KR20200063563A - Similarity information providing system through image comparison - Google Patents

Abstract

The present invention relates to a similarity information providing system through image comparison. According to the present invention, by utilizing deep learning technology, an image file to be searched is directly compared with a figure image of a searched first-applied trademark, and similarity information is numerically provided. Therefore, the completeness of search results can be increased.

Description

Similarity information providing system through image comparison}

The present invention relates to a system for providing similarity information through image comparison, and more specifically, to make it possible to quantify and provide similarity by directly comparing images to be searched when searching for a trademark trademark and images of pre-registered trademark data built in a database. It's about technology.

Trademarks used for the identification of goods or services are composed of symbols, letters, figures, three-dimensional shapes, or combinations and colors thereof. And, according to the trademark law, it is necessary to grant exclusive and exclusive rights to products and other trademarks that are handled by others with their own discernment power. Therefore, in the process of reviewing the trademarks filed, it is necessary to check whether there are any other registered trademarks.

Meanwhile, with the establishment of the electronic filing system, the database of application trademarks was completed, and online search technology also developed. Accordingly, on the basis of the completed database, online searches for similar trademarks or similar trademarks are being actively conducted for the trademarks to be applied for or the trademarks to be reviewed in the application process and the examination process. However, the current level of skill in searching for trademarks remains in text search for character trademarks.

On the other hand, the prior art related to the trademark search service includes Korean Patent Publication No. 10-2002-0006339 (2002.01.19.'Trademark Search Method and Trademark Search System').

The present invention has been devised to solve the problems of the prior art as described above, by directly comparing the image file to be searched with the image of the searched prior application trademark using deep learning technology and providing similarity information as a numerical value. The aim is to provide a technique that can improve the completeness of search results.

A system for providing similarity information through image comparison according to the present invention for achieving the above object includes: an information receiving unit receiving an image file from a user terminal; A search query generator that infers the Vienna code and generates a search query by analyzing the image file received through the information receiver; A search unit that searches for trademark data stored in a database using a search query generated by the search query generator and extracts trademark data having the same or similar number of Vienna codes as the Vienna code of the image file; An image comparison unit that compares the image of the trademark data extracted from the search unit with the image file received through the information receiving unit to calculate similarity information and matches each trademark data; And an information providing unit that provides trademark data matching the similarity information calculated by the image comparison unit and extracted from the search unit to the user terminal.

Here, a deep learning technique may be applied when comparing the image file and the trademark data in the image comparison unit.

Also, the search unit may sort the trademark data according to similarity information calculated by the image comparison unit.

Using the similarity information providing system according to the present invention, even if the general user does not properly grasp the Vienna code of the graphic trademark, it is possible to infer the Vienna code of the input image file and perform an automatic search using a machine learning technique. Therefore, the search efficiency of graphic trademarks is increased.

In addition, unlike text-based search methods, it is possible to search for intellectual property rights such as trademarks because it is possible to search using images. Can help directly.

In addition, the degree of similarity between the target brand image to be searched and the trademark image of the search result is provided together, thereby enhancing the user's analysis convenience.

1 is a block diagram illustrating a system for providing similarity information through image comparison according to an embodiment of the present invention.
2 is a flowchart illustrating a method for providing similarity information according to an embodiment of the present invention.

Hereinafter, preferred embodiments of the present invention will be described with reference to the accompanying drawings. However, some components irrelevant to the subject matter of the invention will be omitted or compressed, but the omitted components are not necessarily necessary in the present invention, and may be used in combination with those skilled in the art to which the present invention pertains. Can be.

In addition, each of the components and servers and systems described below does not necessarily have to be composed of independent components or servers that perform respective functions, and may be implemented as a set of one or more programs or one or more servers or one or more systems, or some It can be shared.

1 is a block diagram illustrating a system for providing similarity information through image comparison according to an embodiment of the present invention. Prior to the description, the similarity information providing system shown in FIG. 1 basically supports a trademark search service function, and provides a service so that similarity information is exposed on a search result screen.

1, the similarity information providing system 100 according to an embodiment of the present invention includes a communication unit 110, an interface providing unit 120, an information receiving unit 130, an information providing unit 140, and a search query generation. It includes a unit 150, a search unit 160, an image comparison unit 170, and a database 190.

