Abstract: Side scan sonar is one of the commonly methods for underwater riprap detection, which has the advantages of high lateral resolution and high efficiency. However, qualitative identification of underwater media according to the relative value of acoustic scattering intensity may have big error. In order to improve the identification accuracy, taking the underwater riprap in Baoding polder as the research object, wavelet analysis was used to denoise the side scan sonar data. Then Lambert law was used to eliminate the effect of grazing angle, the acoustic scattering intensity was related to the type of underwater media directly. Finally cluster analysis was used to classify underwater media. The research results showed that the identification ability of this method was obviously higher than that of clustering analysis based on the original side scan sonar data, and could accurately classify the distribution range of riprap, and greatly improved the identification ability of underwater riprap.