In order to eliminate the ambiguity in the land cover classification of LiDAR point cloud by using the image texture, the texture feature of point cloud based on the searching structure of KD tree and the gray level co-occurrence matrix were proposed, which represents the distribution of attribute values of points and their surrounding neighborhood points. The influence of the parameters, such as search neighborhood, moving step and gray-level, on the texture features of point cloud was analyzed. Using the support vector machine classification method, it was verified that the texture feature of point cloud can effectively assist the elevation and intensity to improve the results of the land cover classification. In addition, the results demonstrated that the land cover classification under the constraint of the texture features of point cloud has higher accuracy than that under the constraint of the rasterbased image texture features, and the texture features of point cloud perform outstandingly in distinguishing tiny land objects and separating the water and land. These excellent characteristics of the texture features of point cloud can contribute significantly to the refined classification of coastal LiDAR data, the construction of highprecision DEM in coastal zone and the extraction of coastlines.