Method and device for three-dimensional reconstruction based on neural radiance field and semantic segmentation

CN116958446BActive Publication Date: 2026-08-28INST OF SOFTWARE - CHINESE ACAD OF SCI
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202310960660.3
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2023-08-02
Publication Date
2026-08-28
Estimated Expiration
2043-08-02

AI Technical Summary

Technical Problem

[0004]1.使用深度信息的重构方法:需要带有深度监督的训练图像或外部深度估计模块,并且容易受到由于深度预测不正确而产生的大投影误差的影响;

Benefits of technology

[0063]与现有技术相比,本发明的特点在于使用语义分割信息以及语义标签内部特征,约束和规范神经辐射场的训练过程,其积极作用在于:

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN116958446B_ABST
    Figure CN116958446B_ABST
Patent Text Reader

Abstract

The application discloses a few-shot three-dimensional reconstruction method and device based on a neural radiance field and semantic segmentation, and the method comprises the following steps: training the neural radiance field based on semantic loss, adjacent semantic loss, random adjacent semantic loss, color loss and texture loss; and reconstructing a test picture based on the trained neural radiance field to obtain a reconstruction result of the test picture. The present application can correct the reconstruction process and bring as few fine-grained constraints as possible, avoid inaccurate prior knowledge from bringing wrong guidance, and thus achieve better reconstruction quality.
Need to check novelty before this filing date? Find Prior Art