Real-time 3D acoustic imaging underwater is becoming increasingly important, as 3D geometry information of underwater objects is required for autonomous underwater vehicles to avoid obstacles. A major problem for real-time imaging is the time required to process the data associated with large planar arrays and the hardware cost of uniform 2D receiving arrays. The solutions to the problem are to reduce the number of elements in the receiving array and find fast beamforming techniques for 3D imaging. In this work, we propose an efficient cross-array beamforming using spatial cross correlation of azimuth and elevation beams for underwater real-time 3D imaging. The simulation result demonstrated that the proposed method could reconstruct a 3d image of an underwater target with an inferior but acceptable resolution. The proposed method is twenty-two times faster than the conventional delay and sum beamforming for 120 x 120 beams.