Optimization method based on non-convex non-smooth second-order regular term and sparse fidelity term

Abstract

The invention discloses an optimization method based on a non-convex non-smooth second-order regular term and a sparse fidelity term, and the method comprises the steps: proposing an optimization model comprising the non-convex non-smooth second-order regular term and the sparse fidelity term, and carrying out the optimization of a surface normal vector of a geometric grid; and updating the vertexposition according to the optimized surface normal vector so as to optimize the geometric grid. Regular terms in the model can retain geometrical characteristics of an original grid to the maximum extent, and sparse fidelity terms can enable the model to process various types of noise, such as impulse noise, Gaussian noise and mixed noise. Compared with the existing geometric grid model quality optimization technology, the method provided by the invention can significantly improve the visualization quality of the geometric grid model, and has the advantages of high calculation speed and highcalculation precision.

Description

technical field [0001] The invention relates to the field of geometric modeling, and more specifically relates to an optimization method based on non-convex and non-smooth second-order regular terms and sparse fidelity terms, which is used for optimizing the quality of three-dimensional grids. Background technique [0002] At present, geometric grid, as the most commonly used 3D data, has been widely used in the fields of surveying and mapping, geology and geography. In recent years, the method of scanning and acquiring grids by laser scanners has gradually become popular. However, due to the influence of the accuracy of acquisition instruments and the external environment, the acquired point cloud data will inevitably contain a lot of redundancy and noise. The resulting geometric grid necessarily also contains noise. These noises will not only seriously affect the visualization quality of the model, but also cause errors in subsequent geometric processing operations. There...

AI Technical Summary

Problems solved by technology

Although the isotropic filtering method can effectively optimize the quality of the geometric grid model, it will blur the geometric features of the model at the same time; the anisotropic filtering method will also make the geometrical characteristics of the model blurred when dealing with high-frequency noise. features become blurred

[0005] 2. Grid optimization algorithm based on compressive sensing. The basic idea of ​​this type of method is to optimize the quality of the geometric grid model by optimizing the global energy function, although using this method can well maintain the geometric characteristics of the model while optimizing the quality , but there will be a problem of staircase effect in the smooth region

[0006] 3. Based on machine learning, although this method can effectively optimize the quality of the model without considering the geometric characteristics and noise patterns of the underlying surface, the performance of this method depends on the integrity of the training data set, and it is very expensive. Time

Images