Focusing on the issue that different scale noise exists in denoising and smoothing of 3D point cloud data model, a bilateral filtering denoising algorithm for 3D point cloud based on noise classification was proposed. Firstly, the noise points were subdivided into the large-scale and the small-scale noise, and the large-scale noise was removed with statistical filtering and radius filtering. Secondly, the curvature of the 3D point cloud data was estimated, and the bilateral filter was improved to enhance the robustness and security. Finally, the small-scale noise was smoothed with the improved bilateral filter to achieve the smoothing and denoising of 3D point clouds. Compared with the algorithms simply based on bilateral filtering or Fleishman bilateral filtering, the smoothing average error index of 3D point cloud data model obtained by the proposed method respectively decreased by 50.53% and 21.67%. The experimental results show that the proposed algorithm increases the efficiency of calculation by scale subdivion of noise points, and avoids excessive smoothing and detail distortion, which can better maintain the geometric characteristics of the model.