Deploying unmanned aerial vehicles (UAVs) as relay nodes to connect ships on the sea with base stations (BS) on the ground is a promising solution in future maritime communications. In this paper, we propose a low-delay UAV deployment(LDUD) algorithm to cover all ships on a given area of the sea, aims to minimize the number of UAVs and the data transmission delay from ships to BS. The LDUD algorithm includes two sub-algorithms: the first one is a coverage algorithm based on convex hull (CACH) and the second one is a network formation game combined with virtual force (NFVF). In the first sub-algorithm, a UAV network with minimum UAVs is deployed to connect all ships with BS. This algorithm maintains the connectivity between UAVs and minimize the number of required UAVs at same time. The second sub-algorithm finds the optimal positions of UAVs and associations between UAVs, ships and BS for the lowest data transmission delay. A virtual force model is designed to adjust UAVs’ positions for lower transmission delay. Moreover, a network formation game model is developed to form a multi-hop network connecting ships to BS. Each ship and UAV selects the path connecting itself to BS for maximum utility. Simulation results shows that, the LDUD algorithm requires the minimum number of UAVs when the number of ships is more than 120. For the transmission delay from ships to BS, our algorithm improves more than 25% than other algorithms.