Abstract: The problem that detecting multiple targets among a finite number of regions is considered. The performance of the search strategy is not only related to the number of observations and the detection error, but also costs associated with switching across regions. We propose a multi-target search strategy based on the active hypothesis testing, which can help decision makers find all targets quickly with little switching cost. It's proved that the policy is asymptotically optimal. Meanwhile, which offers better performance in finite regime by simulations. Our policy can also be applied to anomaly detection, fraud detection and other read-world scenarios.