TPE 2013 - Special Session on Tracking Performance Evaluation

One common area of information fusion is object tracking and identification. Characterizing the performance of the target tracking solution is based on the data (e.g. sensor types, uncertainty), scenario (e.g. objects, context, and environment), and algorithms (e.g. linear versus non-linear filtering, Bayesian and non-Bayesian). It is well known that if an information fusion system is to provide the track and identification of objects (Level 1 Fusion); then the representation for tracking is in spatial (distance) and temporal (time); while identification is in probability of detection and probability of false alarms. When a user (Level 5 fusion) requests uncertainty reduction for tracking systems, then they desire the lowest false alarm rate of the target ID with the minimum mean squared error of position. Given the advances in tracking systems and increased requirements (track lifetime, target numbers, object characterization), there is a need to determine how to represent, evaluate, and characterize successful methods of performance evaluation.