Mendeley  

Zotero  

Khaleghi N, Qanei Yakhdan H. Investigation of target tracking methods based on particle filter. JOTS 2024; 2 (2) :99-133
URL: http://jots.knu.edu.af/article-1-40-en.html

Target tracking requires simultaneous estimation of its position, speed and acceleration. There are different methods with different algorithms for target tracking; Particle filter is a new method to obtain posterior probability distribution function based on Bayesian theory. The particle filter algorithm is based on chain Monte Carlo methods, in which the particle representation of the probability density is used to estimate arbitrary distribution parameters.
Target tracking is the estimation of the posterior density function in each sweep for the target in the observed environment. Some things make this difficult, which include: the lack of full disclosure of the target, the existence of false targets, uncertainty in how to allocate data to the existing target, and non-linear equations and non-Gaussian noises - which makes it possible to use the Kalman filter and its families (extended and intangible Kalman ) limits.- Recently, the efficiency of Monte Carlo methods and particle filters on top of them in solving the mentioned cases has been proven. Monte Carlo methods of multi-objective tracking have replaced classical methods; But they still have room for improvement. In the conventional methods of tracking aerial targets, the distance to the target and the angle to the target side, which are a nonlinear function of the system states, are measured; But they have noise, so it is necessary to use estimation and filtering methods. The generalized Kalman filter has a good performance for dealing with nonlinear systems and Gaussian noises; But in practical implementation, we face non-Gaussian noises (Glint) that particle filters have good performance.
Particle filter performance, despite many advantages, also has disadvantages; Because with the initial selection of a large number of particles, no particle may be placed near the correct state; This weakness is known as the problem of deterioration. Re-sampling is used to reduce degradation in a standard particle filter. Re-sampling, while being vital, causes another phenomenon called poverty of samples, where the diversity among particles is lost and in the worst case, all particles fall to a point in the state space. Researchers have proposed different versions of the particle filter (auxiliary, regularized, and traceless) to improve resampling.