KR100926448B1 - System and Method for Determining Adult Contents - Google Patents

Abstract

Adult content determination method according to an embodiment of the present invention that can determine the adult index by using the image and text collected from the object, the adult index using the adult index of the input image collected from the object Determining a skin index indicative of adult image relevance; Determining a textual adult index indicating an adult term relevance of the input text collected from the object; Determining an adult index using the skin index and the text adult index; And determining whether the object is an adult according to the adult index.

Adult index, skin index, text adult index

Description

System and Method for Determining Adult Contents}

The present invention relates to a method and system for determining adult material, and more specifically, a skin index and a text adult index are determined from a webpage document, an image, a video, etc. collected from an object. The present invention relates to a method and system for determining whether an object is adult or not as an adult index determined using a text adult index.

Due to the development of mass media, adult content is infiltrating into the home in earnest, and teenagers are exposed to harmful adult content through web pages or websites on the Internet. In addition, in most families, teenagers are easily accessing various websites using computers, and are easily accessible to adult content.

Therefore, the Internet site provider is undergoing a screening process for various multimedia content so that adult content is not provided without a predetermined authentication process on the website operated by the Internet site provider.

In particular, in the case of a search site provider or a portal site provider, content such as URLs or images or videos of an adult site is associated with adult content so that a user cannot access or download a certain adult site through a website operated by the user. It warns you that you have a problem, or blocks access to or related to adult sites.

In addition, the search site provider or the portal site provider acquires an image, a video, and the like from a host, a web page, a website, or the like, which is an inspection object, and checks whether they are an adult provider or an image, a video, etc. obtained from the adult.

Furthermore, since web pages, image posts, video posts, etc. are provided to multiple users via a search site or portal site, the search site provider or portal site provider exposes the adult content to the user by checking in advance whether the aforementioned content is adult content. We are trying to minimize the probability.

However, as the number of inspection objects for judging adult content increases rapidly and the inspectors cannot finish the inspection within the intended time, there is a need for a method to prevent the mistakes of the inspectors and even to reinspect the inspection by performing the inspection quickly and automatically. However, there is no effective method of judging adult content yet.

In particular, in the case of content such as an image or a video, it is necessary to use image processing to determine adult content, but there is a problem in that the method and procedure for the image processing are unclear.

In addition, there is a need for a method to significantly increase the detection rate of adult objects over time by setting priorities for inspection targets, but there is a problem that the method of determining an index indicating the priority and the degree of related adult content is unclear. .

The present invention is to solve the above problems, to provide a method and system for determining adult material that can determine whether adult content by quickly inspecting the adult content in an automated method as its technical problem.

In addition, the present invention is to provide a clear image processing method and procedure for the inspection of the content, such as image, video, and the like collected from the inspection object as its technical problem.

In addition, the present invention is to provide a method and procedure for determining the priority of the inspection for determining the adult content and determine the index that can indicate the degree of adult content.

Adult object determination method according to an aspect of the present invention for achieving the above object comprises the steps of determining a skin index (skin index) indicating the adult image relatedness of the input image collected from the object; Determining a textual adult index indicating an adult term relevance of the input text collected from the object; Determining the adult index using the skin index and the text adult index, and determining whether the object is an adult according to the adult index.

In accordance with an aspect of the present invention, there is provided a system for determining adult material, comprising: a skin index determiner configured to determine a skin index representing an adult image related degree of an input image collected from an object; A text adult index determination unit for determining a text adult index indicating an adult term relevance of the input text collected from the object; An adult index determiner that determines an adult index using the skin index and the text adult index; And an adult material determining unit determining whether an adult object of the object is based on the adult index.

As described above, according to the present invention, there is an effect of providing a method and system for determining adult material, which enables to perform the inspection quickly and automatically despite the large amount of inspection targets for adult content inspection.

In addition, according to the present invention, there is an effect of providing a clear image processing method and procedure to determine whether or not adult content by analyzing the image obtained from the inspection target.

In addition, according to the present invention, the priority of the inspection target can be determined by the adult index indicating the degree of adult-related content of the content collected from the inspection target, and the adult index can be determined by referring to the adult index. Therefore, the adult index may determine whether adult content to be blocked from access to the user or adult content requiring re-examination.

