Method and system for distributed multiple target tracking
View Patent ↗A method and system for distributed tracking of multiple targets is disclosed. Multiple targets to be tracked by a plurality of trackers are detected in a frame. The motion state variable of each of the plurality of trackers is calculated in the E-step of a variational Expectation-Maximization algorithm. Further, the data association variable of each of the plurality of trackers is calculated in the M-step of the algorithm. Depending on the motion state variable and the data association variable, the multiple targets are tracked.
1. A method for distributed tracking of multiple targets using a variational Expectation-Maximization (EM) algorithm in a surveillance system, the surveillance system comprising a plurality of trackers tracking the multiple targets, the method comprising: estimating a motion state variable of each of the plurality of trackers, the estimation being performed in the E-step of the variational EM algorithm; estimating a data association variable of each of the plurality of trackers based on a graph-based discrete optimization algorithm, the estimation being performed in the M-step of the variational EM algorithm;
tracking the multiple targets based on the motion state variable and the data association variable of each of the plurality of trackers; and
wherein the data association variable of each of the plurality of trackers is estimated as a maximum a posteriori (MAP) estimate in the M-step of the variational EM algorithm.
2. The method of claim 1 further comprising initializing the plurality of trackers to track the multiple targets, wherein the association between the plurality of trackers and the multiple targets is based on a one-to-one mapping.
3. The method of claim 1 , wherein the motion state variable of each of the plurality of trackers is estimated as a probability distribution function in the E-step of the variational EM algorithm.
4. The method of claim 1 , wherein the E-step and the M-step of the variational EM algorithm are repeated for a predefined number of times to converge the values of the motion state variable and the data association variable in a predefined range.
5. The method of claim 1 , wherein the distributed tracking is performed through decomposing at least one of the multiple targets into a predefined number of parts.
6. The method of claim 1 , wherein the motion state variable of each of the plurality of trackers comprises information related to the motion state of a corresponding target in 3-dimensional space.
7. A distributed tracking system for distributed tracking of multiple targets using a variational Expectation-Maximization (EM) algorithm in a surveillance system, the system comprising:
a processor including a tracking module comprising a plurality of trackers for tracking the multiple targets, each of the plurality of trackers being associated with a motion state variable and a data association variable;
wherein each of the plurality of trackers estimates the motion state variable in the E-step of the variational EM algorithm;
wherein each of the plurality of trackers estimates a data association variable based on a graph-based discrete optimization formulation, the estimation being performed in the M-step of the variational EM algorithm; and
wherein the data association variable of each of the plurality of trackers is estimated as a maximum a posteriori (MAP) estimate in the M-step of the variational EM algorithm.
8. The distributed tracking system of claim 7 , wherein the tracking module initializes the plurality of trackers to track the multiple targets, wherein the association between the plurality of trackers and the multiple targets is based on a one-to-one mapping.
9. The distributed tracking system of claim 7 , wherein the motion state variable of each of the plurality of trackers is estimated as a probability distribution in the E-step of the variational EM algorithm.
10. The distributed tracking system of claim 7 , wherein the E-step and the M-step of the variational EM algorithm are repeated for a predefined number of times to converge the values of the motion state variable and the data association variable in a predefined range.
11. The distributed tracking system of claim 7 , wherein the distributed tracking is performed through decomposing at least one of the multiple targets into a predefined number of parts.
12. The distributed tracking system of claim 7 , wherein the motion state variable of each of the plurality of trackers comprises information related to the motion state of a corresponding target in 3-dimensional space.