Exploitation of track accuracy information in fusion technologies for radar target classification using Dempster-Shafer Rules

The surveillance of the littoral is required by applications in the defence, protection and security area. One might think about anti asymmetric warfare, harbour and coastal surveillance or the prevention of smuggling, illegal fishing, illegal immigration or acts of piracy.To establish situation awareness ground, sea and air targets must be detected and tracked in the littoral simultaneously. Also the detailed classification of targets is of extraordinary importance, e.g. persons vs. vehicles, helicopters vs. planes or buoys vs. ships. This classification can be overtaken by an operator, who listens to the Doppler sound of a target. Unfortunately, obligated to the classification issue an operator gets distracted from the tactical surveillance task. Further, an operator is only able to classify a limited number of targets. Hence, automatic target recognition is an important issue for radar systems applied to the littoral surveillance. Finally, automatic target recognition offers also a synergy with the multi target tracking of such a radar system. In this paper 2 Dempster-Shafer (DS) based fusion methods will be described. Both use tracks and track accuracy information to fuse with Doppler based classified targets, in order to provide a robust classification technique for distinguishes between different kinds of targets. The first classification technique uses a hierarchical tree structured decision method, integrated in a track-based classifier. The second classification technique uses a non-hierarchical decision method also integrated in a track based classifier. In this paper both kinds of DS methods will be compared. The results will be discussed especially with respect to the following performance criteria: track accuracy, classification confusion matrix, targets hit rate, targets rejection probabilities, DS topology requirements, convergence reliability, training duration and generalization efficiency.