An underwater moving target detection algorithm based on water surface characteristic wave is proposed to overcome the shortage of effective detection methods for photoelectric polarization imaging modes. Based on the wind-induced gravity wave model and the water surface characteristic wave model of an underwater moving target, the mixed wave images under different states are simulated and used for the research of the algorithm. The algorithm uses the Radon transform to extract the linear wave characteristic, and average filter and standardization are employed to preprocess images, thereby eliminating the adverse effect of Radon transform on detection. The double-neighborhood adaptive threshold method is employed to extract partial peak points in Radon transform domain. The algorithm employs continuous wavelet transform to extract features and support vector machine to judge the peak points, thereby improving the detection accuracy. The experimental result shows that the algorithm is feasible for characteristic wave detection, which also provides a new way for underwater moving target detection.