In this paper, the monocular depth estimation algorithm based on feature enhanced full convolution residual network is applied to three-dimensional scene reconstruction. Combining FE-FCN[1] (Feature Enhanced Fully Convolutional Residual Networks) algorithm with ORB-SLAM2 algorithm[2], a three-dimensional scene reconstruction system of indoor scene based on monocular image depth estimation algorithm is constructed, which further makes up for the shortcomings of the traditional three-dimensional scene reconstruction algorithm in the less image texture and the reconstruction of the scene in the process of pure rotation of the camera. The practical verification shows that the three-dimensional scene reconstruction system designed in this paper can recover the indoor scene corresponding to the relevant monocular image stably and accurately.
Three-dimensional Monocular Image Reconstruction of Indoor Scene Based on Neural Network
2022-10-12
1862602 byte
Aufsatz (Konferenz)
Elektronische Ressource
Englisch
3D Scene Reconstruction with Sparse LiDAR Data and Monocular Image in Single Frame
SAE Technical Papers | 2017
|3D Scene Reconstruction with Sparse LiDAR Data and Monocular Image in Single Frame
British Library Conference Proceedings | 2017
|Indoor positioning method based on monocular image stitching
Europäisches Patentamt | 2023
|Interactive Indoor Scene Reconstruction from Image Mosaics Using Cuboid Structure
British Library Conference Proceedings | 2002
|