Before the detailed description of the embodiments of the present invention, the term adult index used in the present invention will be briefly described.

The adult index indicates the adult content relatedness of images collected from image contents, video contents, hosts, blogs, web pages, websites, bulletin boards, BBS, mini homepages, cafes, etc. In other words, the adult index means that the inspected object is an adult object.

Therefore, when the search site provider or the portal site provider knows the adult index, the adult index may be used to determine whether the subject to be inspected is an adult to be blocked from access to the user or an adult to be re-examined.

The inspection object may include at least one of an image content, a video content, a host, a blog, a web page, a website, a cafe, a mini homepage, and a bulletin board system (BBS).

Hereinafter, for convenience of explanation, the inspection object is described as a host. That is, an adult host determination method and system will be described using an example that an adult is an adult host according to an embodiment of the present invention.

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

1 is a schematic block diagram of an adult host determination system according to an embodiment of the present invention.

The adult host determination system 102 according to an embodiment of the present invention includes a skin index determination 104, a text adult index determination unit 106, an adult index determination unit 108, and an adult host determination unit 110. do.

The adult host determination system 102 collects content such as web page documents, image posts, video posts, etc. from the host to be inspected. In addition, the adult host determination system 102 may receive content such as web page documents, image posts, video posts, and the like, collected from a host and stored in a predetermined database.

The adult host determination system 102 determines a skin index for an image extracted from the inspected target host by image processing, and provides a web page document, a BBS, an image, a video, or a URL provided by the host. The text adult index is analyzed by analyzing the text extracted from the back. After determining the adult index according to the indexes, it is determined whether the host is an adult host according to the adult index.

The determination result is then transmitted to the operator of the search site, portal site, and the like. Thereafter, the operator of the search site, the portal site, or the like determines whether to provide or re-examine the host and the content obtained from the host according to the received adult host determination result.

Conceptually briefly describing the skin index and the text adult index, the skin index refers to the ratio of the color region of the skin color line of the image acquired from the host, that is, the degree of exposure of the body. In addition, the text adult index means how relevant the text obtained from the host is to an adult term.

Hereinafter, an image acquired by the adult host determination system 102 from the host for determining the skin index is referred to as an input image, and text obtained from the host for determining the text adult index is referred to as input text.

The adult host determination system 102 may use the adult index as a measure of prioritizing host inspection. For example, it may be determined that the inspection target host having a low adult index is not a high priority adult host, and the adult host determination result may be output so that the inspection target host is provided to the user. Therefore, the operator of the search site, portal site, etc. do not re-examine the host to be inspected with a low adult index later.

Hereinafter, the configuration and operation of the skin index determiner 104 will be described with reference to FIGS. 2 to 4.

2 is a schematic block diagram illustrating a skin index determiner 104 according to an embodiment of the present invention, wherein the skin index determiner 104 includes a training unit 202, a color analyzer 204, and a skin color area determination unit. The unit 206 includes a skin region determiner 208 and a skin index calculator 210.

The skin index determiner 104 determines a skin index indicating an adult image relevance of the input image collected from the host. Here, the input image collected from the host is an image obtained from a website, a web page, a bulletin board, or the like operated by the host, or an image extracted from an image post or a video post provided by the host, and the skin index determiner 104 directly inputs the image. May be collected or an input image already collected may be received.

The training unit 202 performs a training for determining a flesh color range using a first training set, and defines a flesh color line according to the flesh color range. Based on the color scheme defined in the training unit 202, the color region determination unit 206 determines whether the input image from the host includes the color region of the color scheme.

The training unit 202 determines a skin color range that defines the color of the skin color system by training the knowledge database using the images included in the first training set. Here, the training unit 202 determines the flesh color range using the relative RGB of the trained images.

Here, relative RGB is used by its definition to largely ignore the effect of luminance on the image. Therefore, if all images having a bright image or a dark image have similar relative RGB values, it is determined that they belong to the same color region. That is, in the case of a flesh-colored color, it is determined that both the dark skin color and the light skin color belong to the same color region.

This is to minimize the influence of both light and race when judging the color gamut of the skin color system of the image. The relative RGB value of the image may be represented by Equation 1 using RGB of the image.

