KR100305592B1 - Building method of content-based image retrieval language using the VQL - Google Patents

Abstract

The present invention relates to a content-based video query generation method using an intermediate language.

The present invention collects content-based retrieval information for content-based retrieval from a user interface, and classifies data types through a query parser based on input image characteristic parameters, and queries based on the data classified from the step. Generating an intermediate language by determining a type, and generating a query sentence via parsing and contextual analysis using the intermediate language generated from the above step. Since it is not possible to base query, it creates user's convenience and maximizes search efficiency by creating a standard query using the middle language (VQL: Video Query Language) suitable for the content-based query. Content-based video using intermediate language We want to provide a query generation method.

Description

Building method of content-based image retrieval language using the VQL}

The present invention relates to a method for generating a content-based video query using VQL (Video Query Language). In particular, in the construction of a video database retrieval system, a query change part according to a user's retrieval requirements according to various application fields is described. The present invention relates to a method of generating content-based video query using an intermediate language that can be minimized.

The conventional method simply uses a static method of mapping a user's requirements and a query one-to-one (Static SQL), or selects and searches a representative query that can be used jointly by predicting several cases (Dynamic). SQL) has been taken. Therefore, each time a user's needs were changed, a new query had to be created, and when the application field was changed, it was inconvenient to rebuild from the beginning, and a lot of time wasted in testing and supplementing the newly constructed system.

On the other hand, as prior art related to this, Patent Registration No. 94-9868 (the right holder; Sakai Yurikoko et al., Two countries, Japan; name; image data management method and image database) collects image data input from the outside and collects the image. The image data management method of adding and storing the collection time as the retrieval information to the data and displaying, retrieving, and outputting the data by the collection time is specified. Patent Registration No. 91-2745 (Owner; Akio Okazaki, Country; Japan, Name) Information retrieval device) specifies an information retrieval device that performs information retrieval according to a retrieval request received through an information retrieval interface for receiving retrieval request information such as symbols, characters, and figures having a spatial positional relationship, and a patent registration number 96-18993 (right holder; Kubotari, country name; Japan, name; information retrieval method and system). By using an index file that contains information, we searched a document containing a string similar to a specified string and a string of characters at high speed, and specified a search accuracy to find a string similar to a string in the document.

Therefore, in order to solve the above-mentioned disadvantages, the present invention expresses the user's query content in an intermediate language and converts it into a query language (SQL) so that a new query can be executed by a simple procedure even if the content of the query is changed. An object of the present invention is to provide a method of generating a content-based video query using an intermediate language that enables a user to query without restricting the query range.

In order to achieve the above object, a method of generating a content-based video query using an intermediate language according to the present invention is a first step of classifying a content-based query type into a text query and a video query through a query interface capable of content-based searching. And a second step of generating an image characteristic parameter in the case of a video query after analyzing the classified query and determining the type, and contents of the user query input through analyzing the image characteristic parameter using the image characteristic structure information. And a fourth step of converting to a middle language, and a fourth step of performing optimization through syntax and context analysis using the intermediate language and generating a query sentence using an optimized query.

In addition, the present invention for achieving the above object is a first step of classifying the form of the content-based query input through the query interface capable of content-based search into a text query and a video query, and the type after analyzing the classified query A second step of generating an image characteristic parameter in the case of a video query, and converting the input user query content into an intermediate language through analyzing the image characteristic parameter using the image characteristic structure information; And a computer-readable recording medium recording a program for executing the fourth step of performing optimization through syntax and context analysis using the intermediate language and generating a query sentence using the optimized query. .

1 is a conceptual diagram of query generation according to the present invention.

2 is a flow chart for a query parser in accordance with the present invention.

3 is a flowchart of a query generator according to the present invention;

<Description of Signs of Major Parts of Drawings>

1: user interface 2: query parser

3: query generator 4: database

21: data type classification function 22: query type determination function

23: intermediate language (VQL) conversion function 31: parsing function

32: Context Analysis Function 33: Optimization & SQL Code Generation Function

Hereinafter, with reference to the accompanying drawings will be described in detail the present invention.

1 is a conceptual diagram of query generation according to the present invention, FIG. 2 is a flowchart illustrating a query interpreter according to the present invention, and FIG. 3 is a flowchart illustrating a query generator according to the present invention.

Referring to FIG. 1, the conceptual diagram of query generation enables a user to construct a content-based query directly through a visual interface 1 that enables content-based search using feature information such as image, color, and texture. The query content is inputted by the user through the query parser (2) to determine the query type and converted into a middle language. The intermediate language can be recognized again by a commercial database using the query generator (3). Conversion to standard SQL (Structured Query Language) statements. In order to explain this process in more detail with reference to the accompanying Figures 2 and 3 will be described in detail.

Referring to FIG. 2, the query parser 2 according to the present invention analyzes a query by analyzing a data type classification function 21 in which a user classifies various types of queries through the interface 1 and the inputted queries. It has a query type determination function 22 for determining the type.

The data type classification function 21 uses a query entered by the user through the interface 1 as an input, and outputs a parameter list according to the user query. The user inputs a keyword, a color, a texture, a temporal characteristic, a spatial characteristic, and a temporal / spatial characteristic through the user interface 1. Here, the keyword is text-based data such as letters or numbers input by the user, the color or texture is video data about image characteristics, and the spatiotemporal characteristics express the temporal and spatial characteristics of the object, thereby expressing the temporal, It consists of spatial and spatiotemporal relationships.

That is, the type of query input through the user interface 1 is classified into text data and video data, and the classified data is analyzed to determine a type. In the case of video data, an image characteristic parameter is generated.

