Sequential Monte Carlo methods for multiple target tracking and data fusion
read more
Citations
Object tracking: A survey
Kernel-based object tracking
Multitarget Bayes filtering via first-order multitarget moments
The Gaussian Mixture Probability Hypothesis Density Filter
Particle filtering
References
Novel approach to nonlinear/non-Gaussian Bayesian state estimation
Sampling-Based Approaches to Calculating Marginal Densities
Sampling-based approaches to calculating marginal densities
C ONDENSATION —Conditional Density Propagation forVisual Tracking
Related Papers (5)
Frequently Asked Questions (11)
Q2. What future works have the authors mentioned in the paper "Sequential monte carlo methods for multiple target tracking and data fusion" ?
This will be addressed in future studies.
Q3. How long does it take to compute the MTPF estimate of three targets?
With a Pentium III 863 MHz, particles, a burn-in period , and a total amount of iterations in the Gibbs sampler, it takes around 1 s per time step to compute the MTPF estimate of three targets with bearings-only measurements.
Q4. How many MTPF estimates can be computed with a Pentium III?
With a Pentium III 863 MHz, particles, a burn-in period , and a total amount of iterations in the Gibbs sampler, it takes around 840 ms per time step to compute the MTPF estimates of two targets with bearings measurements and 20% of range measurements.
Q5. What is the probability of the two objects overlapping?
a probabilistic exclusion principle is integrated to the likelihood measurement to penalize the hypotheses with the two objects overlapping.
Q6. What is the effect of updating the particle weights according to the measurements?
In the updating step of particle filtering, the weights of the particles are updated according to the measurements, but the predicted positions are not modified.
Q7. What is the rate of convergence of the average mean square error?
In the context of sequential Monte Carlo methods [10] that cover most of the particle filtering methods proposed in the last few years, the convergence and the rate of convergence of order of the average mean square error is proved.
Q8. What is the main method used to estimate the parameters of the MTPF?
Two main ways have been found in the literature to estimate the parameters of this model: the expectation maximization (EM) method (and its stochastic version the SEM algorithm [5]) and the data augmentation method.
Q9. How is the probability of the observations conditioned by the th particle obtained?
The likelihood of the observations conditioned by the th particle is readily obtained as(24)There is no strong limitation on the use of the particle filter for multireceiver and multitarget tracking: the MRMTPF is obtained from the MTPF by replacing the likelihood functionsby the functions .
Q10. What is the standard deviation for the th component of the MTPF?
At each time , the bias and the standard deviation for the th component of are defined bybiasstd (30)To avoid the compensation of elementary bias of opposite signs, the authors average the absolute values of the bias bias .
Q11. How do the authors compute the MTPF estimate for particles?
Such estimates are obtained by applying the prediction step and by giving constant weights to the particles instead of computing them given the measurements.