Relative R = R / (R + G + B);

Relative G = G / (R + G + B);

Relative B = B / (R + G + B);

Equation 1 is a method of normalizing a difference by contrast by dividing an R, G or B value by an R + G + B value. The R + G + B values generally represent the light and darkness of the pixel, so the effect on relative RGB can be minimized. Relative RGB values can be represented by only two values by definition. In other words, if only the relative R and G values are known, the B value can be known by a predetermined calculation.

In an embodiment of the present invention, the average and the variance are obtained as the range of the skin color from the first training set, and the skin color area determiner 206 determines that the pixel values of the input image within the average and the variance are the skin pixels.

The training unit 202 defines a relative RGB range of the flesh color line from the relative RGB values of the images trained in the first training set. In one embodiment, the relative RGB range of the flesh line can be defined by the mean and variance of the relative RGB values of the trained images.

The color analyzer 204 extracts an input image from an image, a video, and the like obtained from the host, and analyzes the color of the input image. In the modified embodiment, the color analyzer 204 may receive the input image and analyze the color without directly extracting the input image.

In an embodiment of the present invention, when the content obtained from the host is an image, the input image may be the image content, and when the content is a video, the input image is extracted from the video by a capture method or the like. It may be a still image.

The color analyzer 204 acquires the color of each region constituting the input image through the analysis and generates color information therefor. Here, the color analyzer 204 generates relative RGB of the input image pixel as the color information.

In addition, the color analyzer 204 determines whether the input image is composed of a color region of a gray system based on the color information that is an analysis result, and notifies the skin index calculator 210 of the result. Subsequently, when the skin index calculator 210 receives a notification from the color analyzer 204 that the input image is in the gray region, the skin index is determined as the reference value. This will be described later in detail.

The skin color area determiner 206 determines whether the input image includes the color region of the skin color line based on the color information that is the analysis result of the color analyzer 204. When the input image includes the color gamut of the flesh color line, a color gamut image of the skin color line is obtained from the input image. Here, the skin color area determiner 206 determines whether the input image includes the color area of the skin color line using the skin color line defined in the training unit 202.

In one embodiment, the color of the color scheme by determining whether the relative RGB of the input image according to the analysis result of the color analyzer 204 falls within the range of the relative RGB of the color scheme defined in the training unit 202. It may be determined whether the region is included.

In an embodiment of the present invention, by using a relative RGB color space that is relatively less affected by light, the amount of data required to define the color range of the flesh color system determined by the training unit 202 is reduced, and the color analyzer 204 And a data amount and a calculation time required for determining whether the input image includes the color gamut of the skin color line in the skin area determining unit 206.

In detail, the flesh color area determiner 206 receives the relative RGB values of the pixels of the input image from the color analyzer 202 and receives the average and the variance of the relative RGB values of the flesh color lines from the training unit 202. When the relative RGB value of the pixel of the input image is included in the color range by the mean and the variance, the pixel is determined to be a color region of the flesh color line.

On the other hand, the color gamut image of the color gamut determined by the color gamut determination unit 206 may be an image in which a portion other than the color gamut of the color gamut is removed from the input image.

In one embodiment, when the input image does not include the color region of the skin color system or the gray system, the skin color calculator 206 notifies the skin index calculator 210 of the result, and the skin index calculator ( 210 may accordingly determine the skin index as a default. Here, the default value may be a value small enough to determine that the host is not an adult host.

The skin region determiner 208 determines the skin region using the color region image of the skin color line. Hereinafter, a detailed configuration and operation of the skin region determiner 208 will be described with reference to FIG. 3.

3 is a block diagram schematically illustrating a skin region determiner 208 according to an exemplary embodiment of the present invention. The skin region determiner 208 includes a filter 302 and a connection region remover 304. .

The filter unit 302 filters the color gamut image of the color scheme determined by the color gamut determination unit 206 to remove noise. In one embodiment, the filtering may include median filtering or average filtering. . On the other hand, the noise may mean a skin portion below the threshold value in the skin region determined by the skin region determiner.

The connection area removal unit 304 determines a skin area by removing a connection area having an area smaller than or equal to a set value from the noise-removed image output from the filter unit 302.

