For the terrain classification problem of light detection and ranging (LiDAR) point cloud data in complex scenes, a deep neural network model based on multi-scale features and PointNet is proposed. The method improves the ability of PointNet to extract local features and realizes the automatic classification of LiDAR point cloud under complex scenes. Multi-scale network on the basis of PointNet network is added to extract the local features of points, and the local features of different scale points are formed into a multi-dimensional feature through the full connection layer, and combined with the global features extracted by PointNet, the score of each point class is returned to complete the point cloud classification label. The proposed deep neural network model is verified by using the Semantic three-dimensional dataset and the Vaihingen dataset provided by ISPRS. The research results show that the proposed algorithm achieves higher classification accuracy compared with other neural networks for point cloud classification.