山东电力技术2026，Vol.53Issue(6)：117-128,12.DOI:10.20097/j.cnki.issn1007-9904.250262
面向电力储能设备的门控循环单图像三维重建方法
Gated Recurrent Single-image 3D Reconstruction for Electric Energy Storage Equipment
摘要
Abstract
Single-image three-dimensional(3D)reconstruction of electric energy storage equipment is the development direction of digital geometric modeling for power devices.However,existing single-image reconstruction methods based on deep learning are applied to power storage devices with low accuracy.Aiming at this issue,a 3D modeling method of single image electric energy storage equipment based on gated recurrent unit(GRU)is proposed.Firstly,the diffusion model is used to generate six orthogonal views from a single input image,which are then processed by a U-Net to create high-resolution orthogonal triplanes.DeepCache method is used to enhance the training efficiency of diffusion models.Meanwhile,GRU module with feature interaction function is designed for decoding triplanes features into signed distance field(SDF)values,texture colors,and Flexicubes parameters,which can effectively remove redundant information in features.Finally,3D model is obtained based on Flexicubes reconstruction.This method significantly improves the accuracy of generation compared to existing methods,achieving high-fidelity results within 20 seconds.关键词
扩散模型/DeepCache方法/正交三平面/门控循环/Flexicubes重建Key words
diffusion model/DeepCache method/orthogonal triplanes/gated recurrent/Flexicubes reconstruction分类
信息技术与安全科学引用本文复制引用
田晓,刘勇超,张帝,张亚萍..面向电力储能设备的门控循环单图像三维重建方法[J].山东电力技术,2026,53(6):117-128,12.基金项目
国网山东省电力公司科技项目(520633240008).Science and Technology Project of State Grid Shandong Electric Power Company(520633240008). （520633240008）