Therefore, the skin region according to an embodiment of the present invention corresponds to an image from which noise is removed from the color region image of the skin color system, and in which the connection region having an area smaller than or equal to a set value is removed according to the connected component method. do. Since the connected component method is well known in the art, a detailed description thereof will be omitted.

The connection area removal unit 304 removes small image areas that do not have a significant influence on the measurement of the adult-related degree according to the connected component, and then the skin index calculation unit 210 calculates a skin index indicating the degree of the adult image-related degree. This reduces the computation time and complexity.

In the above-described embodiment, the connection area is removed after removing the noise from the color gamut image of the skin color system. In the modified embodiment, the connection area removal unit 304 removes the connection area and then the filter unit 302. You might be able to remove the noise from the. In addition, by omitting the filter unit 302, the connection area removal unit 304 may remove only the connection area from the color gamut image of the flesh color system without undergoing a noise removing process. In addition, by omitting the connection area removal unit 304, the skin area may be determined by removing only the noise from the filter unit 302 from the image without undergoing the connection area removal process.

Referring again to FIG. 2, the skin index calculator 210 determines the skin index using the area of the skin region.

In detail, the skin index calculator 210 determines the skin index according to the ratio of the area of the skin region to the area of the input image. Here, the skin index can be represented by the following equation (2).

In addition, the skin index calculator 210 determines the skin index as a reference value when receiving a notification from the color analyzer 204 that the input image is composed of a gray area of a color region. This means that when the input image does not include the color region of the skin color line, a skin index related to the degree of skin exposure of a person is not determined, and thus the skin index is determined as a predetermined reference value. This will be described later in detail.

In addition, when the skin index is less than the first threshold value or more than the second threshold value, the skin index calculator 210 determines the skin index again as a default value. This is to determine that if the skin area is too small or too large, the input image is likely to be a non-human image or a simple box image filled with flesh-colored colors, which is low relative to the adult image. to be. For example, if judging an image with a skin index greater than 0.5 as adult content is the adult content criterion, an image with a skin index of 0.01 or less or 0.99 or more is again given a very small skin index of 0.005 so that the adult index is determined to be lower later. do.

On the other hand, the skin index calculator 210 gives a skin index (reference value) of 0.5 when the input image consists of a gray area color area. In this case, whether the host is an adult host is more affected by the text adult index than the skin index.

 For example, when the input image is a nude female image of gray color, the skin index calculator 210 determines the skin index to be 0.5. And since the input image is a nude female image, it is necessary to judge it as adult. Therefore, in this case, the text adult index determining unit 106 determines that the text adult index is high when the input text obtained from the host includes adult text such as "sex", "porno", or the like. As a result, the host to be inspected later is determined to be an adult host.

4 is a diagram illustrating images processed by the skin index determiner 104 according to an embodiment of the present invention.

Referring to FIG. 4, FIG. 4A is an input image extracted from an image or a video provided by a host and is a female image in which a portion of a skin is exposed.

The image of FIG. 4A is analyzed by the color analyzer 204, and the color analyzer 204 outputs color information of the corresponding image that is a result of the analysis.

FIG. 4B illustrates a color gamut image of the color gamut generated by the color gamut determination unit 206 based on the color information.

FIG. 4C shows the image after the noise 402 shown in the flesh color region image is removed by the filter unit 302 of the skin region determination unit 208.

FIG. 4D illustrates an image after the connection area 404 having an area less than or equal to a set value is removed by the connection area removal unit 304 of the skin area determination unit 208.

FIG. 4E illustrates a region finally determined as the skin region 406 by the connection region removing unit 304 of the skin region determination unit 208. The skin index calculator 210 determines the skin index based on the image shown in FIG. 4E.

Referring back to FIG. 1, the text adult index determiner 106 determines a text adult index indicating the degree of adult term relevance of the input text collected from the host.

Hereinafter, the configuration and operation of the text adult index determiner 106 will be described with reference to FIGS. 5 through 8.

5 is a schematic block diagram illustrating a text adult index determiner 106 according to an embodiment of the present invention. The text adult index determiner 106 determines a webpage adult index determiner 502 and a UEL adult index determiner. A unit 504 and a text adult index calculator 506 are included.

