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

Trademark image retrieval method based on deep hash method

An image retrieval and hashing technology, applied in the field of image processing, can solve problems such as limited performance, achieve the effect of improving retrieval accuracy, saving storage space and computing time

Pending Publication Date: 2020-08-21
NORTHWESTERN POLYTECHNICAL UNIV
View PDF6 Cites 2 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

However, the hashing methods applied to image retrieval in the prior art are mainly designed for natural scene images, and the performance is limited when directly applied to trademark image retrieval

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
  • Trademark image retrieval method based on deep hash method
  • Trademark image retrieval method based on deep hash method
  • Trademark image retrieval method based on deep hash method

Examples

Experimental program
Comparison scheme
Effect test

Embodiment

[0066] In this embodiment, the NPU-TM database is used for experiments, and the database contains 7139 trademark images, among which there are 319 groups of visually similar trademark images, and the rest of the trademark images are different from other images. Among similar image groups, the average group has 13.8 images, and the largest group has 52 images.

[0067] (1) Use step 1 to build a deep hash network model, such as figure 2 As shown, the deep hashing network model includes STN, RCN and hashing layer. In the model, the feature vector and hash function of the trademark image are learned at the same time; at the end of the network model, a loss function is used to increase the weight of the sample to ensure that The extracted hash codes are consistent with the feature vectors in the real space. The network model is attached figure 2 shown.

[0068] A learnable module, the spatial transformer, is inserted after the STN to enable the network to automatically spatial...

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

In order to solve the problems that trademark image retrieval is large in cardinal number, low in speed and not accurate enough, The invention provides a trademark image retrieval method based on a deep hash method, which comprises the following steps of: extracting a feature vector of a trademark image by using a convolutional neural network, converting the feature vector into a binary hash code,and comparing similarity features of the image by calculating a Hamming distance of the binary hash code. According to the method, the deep hash method is well applied to trademark image retrieval, so that the trademark image retrieval precision is higher, the speed is higher, and the error is lower.

Description

technical field [0001] The invention belongs to the field of image processing, and in particular relates to a trademark image retrieval method. Background technique [0002] As information is transmitted faster and faster, people's awareness of intellectual property rights is also getting stronger, and trademarks are important symbols used by companies, enterprises, institutions or individuals to identify their goods or services. How to quickly and accurately retrieve trademark information in an image has become the focus of people's concern. However, the number of registered trademarks and trademark applications is very large, which puts forward extremely high requirements for the development of trademark image retrieval technology and trademark image retrieval system: can accurately detect all similar trademarks; for various changes in graphics It has invariance; it can efficiently perform image retrieval, the required computing time should be shortened as much as possibl...

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): G06F16/583G06F16/51G06K9/62G06N3/04G06N3/08
CPCG06F16/583G06F16/51G06N3/08G06N3/045G06F18/22
Inventor 夏召强王晨黄东冯晓毅蒋晓悦
Owner NORTHWESTERN POLYTECHNICAL UNIV
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