CN-121999145-A - Gaussian point information redistribution method

CN121999145ACN 121999145 ACN121999145 ACN 121999145ACN-121999145-A

Abstract

The application relates to a Gaussian point information reassignment method. The method comprises the steps of obtaining a visible Gaussian point set covering pixel points and extracting an attribute parameter set of the visible Gaussian point set, randomly eliminating part of Gaussian points to serve as points to be eliminated, determining a neighborhood Gaussian point set of the points, constructing an approximate model of pixel point colors on opacity, calculating gradient items of colors on the opacity and forming a gradient item set, constructing an incidence matrix of the points to be eliminated and the neighborhood points based on the gradient items, calculating an information distribution coefficient by combining distribution factors, compensating the opacity and color information of the points to be eliminated to the neighborhood points according to the coefficient, updating the attribute parameters of the points, finally progressively adjusting shielding rate in training, modifying the opacity association parameters to maintain physical constraint, finishing multi-round information redistribution, finally realizing Gaussian point information redistribution, relieving overfitting and improving reconstruction efficiency and quality.

Inventors

Yuan Yuelin
LIU XIAOHUI
Yin Tingli
XU GUIRONG
LIAO CHENZHONG
WEN CHAO

Assignees

中国人民解放军国防科技大学

Dates

Publication Date: 20260508
Application Date: 20260409

Claims (10)

1. A gaussian point information reassignment method, said method comprising: acquiring a visible Gaussian point set covering pixel points based on a three-dimensional Gaussian sputtering frame, and extracting attribute parameters of each Gaussian point from the Gaussian point set to form a Gaussian point attribute parameter set; Randomly eliminating a preset number of Gaussian points from the visual Gaussian point set to serve as Gaussian points to be eliminated, and determining a neighborhood Gaussian point set corresponding to each Gaussian point to be eliminated in the visual Gaussian point set; Based on the Gaussian point attribute parameter set, constructing an approximate model of the pixel point color with respect to the opacity of the Gaussian points, and calculating to obtain gradient items of the pixel point color with respect to the opacity of each Gaussian point to form a gradient item set; Based on the gradient item set, constructing an incidence matrix of each Gaussian point to be removed and the Gaussian point in the corresponding neighborhood Gaussian point set, and calculating an information distribution coefficient of each neighborhood Gaussian point according to the incidence matrix and the information distribution factor to form an information distribution coefficient set; According to the information distribution coefficient set, compensating the opacity and color information of each Gaussian point to be removed to the corresponding Gaussian point in the neighborhood Gaussian point set, and updating the opacity and color attribute parameters of the neighborhood Gaussian point to obtain an updated Gaussian point attribute parameter set; In the training iteration process of three-dimensional reconstruction, the Gaussian point shielding rate is progressively and dynamically adjusted, the associated parameters of the Gaussian point opacity are modified based on the adjusted Gaussian point shielding rate, the physical constraint interval of the Gaussian point opacity is maintained, meanwhile, the Gaussian point information of the round is redistributed based on the updated Gaussian point attribute parameter set until all training iterations of three-dimensional reconstruction are completed, a final Gaussian point attribute parameter set is obtained, and the redistribution of the Gaussian point information is realized.
2. The method of claim 1, wherein the gaussian spot property parameter set comprises a color of a pixel, and wherein the color of the pixel is: Wherein, the Indicating the opacity of the ith gaussian point at pixel point p, Representing Gaussian point at pixel The projection weight at which the image is projected, Represent the first The pixel coordinates of the center of projection of the gaussian point at the current viewing angle, Representing pixels The coordinates of the pixels are used to determine, As a scale parameter approximated by the post-projection covariance, For the color characteristic of the ith gaussian point at pixel point p, Expressed as the transmittance of the ith gaussian point at the pixel point p, K denotes the number of gaussian points covering the pixel point p.
3. The method of claim 1, wherein constructing an approximation model of pixel point color with respect to gaussian point opacity based on the gaussian point property parameter set comprises: Based on the Gaussian point attribute parameter set, constructing an approximate model of the color of the pixel point about the opacity of the Gaussian point as follows: Wherein, the The color of the pixel point is indicated, Indicating the opacity of the ith gaussian point at pixel point p, Representing Gaussian point at pixel The projection weight at which the image is projected, Represent the first The pixel coordinates of the center of projection of the gaussian point at the current viewing angle, Representing pixels The coordinates of the pixels are used to determine, As a scale parameter approximated by the post-projection covariance, For the color characteristic of the ith gaussian point at pixel point p, Representing the transmittance of the ith gaussian point at pixel point p, A gradient term indicating that the color at pixel point p is opaque to the ith gaussian point.
4. The method of claim 1, wherein computing gradient terms of the opacity of the pixel point color to each gaussian point to form a set of gradient terms, comprising: combining Gao Sidian projection weights at the pixel points, and calculating to obtain gradient items of the opacity of the pixel point colors to each Gaussian point through the approximate model; And integrating the gradient items of all the Gaussian points according to the corresponding relation between the Gaussian points and the pixel points to form the gradient item set.
5. The method of claim 4, wherein the gradient term set provides basic data for a subsequent build objective function, the objective function being used to minimize the effect on pixel color after gaussian point elimination, expressed as: Wherein, the To get rid of the neighborhood point set around the gaussian point j participating in the current rendering, For a gradient term where the color at pixel point p is opaque to the culled gaussian point j, To reject the amount of opacity change of gaussian point j contributing to the color rendering at pixel point p, For the amount of opacity change of the gaussian point i contributing to the color rendering at the pixel point p, Is a gradient term for the color at pixel point p that is opaque to gaussian point i.
6. The method of claim 5, wherein the compact form of the objective function is: Wherein, the The amount of opacity change for the gaussian points in the neighborhood of points, And And respectively splicing the neighborhood point set and the gradient item corresponding to the Gaussian point to be removed.
7. The method of claim 1, wherein constructing an association matrix of each gaussian point to be culled with a gaussian point in a corresponding neighborhood gaussian point set based on the gradient term set comprises: based on the gradient item set, constructing an incidence matrix of each Gaussian point to be removed and the Gaussian point in the corresponding neighborhood Gaussian point set as follows: Wherein, the Representing the transmittance of the ith gaussian point at pixel point p, Representing the ith Gaussian point at the pixel The projection weight at which the image is projected, The transmittance of the jth gaussian point at the pixel point p is represented, Representing the jth Gaussian point at the pixel The projection weight at which the image is projected, Representing the color characteristics of the ith gaussian point, Representing the color characteristics of the jth gaussian point.
8. The method of claim 1, wherein calculating the information distribution coefficient for each neighborhood gaussian point based on the correlation matrix and the information distribution factor comprises: The information distribution coefficient of each neighborhood Gaussian point is calculated according to the incidence matrix and the information distribution factor and is: Wherein, the The information allocation factor is represented by a number of information, Representing the correlation matrix of the image, Representing the inverse of the correlation matrix.
9. The method of claim 1, wherein compensating the opacity and color information of each gaussian point to be removed to the gaussian point in the corresponding neighborhood gaussian point set according to the information distribution coefficient set, updating the opacity and color attribute parameters of the neighborhood gaussian point, and obtaining an updated gaussian point attribute parameter set, comprises: According to the information distribution coefficient set, the opacity and the color information of each Gaussian point to be removed are compensated to the corresponding Gaussian point in the neighborhood Gaussian point set, the opacity and the color attribute parameters of the neighborhood Gaussian point are updated, and the updated Gaussian point attribute parameters are respectively: Wherein, the 、 The opacity and color characteristics of the previous neighborhood gaussian point i are updated respectively, 、 The opacity and color characteristics of the gaussian point j to be culled, Assigning coefficients to the information of the neighborhood gaussian point i, 、 Is the updated opacity and color characteristics of the neighborhood gaussian point i.
10. The method of claim 1, wherein modifying the associated parameters of gaussian point opacity based on the adjusted gaussian point mask rate comprises: Gaussian point opacity Associated parameters The corresponding relation of (2) is: based on Gaussian point shielding rate Modifying associated parameters The process of (1) is as follows: Wherein, the In order to modify the associated parameters before modification, Is an associated parameter of modified gaussian point opacity.

Description

Gaussian point information redistribution method Technical Field The application relates to the technical field of computer vision and three-dimensional reconstruction, in particular to a Gaussian point information redistribution method. Background The three-dimensional Gaussian sputtering (3 DGS) technology becomes a research hot spot in the field of three-dimensional reconstruction by virtue of high rendering quality and high-efficiency training performance, and is widely applied to scenes such as robot perception, automatic driving, virtual reality and the like. In the three-dimensional reconstruction task of sparse views, the 3DGS technique faces the problem of overfitting. In the training process, because the input visual angle is limited, floating artifacts are easy to appear, and the rendering effect of the new visual angle is reduced. The existing solution mostly adopts a mode of randomly shielding Gaussian points, and the visibility of the residual Gaussian points is improved. However, the method has the problem of information loss, does not directly reduce the Gaussian point rule, has the problem of Gaussian point redundancy in the reasoning stage, and brings additional calculation and storage burden to influence the reconstruction efficiency and quality. Disclosure of Invention In view of the above, it is necessary to provide a gaussian point information reassignment method capable of compensating for information loss, alleviating overfitting, and reducing calculation load. A gaussian point information reassignment method, the method comprising: Acquiring a visible Gaussian point set covering pixel points based on a three-dimensional Gaussian sputtering frame, extracting attribute parameters of each Gaussian point from the Gaussian point set, and forming a Gaussian point attribute parameter set; randomly eliminating a preset number of Gaussian points from the visual Gaussian point set to serve as Gaussian points to be eliminated, and determining a neighborhood Gaussian point set corresponding to each Gaussian point to be eliminated in the visual Gaussian point set; Based on the Gaussian point attribute parameter set, constructing an approximate model of the pixel point color with respect to the opacity of the Gaussian points, and calculating to obtain gradient items of the opacity of the pixel point color to each Gaussian point to form a gradient item set; Based on the gradient item set, constructing an incidence matrix of each Gaussian point to be removed and the Gaussian point in the corresponding neighborhood Gaussian point set, and calculating an information distribution coefficient of each neighborhood Gaussian point according to the incidence matrix and the information distribution factor to form an information distribution coefficient set; according to the information distribution coefficient set, compensating the opacity and color information of each Gaussian point to be removed to the Gaussian point in the corresponding neighborhood Gaussian point set, and updating the opacity and color attribute parameters of the neighborhood Gaussian point to obtain an updated Gaussian point attribute parameter set; In the training iteration process of three-dimensional reconstruction, the Gaussian point shielding rate is progressively and dynamically adjusted, the associated parameters of the Gaussian point opacity are modified based on the adjusted Gaussian point shielding rate, the physical constraint interval of the Gaussian point opacity is maintained, meanwhile, the Gaussian point information of the round is redistributed based on the updated Gaussian point attribute parameter set until all training iterations of three-dimensional reconstruction are completed, a final Gaussian point attribute parameter set is obtained, and the redistribution of the Gaussian point information is realized. According to the Gaussian point information redistribution method, firstly, the visual Gaussian point sets are obtained and the attribute parameters are extracted, so that complete basic data support is provided for information redistribution, the core attribute of the subsequent calculation can be accurately related to the Gaussian points is ensured, the whole scale of the Gaussian points is directly reduced by randomly eliminating the preset number of Gaussian points, the calculation and storage burden is reduced from the source, and the problem of low efficiency caused by Gaussian point redundancy in the traditional method is avoided. The method comprises the steps of establishing a gradient item, establishing an associated matrix and calculating an information distribution coefficient, wherein the gradient item is used for establishing an approximate model of pixel point color with respect to opacity, calculating the gradient item, accurately quantifying the influence of each Gaussian point on pixel color, providing scientific basis for information distribution, establishing an associated matrix based on the