In this paper, an adaptive coarse-to-fine particle filter is proposed to improve the performance of the previous coarse-to-fine particle filter approach which processes the measurement data twice for better tracking. Analysis on the characteristics of coarse-to-fine particle filter is conducted to derive a proper structure for the adaptive filter. Then, the position variance information of particles are used for convergence check. The number of particles is adaptively controlled, depending on the convergence performance, where the range of the number of particles is a design parameter. Monte Carlo simulation is conducted to compare the performance of the proposed adaptive particle filter with the previous coarse-to-fine particle filter with a fixed population size, in terms of tracking success rate and computation time.