The grid-based Bayesian tracker employs a novel sample generation and weighting mechanism that achieves significantly improved visual tracking performance (in terms of accuracy, robustness, and computational burden) over existing active contour trackers and Monte Carlo trackers. This paper presents a method to enhance its capability in accommodating the tracking of targets in video with erratic motion, by introducing adaptation in the motion model and iterative position estimation. Tracking performance of the resulting algorithm is compared with the grid-based Bayesian tracker in the context of leukocyte tracking, UAV-based vehicle tracking, and Drosophila larva tracking to demonstrate its effectiveness in dealing with erratic target movement.