In the marine area, the all-weather atmospheric optical turbulence profile within the boundary layer is obtained by inversion by coherent Doppler laser wind profile radar. The machine learning approach and back propagation neural network is used to train the optical turbulence estimation model of the boundary layer. The conventional meteorological parameters measured by sounding are used as model input parameters to estimate the atmospheric optical turbulence profile in the boundary layer on different days and different times, and compared with the measured values. The error analysis shows that the root mean square errors of the estimated optical turbulence profiles at day and night are 0.4332 and 0.5626, respectively, and the correlation coefficients are 0.8899 and 0.7673, respectively. The research proves that the optical turbulence profile inversion of coherent Doppler laser wind profile radar can realize the function of all-weather estimation of optical turbulence profile of the oceanic atmospheric boundary layer through the neural network model, and the effect is good. It has great engineering reference significance in photoelectric engineering and astronomical site selection.