Markov chain Monte Carlo data association for general multiple-target tracking problems
read more
Citations
Simple online and realtime tracking
Simple Online and Realtime Tracking
Stable multi-target tracking in real-time surveillance video
BLOG: Probabilistic Models with Unknown Objects
BLOG: probabilistic models with unknown objects
References
Markov Chain Monte Carlo in Practice
The complexity of computing the permanent
An algorithm for tracking multiple targets
Related Papers (5)
Frequently Asked Questions (17)
Q2. What are the MCMC techniques used to solve?
MCMC techniques have been applied to the complex probability distribution integration problems, counting problems such as #P-complete problems, and combinatorial optimization problems [7], [2].
Q3. What is the number of objects arising at each time over R?
The number of objects arising at each time over R has a Poisson distribution with a parameter (λbV ) where λb is the birth rate of new objects per unit time, per unit volume.
Q4. What are the types of moves that are used in the MCMC sampler?
They are1) birth/death move pair; 2) split/merge move pair; 3) extension/reduction move pair; 4) track update move; and 5) track switch move.
Q5. What is the criterion used to compare multiple target tracking algorithms?
Since the number of targets is not fixed, it is difficult to compare algorithms using a standard criterion such as the residual mean square error.
Q6. What is the MAP estimate under the data-oriented approach?
Under the data-oriented approach, the multiple target tracking problem is to partition the observations such thatthe posterior is maximized, i.e., the maximum a posteriori (MAP) estimate.
Q7. What is the value of xt without fixing the number of tracks?
Since the size of xt = (x1t T , . . . , xKtT )T depends on the number of tracks K, the estimation of xt without fixing the number of tracks is not meaningful.
Q8. What is the goal of the inverse optimization approach?
Now under the data-oriented, combinatorial optimization approach, their goal is to find a partition of observations such that P (ω|Y ) is maximized.
Q9. What is the way to estimate the number of states?
under the Bayesian approach, if a single set of state estimation is required, the authors might first estimate the most likely number of targets and then estimate the expected values of states given the estimated number of targets.
Q10. What is the main advantage of the MCMCDA algorithm?
Another noticeable benefit of the MCMCDA algorithm is that its running time can be regulated by the number of samples and the number of observations but the running time of MHT depends on the complexity of the problem instance and is not predictable in advance.
Q11. What is the use of this neighborhood tree?
The use of this neighborhood tree makes the algorithm more scalable since distant observations will be considered separately and makes the computations of the proposal distribution easier.
Q12. What is the conditional observation covariance for the track?
For each track τ ∈ ω, the authors apply the Kalman filter to estimate the states x̄t(τ) and covariances Bt(τ), where Bt(τ) = CP̄t(τ)CT + R is the conditional observation covariance at time t for the track τ .
Q13. What is the MAP estimate under the Bayesian approach?
Under the Bayesian approach, if the authors are given a function defined on Ω, the collection of all partitions of observations (see below for its definition), the authors seek the expected value of the function given the observations.
Q14. what is the expected value of xt given y1?
P̄t be the covariance of xt given y1, . . . , yt−1; x̂t be the expected value of xt given y1, . . . , yt; and P̂t be the covariance of xt given y1, . . . , yt.
Q15. What is the effect of pruning on the performance of MCMCDA?
Although the maximum number of hypotheses of 1000 per group is a large number, with increasing number of tracks, the performance of MHT deteriorates as the optimality is compromised by pruning.
Q16. How many samples are used for MCMCDA?
The results for MCMCDA are the average values over 10 repeated runs and the initial state is initialized with the greedy algorithm and 10,000 samples are used.
Q17. What are the two new metrics to measure the effectiveness of each data association algorithm?
The two new metrics the authors will be using are the normalized correct associations (NCA) and incorrect-to-correct association ratio (ICAR):NCA(ω) = |CA(ω)| |SA(ω∗)|(4)ICAR(ω) = |SA(ω)| − |CA(ω)||CA(ω)| .