Unmanned Aerial Vehicles have emerged as the optimal solution for target tracking due to their low cost and high maneuverability. We study the problem of target tracking in partially observable adversarial environments. Building on previous research, we develop a data fusion algorithm to optimize target tracking. To enable a single UAV to track a hard-to-observe target with limited sensor capabilities, we construct a target environment tracking model along with its corresponding reward function. In the algorithm design, we incorporate the Exp4-IX algorithm from the framework of an adversarial multi-armed bandit with advice, and we prove that the regret bound of this algorithm exhibits sub-linear growth. In numerical experiments, the Exp4-IX algorithm integrates the Previous Position algorithm, Particle Filtering algorithm, and Trajectory Fitting algorithm, and is benchmarked against the Average Fusion algorithm. The results demonstrate its effectiveness in online fusion for smooth trajectory scenarios. This integration allows the UAV to predict the target’s position with greater accuracy compared to other algorithms.