The query type determination function 22 has a function of generating an intermediate language code by inputting image characteristic parameters generated by a user input query. The data type input by the user is analyzed by image characteristic structure information. . This is the first step to accurately understand the user's query request and optimize the user's query, and converts the analyzed query content into intermediate language (VQL: Video Query Language) to generate standard SQL.

That is, the content of the input query is analyzed by analyzing the image feature parameter using the image feature structure information and converted into an intermediate language using a Bcaf-Naur Form (BNF) expression, which is a formal definition proposed for the video query language. .

The functions of this module are needed to extend the user interface more efficiently. The user query expression format for generating intermediate languages is as follows.

<User_Deined_Exp> :: = <Video_Exp>; <DB_Exp>; <Keyword_Exp>; <Color_Exp>;

<Texture_Exp>; <Spatial_Temporal_Exp>

<Video_Exp>: Search unit

<DB_Exp>: Search target

<Keyword_Exp>: Search Condition

<Color_Exp>, <Texture_Exp>: Image search condition

<Spatial_Temporal_Exp>: Spatio-temporal search condition

The query interpreter 2 has a function of analyzing a query in which a user selects or inputs information related to a clip to be obtained through a search screen, and the data for classifying a type includes a text-based keyword (character, Input values), content-based color (enter color as an image characteristic), text (texture (enter texture as an image characteristic with text input)), and spatiotemporal query (by expressing the temporal and spatial characteristics of objects) The temporal, spatial and spatiotemporal relations of objects are classified into the form of data.These classified data are the first step to accurately understand the user's query needs and optimize the user's query based on the classified information in the query type determination function. If the search request data entered by the user is a content-based query, it is analyzed by the metadata structure information. There is by calling a function of the type of query based on converted into an intermediate language, which is to express complex queries required a formal history of the user.

Referring to FIG. 3, the query generator 3 according to the present invention uses the intermediate language 24 generated by the user query interpreter module 2 as an input to determine whether the syntax is correct. Classify them into meaningful tokens, perform query optimization through contextual analysis (32) using these tokens (33), convert the optimized query into an SQL statement (34), and request a query from the database (4). Perform the function. The input value of the user query optimization and query generation module is an intermediate code in VQL form, the output value generates a query form in SQL form, and the analyzed user query is restructured for optimization before query generation while ensuring fast retrieval. Reconstruct the optimal SQL to be generated by the conversion rule based on the information. The detailed steps are as follows.

Step 1: Parse the user query expressed in VQL form as intermediate code.

Step 2: Convert to similarity function for comparing image characteristics.

Step 3: Replace the function representing the spatiotemporal properties with a member function of metadata consisting of objects according to the conversion rules.

Step 4: Convert the conditions of the query according to the conversion rules.

<Conversion rule>

(1) Reconstruct with subquery when there are many conditions.

(2) Reorganize into a UNION statement if there is an OR operation.

Step 5: Create a Query

That is, the query generator 3 performs an optimization function to ensure fast search by using the intermediate code generated by the query interpreter 2 and converts the optimized query into an SQL statement that can be recognized by a commercial database. After executing this function, you can query the database. Here, parsing or contextual analysis refers to analyzing the intermediate code by the type of query in order to convert the intermediate code into SQL statements that can be recognized by the database, and analyzing the syntax and context of the intermediate language in order to convert the intermediate code into appropriate SQL. In order to ensure fast retrieval according to the user's query, the optimization function is to reconstruct the optimal SQL to be generated by the conversion rule based on metadata structure information before the query is generated.

The present invention described above is capable of various substitutions, modifications, and changes without departing from the spirit of the present invention for those skilled in the art to which the present invention pertains. It is not limited.

As described above, the present invention enables not only image-based data retrieval from content-based video data, but also retrieval reflecting spatio-temporal characteristics of specific objects in the video, thereby providing a video content that is the core of multimedia data. It can be used more efficiently to search based on.

In addition, the present invention facilitates the expansion to the content-based search in the conventional text-based search system, the user can perform the content-based search without using the SQL statement required for the database search directly through their interface environment .

Claims (6)

A first step of classifying a content-based query type into a text query and a video query through a content-based search query interface; A second step of analyzing the classified query and determining a type to generate an image characteristic parameter in the case of a video query; A third step of converting a user query content into an intermediate language by analyzing the image characteristic parameter using image characteristic structure information; And a fourth step of performing optimization through syntax analysis and context analysis using the intermediate language and generating a query sentence using the optimized query. The method of claim 1, The form of the query of the first step is classified by text data, color and texture image characteristic information input through a user interface, and temporal / spatial characteristics of an object. . The method of claim 1, In the intermediate language of the third step, a parameter having a text and an image characteristic is classified into a unit, object, condition, image condition, and space-time condition of a search using a predefined user query expression format, and then mapped to a specific format. Content-based video query generation method using the intermediate language, characterized in that converted through. The method of claim 1, The fourth step query sentence is a content-based video query using the intermediate language, characterized in that it is generated by the step of determining whether the syntax is correct using the generated intermediate language, and the context analysis step using the classified meaningful tokens How to produce. The method of claim 5, The generating of the query statement uses a similarity function to compare image characteristics, and converts a function representing a spatiotemporal characteristic to a metadata member function configured according to the conversion rule to replace the function representing the spatiotemporal characteristic. Content-based video query generation method using an intermediate language characterized in that the. A first step of classifying a content-based query type into a text query and a video query through a content-based search query interface; A second step of analyzing the classified query and determining a type to generate an image characteristic parameter in the case of a video query; A third step of converting a user query content into an intermediate language by analyzing the image characteristic parameter using image characteristic structure information; A computer-readable recording medium having recorded thereon a program for executing the fourth step of performing optimization through syntax and context analysis using the intermediate language and generating a query sentence using the optimized query.