Abstract: To meet the accuracy requirements for intelligent identification of rock mass structural planes from point cloud data,this study proposes a vegetation filtering algorithm that integrates spatial position and color information of point clouds. First,based on spatial differences between rock mass and vegetation,point cloud data were gridded and subjected to adaptive plane fitting to select vegetation seed points. Spatial processing allowed plane fitting for slopes other than vertical ones,making the method well-suited to high-angle rock slopes. Next,using the color information of vegetation seed points,the grid cells were traversed to filter out vegetation noise. Finally,a retention safeguard mechanism based on normal distribution was implemented to prevent excessive filtering of rock mass features. Applying the algorithm to steep rocky slopes of Tiantai Mountain,the filtering performance was evaluated through sensitivity analysis of parameters,comparison with existing filtering methods,and assessment of structural plane identification before and after filtering. The results show that the proposed algorithm significantly improves vegetation filtering,with a type-Ⅱ error of only 3.74%,substantially lower than errors from slope-based,fitting-based,and red-green difference index filtering methods. At the same time,the retention rate of gridded rock surface features reached 100%,preserving rock mass structural characteristics effectively. The original point cloud identified only one dominant structural plane(97% coverage),failing to delineate additional effective planes,whereas the filtered point cloud identified two main structural planes covering 58% and 36% of the area,respectively. This indicates that the filtered point cloud enables more effective identification of rock mass structural planes. The study provides a new approach for vegetation filtering on high-steep rocky slopes and significantly improves the accuracy of rock mass structural plane identification.