Building entity model lightweight method considering visual subject

The invention relates to a building entity model lightweight method considering a visual subject, and the method comprises the following specific steps: carrying out the orthographic projection, height layering and boundary extraction of a triangular surface, and recognizing a roof, a wall body, a water tank and a decoration according to a spatial inclusion relation, a height relation and an area ratio relation; generating a five-level LOD model for a single building by taking a roof as a main body structure and combining with other parts; loading an original model on a three-dimensional platform, generating a model texture in a scene drawing mode, and reconstructing texture coordinates for each level of LOD model according to a principle that the geometric position of the model is close to a scene camera; all models are traversed, a scene bounding box is obtained, scenes are divided and tiles are organized according to a quadtree, the tiles where the models are located are determined according to the center