Three-dimensional reconstruction method based on gaussian point information redistribution and geometric structure constraint

Abstract

The application relates to a three-dimensional reconstruction method based on Gaussian point information redistribution and geometric structure constraint. The method comprises the following steps: acquiring target scene image data, extracting a visible Gaussian point attribute parameter set based on a 3D Gaussian sputtering framework; randomly removing part of the Gaussian points and determining a neighborhood point set, constructing an approximate model to obtain a gradient term set; constructing a correlation matrix and calculating an information distribution coefficient, compensating the opacity and color information of the removed points to the neighborhood points to update the parameter set; constructing a total loss function containing a basic reconstruction, local direction consistency and main direction alignment loss, iteratively optimizing the parameters; dynamically adjusting the Gaussian point shielding rate and modifying the opacity correlation parameter, maintaining the physical constraint, repeating the information redistribution to complete the training, and realizing high-quality three-dimensional reconstruction. The method can improve the clarity of the geometric structure.

Images