This paper presents a novel adaptive sampling method using intelligent UAVs in battlefields to help soldiers with awareness of environments. A UAV can perform as a robotic wingman in soldier formations, compensating for that cannot be scouted by soldiers, even being exposed to enemy fire. With portable size, the UAV is easily carried and flown for scouting tasks anytime. The flexibility of UAVs makes it possible to collect measurements sequentially. Each measurement is adaptively designed and determined from the Bayesian perspective to increase the fidelity of battlefields. Wavelet structure is considered to optimize measurement projections to substantially reduce the number of measurements based on compressive sensing framework. More specifically, each measurement is optimized by maximizing the posterior variance inferred from existing informative data. A motion planning algorithm for UAVs is designed based on the distribution of optimal measurements, striking a balance between moving cost and measurement value. Simulation results and future experimental environments are presented at last.