Given the low resolution of classification based on angular response features and the poor anti-noise performance of classification based on pixel statistical features in traditional sediment classification with multibeam backscatter images，this paper proposes an object-based sediment classification method. For this purpose，a general backscattering processing flow was adopted to obtain geocoded backscatter images after radiation distortion correction. Then，the simple linear iterative clustering（SLIC）algorithm was employed to segment the backscatter images into internally uniform and well-defined object blocks. Finally，the statistical features of each object block were extracted to construct feature vectors，and sediment classification was performed using K-Means ++ as a classifier. The proposed method improves the reliability of sediment classification to a classification accuracy of 86.96%.