The webpage adult index determination unit 502 determines the webpage adult index by analyzing input text extracted from a webpage document, a BBS, an image post or a video post provided by a host. Here, the webpage adult index determination unit 502 may directly extract the input text from the host or receive the extracted input text.

FIG. 6 is a schematic block diagram illustrating a webpage adult index determiner 502 according to an embodiment of the present invention. The webpage adult index determiner 502 may include a second training set 602 and a first candidate term. The selection unit 604, the first dictionary generation unit 606, and the webpage adult index calculation unit 608 are included. Hereinafter, the input text will be described as text included in a web page document operated by the host.

The second training set 602 consists of a plurality of adult webpage documents and non-adult webpage documents. Here, the adult webpage document and the non-adult webpage document serve as a population necessary to calculate the frequency with which a certain term appears in the adult webpage document or the non-adult webpage document. In addition, adult webpage documents and adult webpage documents may be randomly collected from a plurality of hosts on the Internet.

The first candidate term selector 604 selects the first candidate term from the second training set 602. Here, the first candidate term is selected from terms included in the adult webpage document and the non-adult webpage document.

In an embodiment, the first candidate term may be a word subtracted from a word included in the adult webpage document or a non-adult webpage document and given a general word such as an article.

The first dictionary generation unit 606 generates mapping information of frequency scores determined according to the frequency of appearance of the first candidate term and the first candidate term in an adult webpage document and the frequency of appearance in an adult webpage. Hereinafter, the mapping information will be referred to as a first dictionary.

For example, when the first candidate term "bikini" exists, the first dictionary generation unit 606 calculates a frequency in which "bikini" appears in the adult webpage document of the second training set 602. The frequency appearing in the adult webpage document is calculated, and the frequency score is calculated using the two frequencies. Then, the first dictionary generation unit 606 adds "bikini" and the corresponding frequency score as entries of the first dictionary. Therefore, mapping information of "bikini" and the corresponding frequency score is generated. An example of the first dictionary is shown in FIG. 7.

가 성인 웹페이지 문서에 출현한 횟수를 구한다. 마찬가지로,

가 비성인 웹페이지 문서에 출현한 횟수를 산출한다.In detail, the method of calculating the frequency score by the first dictionary generation unit 606 first obtains the total number of terms included in the adult webpage document and the total number of terms included in the non-adult webpage document. And the i-th first candidate term

Find the number of times a has appeared in an adult web page document. Likewise,

The number of occurrences of a non-adult web page document is calculated.

Accordingly, the frequency of appearance in the adult webpage document may be calculated through the count and the number information calculated through the above-described process, and the frequency of appearance in the non-adult webpage document may be calculated. Then, the frequency score is obtained using the ratio of the frequencies.

Referring back to FIG. 6, the webpage adult index calculator 608 refers to the mapping information of the first dictionary and uses the frequency score mapped to the term included in the webpage operated by the host to calculate the webpage adult index. Decide In an embodiment, the term included in the webpage operated by the host may be input text extracted from a webpage document, an image post, a video post, or the like of the webpage provided by the host.

In detail, the process of the webpage adult index calculator 608 determining the webpage adult index may be represented by Equation 3 below.

In Equation 3, Adult (D) represents a webpage adult index, and P (Adult | Doc) represents an adult webpage (Adult) when a webpage document (Doc) provided by a host to be inspected is input. Conditional probability. And P (Normal | Doc) represents a conditional probability that the webpage document is an non-web page.

In addition, since Equation 3 derives P (Doc | Adult) P (Adult) / [P (Doc | Normal) P (Normal)], Bayesian theorem (Bayes) for estimating the conditional probability of the current event from prior probability information. 'theorem). In Equation 3, for convenience of calculation, P (Normal) was considered to be the same as P (Adult).

, i는 1에서 n사이의 정수)을 획득하고, 그 용어들이 가지는 빈도 점수들의 합에 의해 결정된다. In the end, Adult (D) refers to terms that the webpage document (Doc) contains from the host (

, i is an integer between 1 and n), and is determined by the sum of the frequency scores of the terms.

