Speaker
Description
This paper presents a Bayesian method for recursive fusion of prior and measurement information aimed at improving accuracy in video-based measurements. The approach is based on a probabilistic formulation of the estimation problem, where measurements are modeled as random variables with additive noise, and the system state is described by a prior distribution.
By applying a recursive Bayesian scheme, an algorithm for state estimate updating is derived, in which the estimate is formed as a weighted combination of prior information and the current measurement, with weights determined by the corresponding variances.
It is analytically shown that, under Gaussian assumptions, the proposed approach leads to a reduction in the dispersion of the estimate and consequently to improved accuracy in noisy measurement conditions. The theoretical results are validated through simulation studies under varying levels of measurement noise, demonstrating the robustness of the method under changing measurement conditions.