Looking for breakthrough ideas for innovation challenges? Try Patsnap Eureka!

Local feature descriptor representation method and system

A feature descriptor and local feature technology, which is applied in the field of representation methods and systems of local feature descriptors, can solve the problems of only considering the correlation of sub-region blocks, but not considering the spatial relationship of sub-region blocks.

Pending Publication Date: 2020-06-12
北京腾信软创科技股份有限公司
View PDF9 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

This invention only considers the relevance of the position of the sub-area blocks, and does not consider the spatial relationship between all sub-area blocks

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
  • Local feature descriptor representation method and system
  • Local feature descriptor representation method and system
  • Local feature descriptor representation method and system

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0035] Such as figure 1 As shown, execute step 100 to generate sift feature descriptors, use 4×4 regions to describe key points, divide 0° to 360° into 8 direction ranges, each range is 45°, and calculate the cumulative value.

[0036] Execute step 110, respectively count the gradient direction histograms in 8 directions in 8 sub-regions, and there are 8 directions of gradient strength information in each sub-region to form a 4x4x8=128-dimensional SIFT feature descriptor, and the result is [x 1 ,x 2 ,...,x j ,...,x 16 ], where, 1≤j≤16,x j ∈ R 8 , R 8 Represents an 8-dimensional real number space.

[0037] Step 120 is executed, and the gradient direction histograms of the 8 directions of the 4 sub-region blocks are respectively counted. Transform the 4×4 region block into a 2×2 region block, and count the gradient direction histograms of the 8 directions of the 4 sub-region blocks respectively, and the result is [y 1 ,y 2 ,y 3 ,y 4 ], where y1, y2, y3, y4∈R 8 , R ...

Embodiment 2

[0041] Such as figure 2 As shown, a local feature descriptor representation system includes a sift descriptor generating module 200 , a statistical module 210 and a generating module 220 .

[0042] The sift descriptor generation module 200 is used to generate a sift descriptor, which uses a 4×4 region to describe key points, divides 0° to 360° into 8 direction ranges, each range is 45°, and calculates the cumulative value.

[0043] Statistics module 210 has the following functions:

[0044] 1) Count the histograms of gradient directions in 8 directions in 8 sub-regions. There are 8 directions of gradient strength information in the sub-regions to form a 4x4x8=128-dimensional SIFT feature descriptor, and the result is [x 1 ,x 2 ,...,x j ,...,x 16 ], where, 1≤j≤16,x j ∈ R 8 , R 8 Represents an 8-dimensional real number space.

[0045] 2) Transform the 4×4 region block into a 2×2 region block, and count the gradient direction histograms of the 8 directions of the 4 sub-...

Embodiment 3

[0049] The invention discloses a method for calculating image local feature descriptors. Although the SIFT descriptor can use the rotation invariance and translation invariance of the image very well, it does not consider the spatial information between the area blocks when calculating the feature descriptor of the feature point. The present invention adds more spatial information by integrating the image pyramid space division method into the representation of the SIFT descriptor. The concept of the method of the invention is simple, and the new descriptor has stronger discrimination ability and robustness.

[0050] image 3 Is the generation process diagram of the feature descriptor based on the pyramid space division. The process of generating feature descriptors based on pyramid space division is as follows:

[0051] (1) Firstly, the traditional sift feature descriptor is generated, and a 4×4 region is used to describe the key points. Divide 0°~360° into 8 direction ra...

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 provides a local feature descriptor representation method and system, and the method comprises the steps: generating a sift feature descriptor, and also comprises the following steps: respectively carrying out the statistics of gradient direction histograms in eight directions in eight subareas; respectively counting gradient direction histograms of the four sub-region blocks in eight directions; counting gradient direction histograms of one region block in eight directions; and connecting the statistical results obtained in the previous steps to generate a feature descriptor ofthe key point. According to the representation method and system of the local feature descriptor provided by the invention, the pyramid space division method of the image is fused into the representation of the SIFT descriptor, so that more space information is added. The method provided by the invention is simple in concept, and the new descriptor has stronger discrimination capability and robustness.

Description

technical field [0001] The invention relates to the technical field of computer vision, in particular to a representation method and system of local feature descriptors. Background technique [0002] The SIFT descriptor proposed by David Lowe in 2004 is the most representative one among image local feature descriptors. The SIFT descriptor divides the 4×4 window around the feature point into 16 sub-regions, counts the 8 direction gradient histograms of each sub-region, and finally forms a 128-dimensional feature vector. For details, see David Lowe.Distinctive image features from scale invariant keypoints .IJCV 2004. The SIFT descriptor has good invariance to illumination changes, background occlusion, rotation and scale transformation, and is considered to be one of the best-performing descriptors. Although the SIFT descriptor can make good use of the image's rotation invariance and translation invariance, there is still room for improvement. [0003] Lowe[David Lowe. Dist...

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): G06K9/46
CPCG06V10/50G06V10/462
Inventor 梁晔马楠何勤许亮
Owner 北京腾信软创科技股份有限公司
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Patsnap Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Patsnap Eureka Blog
Learn More
PatSnap group products

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

© 2024 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap|About US| Contact US: help@patsnap.com