에 매핑되는 빈도 점수를 찾는다. 그리고, 매핑 과정을 n번 반복하여 빈도 점수들을 모두 더함으로써 웹페이지 성인 지수를 결정한다. 일 실시예에 있어서, 상기 웹페이지 성인 지수는 0에서 1사이의 값으로 표준화(normalize)된 값일 수 있다.Specifically, with reference to the first dictionary

Find the frequency score that maps to. Then, the mapping process is repeated n times to add all the frequency scores to determine the webpage adult index. In one embodiment, the webpage adult index may be a value normalized to a value between 0 and 1.

As a result, the process of determining the webpage adult index using the above-described first dictionary may be briefly expressed by Equation 4 below.

(i는 1에서 n까지의 정수)는 검수를 위해 호스트가 운영하는 웹페이지로부터 특정 용어(입력 텍스트)를 획득한 경우, 그 특정 용어와 동일한 제1 후보 용어에 매핑되는 빈도 점수이다.In Equation 4,

(i is an integer from 1 to n) is a frequency score mapped to the same first candidate term when the specific term (input text) is obtained from the web page operated by the host for inspection.

Referring back to FIG. 5, the UEL adult index determination unit 504 determines a UEL adult index indicating the relevance of the adult term of the UAL of the host among the input texts acquired by the host. Here, the UEL adult index determination unit 504 may directly extract the UEL of the host or receive the UEL already extracted.

FIG. 8 is a schematic block diagram illustrating a UEL adult index determiner 504 according to an embodiment of the present invention. The UEL adult index determiner 504 includes a third training set 802 and a second candidate term selector. 804, a second dictionary generator 806, and a UEL adult index calculator 808.

The third training set 802 consists of a number of adult UELs and non-UALs. Here, the adult Ualel and the adult Ualel serve as a population necessary to calculate the frequency at which a given Ual appears in the adult Ual or the adult Ualel. In addition, adult Ualel and adult Ualel may be randomly collected from a number of hosts on the Internet. Here, the adult UEL refers to the UAL of the adult host. However, in the modified embodiment, when the adult content is the adult image content, the adult UEL refers to a UAL such as a host or a web page that provides the adult image content.

The second candidate term selector 804 selects a second candidate term from the third training set 802. Here, the second candidate term is selected from the terms included in the UEL of the adult host and the UEL of the adult host.

In one embodiment, the second candidate term may be a word subtracted from a general word such as “WWW” among words included in the UAL of the adult host or the UAL of the adult host.

The second dictionary generation unit 806 generates mapping information of frequency scores determined according to the frequency in which the second candidate term and the second candidate term appear in the UAL of the adult host and the frequency in the UAL of the adult host. Hereinafter, the mapping information will be referred to as a second dictionary. The detailed configuration of the second dictionary can be understood as described above with reference to the webpage adult index and FIG. 7, and a description thereof will be omitted.

In addition, since the process of calculating the frequency score by the second dictionary generator 806 may be understood as described above with respect to the first dictionary generator 606, a detailed description thereof will be omitted.

The UEL adult index calculator 808 determines the UEL adult index by using a frequency score mapped to a term included in the UEL of the host by referring to the mapping information of the second dictionary. Here, the process of determining the UEL adult index by the UEL adult index calculation unit 808 may be understood by the above Equation 3 to Equation 4, a detailed description thereof will be omitted. In one embodiment, the UEL adult index may be a value normalized to a value between 0 and 1.

Referring back to FIG. 5, the text adult index calculator 506 determines the text adult index using the webpage adult index and the UEL adult index. In one embodiment, the text adult index calculator 506 may determine the text adult index according to the maximum value of the webpage adult index and the UEL adult index. This is shown in Equation 5.

Text Adult Index = Max {Webpage Adult Index, EU Adult Index}

Referring back to FIG. 1, the adult index determiner 108 determines the adult index by using the skin index obtained from the skin index determiner 104 and the text adult index obtained from the adult index determiner 106.

In one embodiment, the adult index determiner 108 may determine the adult index according to the product of the skin index and the text adult index.

In one embodiment, the adult index can be represented by the following equation 6, considering the equation (5).

Adult Index = Skin Index × Max {Webpage Adult Index, EU Adult Index}

In one embodiment, the adult host determination system 102 determines a plurality of skin indices and text adult indices for the plurality of input images when there are a plurality of input images obtained from the host. In addition, the adult index determiner 108 may determine a plurality of primary adult indices, and determine an average of the plurality of primary adult indices as the adult index of the final host.

