Comparing future situation pictures

In this paper, we suggest a practical method for measuring the difference between situation pictures obtained independently from each other. In previous work, we have considered the output of two independent tracking methods, while we here consider the output from a threat analysis/prediction system at two time instances. At two instances in time, a hypothesized threat analysis/prediction system generates future situation pictures, describing the world at the same future point in time. Qualitative and quantitative differences between the two future situation pictures are then detected. The detected difference map serves as an indication of the amount of unpredictability in the world; or alternatively, of the ability of the threat analysis/prediction system to model different areas in the future situation picture accurately. This could either be used as feedback in training a learning threat analysis/prediction system, or to inform a human user about the reliability of the future situation picture.