The communication unit 110 is provided to connect a communication channel for transmitting and receiving information with the user terminal 200 through a wired or wireless network. Here, the user terminal 200 refers to a terminal used by a user who searches for a figure trademark and checks the result, inputs a search command through the user terminal 200, uploads an image file, and outputs a search result. . As the user terminal 200, a desktop computer, a laptop computer, a tablet PC, or a smart phone that can connect and transmit information to and from the similarity information providing system 100 through an Internet network or a mobile communication network may be used.

The interface providing unit 120 is provided to provide various user interfaces for providing a search service to the user terminal 200.

The information receiving unit 130 is provided to receive a command or image file for searching from the user terminal 200, and the information providing unit 140 is provided to provide a search result to the user terminal 200 side.

The search query generation unit 150 is provided to analyze the image file received through the information receiving unit 130 to infer the Vienna code and generate a search query.

The search unit 160 is provided to search the database 190 by using the search query generated by the search query generation unit 150, and to extract and organize trademark data similar to the input trademark image file.

The image comparison unit 170 is provided to directly compare images of uploaded image files and graphic trademarks of trademark data searched by the search unit 160 to calculate similarity information numerically. The similarity information calculated by the image comparison unit 170 is matched to each trademark data organized in the search unit 160.

When similar trademark data is extracted and organized in the search unit 160, and similarity information is calculated in the image comparison unit 170 to match each trademark data, the information providing unit 140 searches the user terminal 200. Results can be sent.

The database 190 is a storage medium that stores data of pre-applied trademarks.

Hereinafter, a method of providing a search service using the similarity information providing system shown in FIG. 1 will be described with reference to FIG. 2.

Prior to the description, it is assumed that the data of the pre-applied trademarks is pre-built in the database 190 of the similarity information providing system 100. Trademark data is stored in the form of matching, including trademark name, trademark sample (figure trademark image), application number, application date, applicant, registration number, registration date, product classification, designated product, legal status, etc. In addition, in the trademark data, the figure classification code, that is, the Vienna code and its description, are also stored in a match.

First, in order to search for a graphic trademark, the user accesses a website operated by the similarity information providing system 100 through the user terminal 200 or a dedicated application to access the similarity information providing system 100. When the user terminal 200 accesses the similarity information providing system 100, the interface providing unit 120 provides a user interface capable of inputting a search command and uploading an image file.

When the user interface is output from the user terminal 200, the user inputs an image file for the graphic label to be searched <S205> and transmits it to the similarity information providing system 100. Depending on the implementation, the format of the image file that can be uploaded or the maximum file size that can be uploaded may be determined in advance.

When an image file transmitted from the user terminal 200 is received through the information receiving unit 130 of the similarity information providing system 100, the search query generator 150 analyzes the image and generates a search query <S210>. . That is, the search query generator 150 analyzes an image according to a preset algorithm and infers the Vienna code to generate a search query.

The Vienna Code (Shape Classification Code) was established by the Vienna Convention in 1973 and entered into force in 1985 for the international unification of the classification of graphic elements on the emblem. The structure of the figure classification code is divided into categories, groups, and terms, and each number is assigned two digits. The Vienna Code was introduced to improve the quality of examination by increasing the efficiency of trademark examination by systematically managing and classifying the increasing number of trademarks each year. However, if you use these Vienna codes, you can also search for similar or identical pre-existing trademarks before filing a trademark. That is, the user enters the Vienna code in addition to the textual brand name on the search site to search for the graphic trademark. However, it is not easy for ordinary users to find out the Vienna code of a graphic trademark. To this end, the similarity information providing system 100 according to an embodiment of the present invention infers a Vienna code by analyzing an image file of a graphic trademark input by a user.

That is, the search query generator 150 infers the Vienna code through the analysis of the image file. Machine learning techniques are used to infer the Vienna code.

Machine learning is a field of artificial intelligence in computer science that evolved from the study of pattern recognition and computer learning theory. Machine learning is a technology that studies and builds systems and algorithms for learning, making predictions, and improving one's own performance based on empirical data. Rather than performing strictly defined static program instructions, machine learning algorithms take the form of building a specific model to make predictions or decisions based on input data.

