Abstract: Aiming at the problem of shallow seabed layered structure and geoacoustic parameter inversion with ship radiated noise as the reference sound source, a shallow multi-layered seabed geoacoustic parameter inversion method based on Bayesian theory is studied. In the inversion, the line spectrum component of the ship radiated noise is the research object, and then the nonlinear Bayesian inversion method is used to invert the parameters of the seabed structure, the wave speed and density in the layer, and the uncertainty of the inversion result is analyed. The Maximum A Posterior (MAP) estimate and the marginal probability distribution of the inversion result are calculated by the Optimization Simulated Annealing (OSA) and Metropolis-Hastings Sample (MHS) method in the prior interval of each parameter, and the best seabed layering is determined according to the Bayesian Information Criterion (BIC). The sea experiments show that the seabed layered structure and geoacoustic parameters are obtained by inversion according to this method, and the error between the calculated sound pressure field and the measured ship radiated noise transmission loss not exceed 10%, and the inversion result can accurately represent the seabed characteristics of the experimental sea area. The uncertainty of the inversion result shows that the uncertainty of the compression wave speed, shear wave speed and density is smaller, and it is more sensitive to changes in the sound pressure field, and the inversion result is more effective and accurate.