A trademark image retrieval method integrating deep convolutional network and semantic analysis
A deep convolution and image retrieval technology, which is applied in the field of deep learning and image retrieval, can solve the problems of image complexity, difficulty in defining specific categories, slow speed, etc., achieve high-efficiency and accurate trademark image retrieval effects, and reduce retrieval effects Affect and solve the effect of inaccurate information
Active Publication Date: 2021-04-06
ZHEJIANG UNIV OF TECH
View PDF6 Cites 0 Cited by
- Summary
- Abstract
- Description
- Claims
- Application Information
AI Technical Summary
Problems solved by technology
However, this method has disadvantages such as slow speed and being affected by image complexity.
Moreover, the traditional method is seriously affected by human subjective factors for abstract images and more complex images.
Especially for those pure graphic trademarks and trademark images with incomplete descriptions, the traditional trademark retrieval method is not only difficult but also inefficient to use, and it is not suitable for the trademark registration application requirements under the rapid economic development of our country
At present, the number of registered trademarks in my country is increasing year by year, and the problems of traditional trademark retrieval methods such as subjective manual assignment, difficulty in defining specific classifications, and difficulty in describing trademark image similarities have become increasingly prominent, seriously restricting the development of trademark registration in my country. Automated and Efficient Trademark Search Technology Is Not Only Important, It Is Urgent
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 moreImage
Smart Image Click on the blue labels to locate them in the text.
Smart ImageViewing Examples
Examples
Experimental program
Comparison scheme
Effect test
Embodiment Construction
[0045] In order to understand the process of the present invention more easily, combine figure 1 The flow chart is described in detail as follows:
[0046] The first part uses deep learning to extract trademark image features and calculate similarity, mainly including steps 1-3.
[0047] Step 1. Preprocess the image.
[0048] Read the image of the trademark that needs to be detected input by the user, detect the position of the trademark in the image, detect the image and text part of the trademark, perform an alignment operation on the trademark image, and finally perform a normalization operation on the size of the trademark image, and further package it into lmdb The file format lays the foundation for deep learning;
[0049] Step 2. Train the deep convolutional neural network model.
[0050] Build a deep convolutional neural network model with a total of 10 layers. Among them, the first layer is the input layer, which inputs the preprocessed trademark image. Connected t...
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
Login to View More Abstract
A trademark image retrieval method that integrates deep convolutional neural networks and semantic analysis, including: Step 1. Preprocessing pictures; Step 2. Training deep convolutional neural network models; Step 3. Inputting pictures into trained models for image matching; Steps 4. Calculate the similarity of two keyword groups; Step 5. Calculate the similarity of two concepts; Step 6. Make a decision based on the feature fusion algorithm of Bayesian theory; Step 7. Use Euclidean distance to judge the feature vectors of two images The distance between them; Step 8. Calculate the similarity between trademark images; Step 9. Build a trademark image retrieval tree. This paper reduces the influence of subjective factors on the retrieval effect, solves the problem of inaccurate image retrieval information, and realizes the efficient and accurate trademark image retrieval effect.
Description
technical field [0001] The present invention relates to deep learning and image retrieval. A method of applying a deep convolutional neural network is proposed for trademark image retrieval, combined with the use of keyword groups for semantic matching. Background technique [0002] A trademark is a sign of goods or services, a symbol of corporate reputation and reliability, and has increasingly become an indispensable weapon in fierce market competition. The new mark must be sufficiently distinctive to avoid confusion or conflict with the registered mark. Based on computer vision technology, and using pattern recognition and other related computer-aided knowledge for image retrieval, it provides a good way to solve the current trademark registration problems. However, this method has disadvantages such as slow speed and being affected by image complexity. Moreover, the traditional method is seriously affected by human subjective factors for abstract images and more compl...
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
Login to View More Patent Type & Authority Patents(China)
IPC IPC(8): G06F16/58G06K9/62
CPCG06F16/5866G06V10/751G06F18/241
Inventor 高楠祝建明李利娟李伟
Owner ZHEJIANG UNIV OF TECH