Abstract: In order to improve the accuracy, efficiency and adaptability of the dense matching point cloud filtering algorithm for oblique images in complex terrain scenes, for the shortage of the traditional cloth simulation filtering algorithm, this paper proposes a cloth simulation filtering method based on virtual grid classification. Firstly, this method constructs a virtual grid with a buffer zone; And then divides the virtual grid into three types: flat ground, gentle slope, and steep slope according to the point cloud slope within its range; Finally, different types of virtual grids use the corresponding cloth parameters for cloth simulate filtering, then realize the point cloud classification of oblique image matching point cloud in complex terrain scenes. The results are compared with the single-parameter cloth simulation filtering. The experiment show that the results of the cloth simulation filtering method based on virtual grid classification are closer to the result of visual interpretation, and this method is suitable for hybrid terrain filtering.

Key words: matching point cloud, point cloud filtering, virtual grid, terrain features, cloth simulation filtering