A Novel Algorithm Based on Particle Filter for Advanced Target Tracking

In this paper, we present a novel particle filter-based algorithm that incorporates MCMC Bayesian Model Selection and Markov Chain Monte Carlo methodologies for target tracking applications. The algorithm is designed to handle single-target tracking, multi-target tracking, and video-based target localization with enhanced capabilities for solving nonlinear problems. Implementation-wise, the algorithm employs sequential importance sampling with systematic resampling techniques and utilizes Metropolis-Hastings within Gibbs sampling for parameter estimation. Compared to traditional Kalman Filter, Extended Kalman Filter (EKF), and Unscented Kalman Filter (UKF) approaches, this method demonstrates significantly better performance in handling complex nonlinear systems. The core algorithm structure includes particle initialization, importance sampling based on proposal distributions, weight computation using likelihood functions, and effective sample size monitoring for resampling triggers. We have chosen to share this valuable technical innovation hoping it will assist others in addressing similar challenges, while also seeking collaborative support to advance this technology further through community contributions and improvements.