Method and medium for parametric 3D model reverse modeling based on large language models

Abstract

The present disclosure discloses a method and medium for parametric 3D model reverse modeling based on a large language model. The method includes: acquiring an intrinsic parameter matrix of a camera and extrinsic parameter matrices of a laser radar and the camera; triggering a synchronous reset operation of the laser radar and the camera when a time deviation of data output by the laser radar and the camera exceeds a preset threshold; inputting image data into a multimodal large model to identify building components in the image data and outputting labeled segmentation masks; performing distortion correction on pixel coordinates of the segmentation masks and converting them into a LiDAR coordinate system using the intrinsic and extrinsic parameter matrices; extracting point cloud clusters from point cloud data and performing Euclidean cluster segmentation on the point cloud clusters; assigning semantic labels to the point cloud clusters and inputting them into the multimodal large model to extract geometric parameters of components and complete attribute information; and generating parametric BIM components according to the geometric parameters, the attribute information of the components, and API specifications of BIM software. The present disclosure can improve modeling accuracy.