Convolutional network-based stereo matching method and related device

A convolutional network and stereo matching technology, applied in the field of image processing, can solve problems such as poor accuracy and slow speed, and achieve the effect of improving accuracy and avoiding slow speed.

Pending Publication Date: 2022-03-01
GUANGZHOU VOCATIONAL COLLEGE OF SCI & TECH
View PDF0 Cites 1 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

The calculation speed of the two-volume convolutional network is faster than that of the three-dimensional convolutional network, but the accuracy is slightly worse than that of the three-dimensional convolutional network.
Although the three-dimensional convolutional network has higher precision than the two-dimensional convolutional network, it is slower

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
  • Convolutional network-based stereo matching method and related device
  • Convolutional network-based stereo matching method and related device
  • Convolutional network-based stereo matching method and related device

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0033] The method and technical effects of the present invention will be described below through specific examples.

[0034] refer to figure 1 and figure 2 , the present invention is based on the convolutional network stereo matching method, including obtaining the left and right images of the binocular camera or video images containing the left and right images, and adopting the method of combining the scale-balanced pyramid convolutional network and the ConvLSTM convolutional long-short memory network, for binocular camera images Feature extraction is performed, and a preliminary disparity map is generated through threshold filtering similarity calculation, and then a final high-precision disparity map is generated through a convolutional long-short memory network.

[0035] The concrete process of this method includes:

[0036] Use the scale-balanced pyramid convolutional network to perform feature extraction on the left and right images obtained through the binocular cam...

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

The invention discloses a three-dimensional matching method based on a convolutional network and a related device, and a new end-to-end network model is formed by combining a scale equalization pyramid convolutional network and a ConvLSTM convolutional long-short memory network model. According to the network model, multi-scale image features are extracted by using a scale-balanced pyramid convolutional network, information between feature layers is fused, and a preliminary disparity map is generated through threshold filtering similarity calculation and is used for ConvLSTM training. And for the whole network, the problem of low speed of the three-dimensional convolutional network is avoided, and the matching precision is also improved.

Description

technical field [0001] The invention belongs to the technical field of image processing, and in particular relates to a convolution network-based stereo matching method and a related device. Background technique [0002] Image depth calculation is of great significance in the fields of autonomous driving, 3D face recognition, robot navigation, and 3D reconstruction. Stereo matching calculation methods are mainly divided into two types: traditional methods and deep learning methods. In the traditional method, it is mainly divided into four steps: cost calculation, cost aggregation, disparity estimation, and disparity optimization. Two branches are formed in deep learning methods: 2D convolution-based methods and 3D convolution-based methods. Two-dimensional convolutional networks mainly include: DispNet, CRL, and FADNet, and three-dimensional convolutional networks mainly include: GCNet, PSMNet, and GANet. The two-volume convolutional network is faster than the three-dimen...

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
Patent Type & Authority Applications(China)
IPC IPC(8): G06T7/55G06N3/04G06N3/08G06V10/52G06V10/74G06V10/80
CPCG06T7/55G06N3/08G06T2207/20228G06T2207/20081G06T2207/20084G06T2207/20016G06N3/044G06N3/045G06F18/22G06F18/253
Inventor 杨玉卢爱芬
Owner GUANGZHOU VOCATIONAL COLLEGE OF SCI & 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