The adult host determiner 110 determines whether the inspected host is an adult host by comparing the adult index with a specific reference value. That is, the adult host determination unit 110 determines whether the host to be inspected is an adult host according to the large and small adult index.

For example, the adult host determination unit 110 determines the inspection target host as an adult host when the adult index is greater than 0.5. The result is then provided to the operator of the search site, portal site, and the like.

9 is a diagram illustrating a result of determining the adult index of the host to be inspected by the adult index determiner 108 according to an embodiment of the present invention.

Referring to FIG. 9, the adult host determining unit 110 may refer to the above results, and refer to the above results, "www.hitmodel.biz", "gxyz.com", and "www.redma.info", which hosts have an adult index greater than 0.5. Can be determined to be an adult host that will block the user's access or an adult host that will perform re-examination.

FIG. 10 is a flowchart illustrating a method for determining adult material according to an embodiment of the present invention. Hereinafter, an adult host determination method will be described with an example that an adult material is an adult host.

First, a skin index indicating an adult image relevance of an input image collected from a host is determined (S1002). In one embodiment, the input image may be collected from at least one of a web page, an image post, and a video post provided by the host.

Next, the text adult index indicating the adult term relatedness of the input text collected from the host is determined (S1004).

In the above-described embodiment, the determining step S1004 for the text adult index is performed after the determining step S1002 for the skin index. In the modified embodiment, the determining step S1002 for the skin index is determined for the text adult index. This may be performed after step S1004, and two steps S1002 and S1004 may be simultaneously performed.

Next, the adult index is determined using the skin index and the text adult index (S1006). In one embodiment, the adult index may be determined according to the product of the skin index and the text adult index.

Next, it is determined whether the host is an adult host according to the adult index (S1008). In one embodiment, it is determined whether the adult index is greater than the specific reference value (= 0.5), and if the skin index is greater than the specific reference value (= 0.5), the inspected host may be determined to be an adult host.

11 is a flowchart specifically showing a determination process (S1002) for the skin index in the adult host determination method according to an embodiment of the present invention.

First, the color of the input image acquired from the inspection target host is analyzed (S1102). In one embodiment, the input image may be extracted from at least one of a web page, an image post, a video post provided by the host.

Next, according to the analysis result, it is determined whether the input image is composed of a color region of a gray system (S1104). If the input image is composed of a gray area color area, the reference value is determined as the skin index (S1106).

Meanwhile, as a result of the analysis, when the input image is not composed of the color gamut of the gray system (S1104), it is determined whether the input image includes the color gamut of the flesh color system (S1108). If the input image does not include the color gamut of the skin color system, the skin index is determined as a default value (S1110). The default value is so small that the host to be inspected is not considered an adult host. However, if the input image includes the color gamut of the skin color system, a color gamut image of the skin color system is obtained from the input image (S1112).

In one embodiment, the flesh line may be defined by a flesh range determined using the first training set.

In one embodiment, the determination of whether the color gamut is included in the flesh color line may include determining whether the relative RGB of the input image according to the analysis result is included in the range of the relative RGB corresponding to the flesh color line. .

Here, the relative RGB of the input image may be a ratio of each RGB of the input image to a value obtained by adding each RGB of the input image, and the relative RGB corresponding to the flesh color system may be calculated in the same manner. In addition, since the relative RGB can be immediately understood through Equation 1, a detailed description thereof will be omitted.

Next, the color gamut image of the skin color system is filtered to remove noise (S1114). In one embodiment, the filtering may include median filtering or average filtering.

Next, an image from which the connection area whose area is less than or equal to the set value is removed from the image from which the noise is removed is obtained (S1116), and the color area of the flesh color line of the image from which the connection area is removed is determined as the skin area (S1118).

In the above-described embodiment, the connection area is removed after removing the noise from the color gamut image of the flesh color system, but in the modified embodiment, the noise may be removed after removing the connection area. In addition, only the connection area may be removed from the color gamut image of the flesh color system without going through the noise removing process (S1114), and only the noise may be removed from the corresponding image without going through the connection area removing process (S1116).

