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

an image retrieval method

An image retrieval and image technology, which is applied in the field of image processing, can solve problems such as error-prone, single extended image features, and heavy manual labeling workload, so as to achieve accurate image retrieval and improve training accuracy

Active Publication Date: 2019-02-22
北京八月瓜科技有限公司
View PDF3 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0004] The purpose of the present invention is to provide an image retrieval method to solve the problem that the image retrieval accuracy is affected by the single extended image feature and error-prone in the current image retrieval process, as well as the large workload of manual labeling

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
  • an image retrieval method
  • an image retrieval method
  • an image retrieval method

Examples

Experimental program
Comparison scheme
Effect test

Embodiment

[0051] (1) Use open source tools such as VLFeat to extract SIFT feature points for each checked image in the checked image database, and perform L2 normalization processing on the SIFT feature points (that is, change the L2 modulus length of the SIFT feature points to 1), and randomly sample Part of the feature points, and use the K-Means method to train D cluster centers, and all cluster centers form a D-dimensional dictionary;

[0052] (2) Use the D-dimensional dictionary obtained in the previous step to describe the features of the searched image and the query image, and obtain the D-dimensional feature vectors of the searched image and the query image respectively. Let Q be the feature vector of the query image, and I i (i=1,2,...,N) is the feature vector of the image to be checked;

[0053] (3) Use the convolutional neural network AlexNet to extract the 4096-dimensional image features of the last fully connected layer of the checked image, and perform feature description ...

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 an image retrieval method, comprising the following steps: performing feature description on the searched image and the query image; performing deep learning on the searched image and the query image; using the features of the query image to measure the similarity of the searched image, Get a feedback list sorted by similarity; use the query image and the image training samples in the feedback list to select a classifier; use the sample selection classifier to classify and predict the first n' images and pseudo-negative images, and take g images that are closest to the classification Annotate the first m images of the feedback list and the g images obtained in the previous step to obtain positive and negative images; fuse the features of the positive image and the query image, and use the fused features to re-search Image similarity measurement to get the final ranking result. The invention utilizes an active learning method to realize a query expansion method in image retrieval, and can realize more accurate image retrieval under the premise of a small number of user annotations.

Description

technical field [0001] The invention relates to the field of image processing, in particular to an image retrieval method. Background technique [0002] At present, content-based image retrieval methods have been more and more widely used, and query expansion method is one of the most effective methods to improve its query performance, and choosing a good expanded image is an important step in query expansion method. The existing extended image selection method is based on the first query, and the extended image is selected through the geometric verification technology based on feature points. The extended image selection performed by this method has problems such as single extended image features and error-prone. The traditional image retrieval technology based on correlation feedback focuses on constructing a better retrieval model by utilizing the results of multiple correlation feedbacks. They generally require multiple feedbacks and a relatively large amount of manual...

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 Patents(China)
IPC IPC(8): G06F16/53G06K9/62
CPCG06F16/583G06F18/2113G06F18/2155G06F18/217G06F18/24
Inventor 赵鑫李长青孙鹏
Owner 北京八月瓜科技有限公司
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