Unmanned Underwater Vehicle (UUV) can work in larger depth. If a towed array applied on UUV, it can detect the acoustics signals from a target in the direct arrival zone with lower transmission loss in deep water. Based on sound field characteristics in deep water, a source detection scheme with multiple UUVs is proposed. The aperture of sonar array is limited by the carrying capacity of UUV. Therefore, a passive aperture extension with sparse Bayesian learning (SBL) is proposed for a moving short array to improve the performance of direction-of-arrival estimation by constructing a larger virtual aperture. The phase correction factors and bearings of different targets are estimated simultaneously by SBL. The experiment results show that the method extends the aperture effectively and obtains higher azimuth resolution and accuracy for the direction-of-arrival (DOA) estimation than the Conventional Beamforming (CBF). The detection of an underwater target by using a towed array from UUV is possible. [Work supported by the National Natural Science Foundation of China, Grant No. U22A2012.]