面向电力储能设备的门控循环单图像三维重建方法OA

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.