Building facade information refers to the spatial distribution and attribute information of the contact surface between buildings and external space. How to extract building facade information from point cloud data is a hot and difficult problem in point cloud data processing. In order to solve the problems of single evaluation standard and weak adaptability of traditional grid density algorithm in building facade point cloud extraction, this paper analyzed the local and overall spatial characteristics such as elevation distribution, projection density and normal vector distribution of various typical surface feature point clouds in the construction area, and constructed a multi-level semantic feature descriptor composed of point cloud single point semantics, grid semantics and regional semantics. Based on this descriptor and the reasonable threshold which was set according to the semantic characteristics of building facade point cloud at different levels, a multi-level semantic feature extraction method was proposed to extract the building facade point cloud accurately layer by layer. The experimental results show that this algorithm can be used to quickly and accurately extract the building facades of low, high buildings and super high buildings from point clouds. Overall, this algorithm achieves a high precision, a high efficiency and a good adaptability.