Gaussian Mixture Probability Hypothesis Density Filter

This is an implementation of a gaussian mixture probability hypothesis density filter (GM-PHD) for a simulated tracking problem. The problem specification is given in the above paper - in summary, two targets move through the environment, there is a lot of clutter on the measurement, and about halfway through a third target spawns off one of the two targets. I made a few changes, either because I couldn"t understand how Vo&Ma did it, or because I wanted to make it closer to my target problem. I extended the measurement vector to include both target position AND velocity (the filter they describe tracks position and velocity but only observes position). Velocity is observed as just being dx/dt, change in position over time, between this new observation and the previous position of this target. Targets are either birthed or spawned depending on which initialisation weight function would give a higher weight; they are given the appropriate initialisation covaria