Abstract: The geoacoustic parameters can hardly be inverted using far-field controlled-source due to the sound propagation in deep water. To aim at this problem, we developed a near-field Matched-Field Inversion (MFI) method using tow-ship noise data in deep water. First, the multipath behaviour of tow-ship noise acquired via towed array was investigated by performing near-field beamforming. Then, the multi-parameter geoacoustic model was established to calculate replica fields using wavenumber integration approach, and multi-frequency MFI was performed using genetic algorithm. Finally, The Posterior Probability Density (PPD) of the parameters was investigated using Monte Carlo method. The tow-ship noise near-field MFI method was verified by simulations and real trial data processing. The results indicated that tow-ship noise mainly came from the seabed reflection path in deep water, and multiple parameters such as water depth, source range, array depth, and sediment thickness were inverted by the proposed method. And a more accurate result of sediment thickness and other parameters was achieved by performing multi-frequency MFI. The wideband Matched Field Processing (MFP) ambiguity surface based on the inverted acoustic model could give the precise location of tow-ship radiated noise source.