Single-frame depth image three-dimensional model reconstruction method and device based on adversarial network

A deep image and three-dimensional model technology, applied in biological neural network models, image data processing, neural learning methods, etc., can solve problems such as redundant shapes, unable to fully restore occluded areas, unable to generate details of objects, etc. The effect of proliferation, avoidance of camera calibration and fine process design, improving the quality of detail

Pending Publication Date: 2021-04-09
BEIJING UNIV OF TECH
View PDF0 Cites 2 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

Although methods based on this type of network improve the accuracy of 3D reconstruction, they still have certain limitations: (1) cannot fully restore the occluded area; (2) cannot generate object details; (3) are prone to redundant shapes

Method used

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View more

Image

Smart Image Click on the blue labels to locate them in the text.
Viewing Examples
Smart Image
  • Single-frame depth image three-dimensional model reconstruction method and device based on adversarial network
  • Single-frame depth image three-dimensional model reconstruction method and device based on adversarial network
  • Single-frame depth image three-dimensional model reconstruction method and device based on adversarial network

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0016] In order to solve the above problems, the present invention proposes a pose and structure aware based adversarial network model, namely PSAAN (Pose and Structure Aware based Adversarial Network), which realizes the three-dimensional reconstruction of a single frame image. Compared with the traditional 3D reconstruction method, this method does not need to manually design complex feature algorithms, avoids complex camera calibration and fine process design, and has the ability to reconstruct the occluded area of ​​the object; compared with the existing deep learning-based Compared with 3D reconstruction methods, this method not only improves the reconstruction accuracy, but also can suppress the proliferation of redundant shapes while improving the quality of object details. A large number of experiments also show that the 3D reconstruction method based on the PSAAN model is superior to the current mainstream reconstruction algorithms in terms of overall accuracy and loca...

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to view more

PUM

No PUM Login to view more

Abstract

Disclosed are a single-frame depth image three-dimensional model reconstruction method and device based on an adversarial network, a complex feature algorithm does not need to be manually designed, complex camera calibration and fine process design are avoided, the capacity of reconstructing the shielded area of the object is achieved, the reconstruction precision is improved, the detail quality of the object can be improved, and meanwhile the reconstruction efficiency is improved. Hyperplasia of redundant shapes is inhibited. The method comprises the following steps: (1) in a generator part, encoding an input single-frame depth image into a plurality of potential capsules by using a capsule encoder, and then decoding and fusing the potential capsules by using an attention decoder to generate a 3D initial shape; (2) in the discriminator part, optimizing the 3D initial shape through integral adversarial learning with a generator; and (3) in the optimizer part, carrying out local structure optimization on the 3D initial shape to generate a 3D refined shape.

Description

technical field [0001] The present invention relates to the technical fields of computer vision and computer graphics, in particular to a method for reconstructing a three-dimensional model of a single-frame depth image based on an adversarial network, and a device for reconstructing a three-dimensional model of a single-frame depth image based on an adversarial network. Background technique [0002] With the development of artificial intelligence, computer vision, and image processing, 3D reconstruction, as one of the key technologies for environmental perception, is in increasing demand in the fields of robotics, autonomous driving, virtual reality, and augmented reality. Traditional 3D reconstruction methods, such as Structure from Motion (SFM), Simultaneous Localization and Mapping (SLAM), etc., require dense viewpoint images and rely heavily on feature matching across views. The reconstruction process involves many links, and it is difficult to reconstruct the shape of ...

Claims

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to view more

Application Information

Patent Timeline
no application Login to view more
IPC IPC(8): G06T17/00G06N3/04G06N3/08
CPCG06T17/00G06N3/08G06N3/045
Inventor 孔德慧刘彩霞王少帆李敬华王立春尹宝才
Owner BEIJING UNIV OF TECH
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Try Eureka
PatSnap group products

Browse by: Latest US Patents, China's latest patents, Technical Efficacy Thesaurus, Application Domain, Technology Topic.

© 2024 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap