To solve the low-classification accuracy problems of urban point clouds in complex environments, we propose a classification method based on multiscale adaptive features herein. Firstly, the classical geometric statistical features and point histogram features are combined; then, the combined feature set is used for classification basis. Random forest is then used to assess the importance of the features and adaptively select important feature sets. Finally, the point clouds are classified based on these multiscale adaptive features. Experimental results reveal that this method can achieve a high-accuracy classification for point clouds in urban areas. The proposed method can be applied to the classification of point cloud data with different resolutions at arbitrary scale.