Using such a machine learning technique, the search query generator 150 continually performs machine learning for analyzing the graphic image of the figure, infers the Vienna code through this, and then generates a search query for the search.

When a search query is generated in the search query generation unit 150, the search unit 160 searches the database 190 using the search query (S215). That is, the search unit 160 searches the database 190 for an image similar to the image file for the graphic trademark input from the user terminal 200, and prioritizes trademark data having the same Vienna code as the input image file. Search extract. In addition, if one or more Vienna codes of the image file are derived, the trademark data matching a certain number of Vienna codes is searched and extracted preferentially.

When trademark data having the same or similar Vienna code as the image file input through the search unit 160 is searched and extracted, the search unit 160 organizes the extracted trademark data. For example, the search unit 160 classifies the extracted trademark data by class (product classification), and sorts the trademark data in the order in which the Vienna code is most similar within the class. Of course, if a user inputs a specific product classification in advance while inputting an image file, only the same product classification or trademark data of a similar group can be used as a search result.

Thereafter, the image comparison unit 170 directly compares the trademark image (figure) of the input image file and the extracted trademark data to calculate similarity information numerically <S220>, and the calculated similarity information is matched to each trademark data do. In this case, the search unit 160 may sort the trademark data in descending order based on the similarity value calculated by the image comparison unit 170.

Here, the image comparison unit 170 uses a deep learning technique for direct comparison between images.

Deep learning refers to attempting to achieve a high level of abstraction (abstraction, summarizing features, key content or functions in a large amount of data or complex data) through a combination of several nonlinear transformation methods. It is defined as a set of machine learning algorithms, and is a field of machine learning that teaches a person's way of thinking to computers. Using this deep learning technique, the similarity can be quantified by directly comparing the input image with the trademark data.

After the similarity value calculated by the image comparison unit 170 (similarity information calculated as a numerical value) is matched and arranged for each trademark data, the trademark data is provided and output to the user terminal 200 through the information providing unit 140 When <S225>, the user can check the trademark data having the same Vienna code as the input image file, and also check the image similarity value of each trademark data.

As described in detail above, when the similarity information providing system 100 according to the present invention is used, even if an ordinary user does not properly grasp the Vienna code of the graphic trademark, the Vienna code of the image file input using the machine learning technique can be used. Can infer and perform automatic searches. Therefore, the search efficiency of graphic trademarks is increased.

In addition, unlike text-based search methods, it is possible to search for intellectual property rights such as trademarks because it is possible to search using images. Can help directly.

In addition, the degree of similarity between the target brand image to be searched and the trademark image of the search result is provided together, thereby enhancing the user's analysis convenience.

The above-described preferred embodiments of the present invention are disclosed for purposes of illustration, and those skilled in the art having ordinary knowledge of the present invention will be able to make various modifications, changes, and additions within the spirit and scope of the present invention. And additions should be considered to fall within the scope of the claims of the present invention.

100: similarity information provision system
110: communication unit
120: interface providing unit
130: information receiving department
140: information providing department
150: search query generator
160: search unit
170: Image comparison department
190: database
200: User terminal

Claims (3)

An information receiving unit that receives an image file from a user terminal;
A search query generator that infers the Vienna code and generates a search query by analyzing the image file received through the information receiver;
A search unit that searches for trademark data stored in a database using a search query generated by the search query generator and extracts trademark data having the same or similar number of Vienna codes as the Vienna code of the image file;
An image comparison unit that compares the image of the trademark data extracted from the search unit with the image file received through the information receiving unit to calculate similarity information and matches each trademark data; And
And an information providing unit that provides trademark data matching the similarity information calculated by the image comparison unit and extracted by the search unit to the user terminal.
According to claim 1,
A system for providing similarity information through image comparison, characterized in that a deep learning technique is applied when the image comparison unit compares the shape of the image file and the trademark data.
According to claim 2,
The search unit provides the similarity information providing system through image comparison, wherein the trademark data is sorted according to the similarity information calculated by the image comparison unit.

Images