Abstract: We propose a linear geometric coding method for efficient transmission of complicated 3D building models.Firstly,the building model is converted into volumetric representation,and the octree index of the original model is established.Then the geometric information of the building model is linearly encoded and compressed according to the expansion patterns of the spatial index structure between levels and their empirical probabilities.Finally,the spatial density field is constructed,and the linear data stream is decompressed and reconstructed based on the improved MC algorithm.Experimental results show that the proposed method is not constrained by the topology of the original models' surfaces,and it could greatly reduce the data volume and the response time in network transmission of complicated 3D building models with preservation of visual quality.