Sparsely distributed sensors or sparse arrays can be associated with high positional accuracy and capability for large area surveillance. This paper shows, both through simulations and with real measurement data, that sparse sensor systems can be used to accomplish high-performance underwater surveillance. The paper focuses on measurement setups with several passive single hydrophones placed hundreds of meters apart in a water depth of dozen of meters. By such a setup, a sound source is more likely to move inside the array, and thus near-field processing can be considered. The sensor pairs will also be located in different directions in relation to the sound source. These two prerequisites give the possibilities to high spatial resolution. Images of the coherent sound activity for different sensor pairs are formed, and every sensor pair map is summarized into a resulting map. In this way, an arbitrary number of sound sources of a target can be resolved. Good correlation results are shown even when the sound sources are at the same distance as the baseline of the sensor pairs. This is achieved by Doppler compensating for target movement and also by whitening of the cross spectra. The analyses of the measurements also show that baseline-dependent bandwidth can improve the results. In this paper, it is indicated from experimental data that two propeller sound sources could be resolved, and in another measurement setup, the engine could be separated from the propellers.