In this work, we propose a physical approximation scheme - Random Opportunistic and Selective Exploration (ROSE) - for aerial localization of survivors by using a collaborative swarm of IoT-based unmanned aerial vehicles (UAVs). The UAV swarm performs a simultaneous multi-pronged search of a given zone by dividing the search region among the swarm members. This multi-pronged search strategy speeds-up the search, and the division of search areas among the swarm members avoids redundant exploration of an already explored location. As the communication range of the member UAVs is limited, the swarm members communicate opportunistically among themselves to share the information of the visited sites. We formulate the various probabilities associated with opportunistic communication of these aerial IoT nodes and simulate the performance of the approximation algorithm based on these formulations. Simulation results of the proposed approach successfully locate 100% of the ground targets within an acceptable time-frame, and out-performs established searching schemes such as the truncated Levy walk, frontier-based search, and sweep search.