Method, device and system for determining image parallax

An image and parallax technology, applied in the field of computer vision, can solve problems such as low accuracy, low parallax accuracy, and high difficulty

Active Publication Date: 2019-10-15
HANGZHOU HIKVISION DIGITAL TECH
View PDF3 Cites 4 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

In this scheme, the real disparity needs to be obtained in advance, but it is more difficult to obtain the real disparity. Generally, the accuracy of the obtained real disparity is low, which leads to the low accuracy of the calculated disparity.

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
  • Method, device and system for determining image parallax
  • Method, device and system for determining image parallax
  • Method, device and system for determining image parallax

Examples

Experimental program
Comparison scheme
Effect test

Embodiment approach

[0142] Usually, the feature tensor dimension of an RGB image is H×W×3, where W is the width of the image to be processed, H is the height of the image to be processed, and 3 represents the number of RGB channels of the image to be processed; as a In one embodiment, S103 may include:

[0143] For each image to be processed, the feature extraction layer is used to convolve the image to be processed to obtain a feature tensor dimension of , wherein F represents the number of output channels of the feature extraction layer, and x represents the first preset downsampling multiple.

[0144] For example, the feature extraction layer can be a 5-layer convolutional neural network, such as figure 2 As shown in , conv_f1-conv_f5 are the five two-dimensional convolutional layers, ⊕ means to add two inputs and perform BN (BatchNormalization) and ELU (Exponential Linear Unit, exponential linear unit) operations, and ELU operations are also It is a convolutional layer followed by a BN la...

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 embodiment of the invention provides a method, a device and a system for determining image parallax. According to the embodiment, the unsupervised neural network is utilized, the parallax betweenthe plurality of images is determined, the unsupervised neural network is trained by using the loss function, the real parallax does not need to be supervised, the loss function contains one or more error parameters, and the error parameters are gradually reduced in the training process, that is, the parallax determination accuracy is improved, so that the parallax determined by applying the embodiment of the invention is relatively high in accuracy.

Description

technical field [0001] The present invention relates to the technical field of computer vision, in particular to a method, device and system for determining image parallax. Background technique [0002] Multi-camera can capture multiple images of the same scene at the same time, increasing the viewing angle range. The binocular camera in the multi-eye camera can also simulate the binocular vision of the human eye to provide better visual effects. Usually, it is necessary to calculate the disparity between multiple images collected by a multi-camera. [0003] The scheme of calculating disparity generally includes: using real disparity as supervision information, using multiple images collected by multi-camera as input, training the neural network, and using the trained neural network to calculate the difference between the multiple images collected by the multi-camera. parallax between. In this solution, the real disparity needs to be obtained in advance, but it is difficu...

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): G06T7/00G06N3/08
CPCG06N3/08G06T2207/20228G06T7/97
Inventor 张奎熊江杨平谢迪
Owner HANGZHOU HIKVISION DIGITAL 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