Hi all,
This pseudo-code description of Bayesian tracking (without a motion model) might help some people:
P(o|m) = observation probability distribution
P(m) = initial prior
While not done {
Calculate P(m|o) using P(m), P(o|m) and Bayes' rule (and trick where we normalise instead of calculating P(o))
P(m) = P(m|o) # move on to next timestep
}
Will