In sparse target detection problems, it has been shown that significant gains can be achieved by adaptive sensing. We generalize previous work on adaptive sensing to (a) include targets of multiple classes with different levels of mission importance and (b) account for multiple sensor models. New optimization policies are developed to simultaneously locate, classify and estimate a sparse number of targets with limited resource budget. More specifically, three sensor models are considered: global adaptive (GA) sensor that allocates different amounts of resource to each location in the space; global uniform (GU) sensor that allocates resources uniformly across the scene; and local adaptive (LA) sensor that allocates fixed amount of resources to a subset of locations. Based on the sensor model, we propose 3 policies: GA policy that uses a GA sensor; LA policy that uses only LA sensor; and GU/LA policy that uses a mixture of GU and LA sensors. The performances of proposed policies with these sensor models are compared numerically with a baseline policy that allocates resources uniformly and an oracle policy with known target locations. Results indicate that the GA policy performs closely to the oracle policy with sufficient resources, and the GU/LA policy performs similarly to that the GA policy but it is cheaper and more easily implementable.