the legacy system.