Pre-training generative three-dimensional grid reconstruction method based on oblique photography data

The invention discloses a pre-training generative three-dimensional grid reconstruction method based on oblique photography data, which relates to the technical field of three-dimensional reconstruction and comprises the following steps: S1, constructing an initial three-dimensional network reconstruction model; s2, obtaining a training data set, importing an initial three-dimensional network reconstruction model, and carrying out training optimization on the initial three-dimensional network reconstruction model to obtain an optimized three-dimensional network reconstruction model; s3, acquiring oblique photography data to be reconstructed, and importing the optimized three-dimensional network reconstruction model to generate a reconstructed three-dimensional network; a three-dimensional grid is used as a new object generated by a large model, and three-dimensional patch display expression with clear edges and a complete topological structure is achieved. The method for generating a complete continuous objec