Next, the skin index is determined according to the ratio of the area of the skin area to the area of the input image (S1120). In one embodiment, the skin index may be determined again as a default value when the skin index determined using the area of the skin area is less than or equal to the first threshold or greater than or equal to the second threshold.

12 is a flowchart specifically showing a determination process (S1004) for a text adult index in the method of determining an adult host according to an embodiment of the present invention.

First, a webpage adult index indicating an adult term relation degree of text extracted from a webpage operated by the host among input texts acquired from the host is determined (S1202).

Next, in the input text, the UEL adult index indicating the adult term relevance of the URL of the host is determined (S1204).

In the above-described embodiment, the determining step S1204 for the UEL adult index is performed after the determining step S1202 for the webpage adult index, but in the modified embodiment, the determining step S1202 for the webpage adult index is UAL adult. It may be performed after the step S1204 of determining the index, and the two steps S1202 and S1204 may be simultaneously performed.

Next, the text adult index is determined using the webpage adult index and the UEL adult index (S1206). The text extracted from the web page may be extracted from at least one of a web page document, a BBS, an image post, and a video post of a web page operated by the host.

In one embodiment, the text adult index may be determined by a maximum value of the webpage adult index and the UEL adult index.

FIG. 13 is a flowchart specifically illustrating a determination process (S1202) of a webpage adult index in a method of determining an adult host according to an embodiment of the present invention.

First, a first candidate term is selected from a second training set consisting of an adult web page and an adult web page (S1302).

Next, mapping information of frequency scores determined according to a frequency in which the first candidate term and the first candidate term appear in the adult webpage and the frequency in the non-adult webpage is generated (S1304).

Next, referring to the mapping information, a webpage adult index is determined using a frequency score mapped to a term included in a webpage operated by the host (S1306).

In one embodiment, a first candidate term is found that is identical to a term included in a webpage operated by the host, and a frequency score mapped to the first candidate term is found. The frequency score is then used to determine the webpage adult index. Here, since the process required for calculating the webpage adult index using the frequency score and the frequency score is described above as Equation 3 and Equation 4, description thereof is omitted.

14 is a flowchart specifically illustrating a determination process (S1204) of a UEL adult index in the method of determining an adult host according to an embodiment of the present invention.

First, a second candidate term is selected from a third training set consisting of UALs of an adult host and an adult host (S1402).

Next, mapping information of frequency scores determined according to the frequency of appearance of the second candidate term and the second candidate term in the UAL of the adult host and the appearance of the UAL of the adult host is generated (S1404).

Next, the UEL adult index is determined using the frequency score mapped to the term included in the UEL of the host by referring to the mapping information (S1406). Here, the process required to determine the UEL adult index using the calculation of the frequency score for the UEL adult index and the frequency score can be understood as Equation 3 and Equation 4 will not be described in detail.

Adult content determination method according to an embodiment of the present invention can be implemented in the form of a program that can be performed using a variety of computer means, the program for performing the adult content determination method is a hard disk, CD-ROM, DVD, It is stored in a computer-readable recording medium such as a ROM, a RAM, or a flash memory.

Those skilled in the art to which the present invention pertains will understand that the present invention can be implemented in other specific forms without changing the technical spirit or essential features.

Therefore, it is to be understood that the embodiments described above are exemplary in all respects and not restrictive. The scope of the present invention is shown by the following claims rather than the detailed description, and all changes or modifications derived from the meaning and scope of the claims and their equivalent concepts should be construed as being included in the scope of the present invention. do.

1 is a schematic block diagram of an adult host determination system in accordance with an embodiment of the present invention.

Figure 2 is a schematic block diagram showing a skin index determination unit according to an embodiment of the present invention.

3 is a block diagram schematically illustrating a skin region determiner according to an embodiment of the present invention.

4 is a view showing images processed by the skin index determiner according to an embodiment of the present invention.

5 is a schematic block diagram illustrating a text adult index determiner according to an embodiment of the present invention.

6 is a schematic block diagram showing a webpage adult index determination unit according to an embodiment of the present invention.

7 illustrates an example of a first dictionary according to an embodiment of the present invention.

8 is a schematic block diagram showing a UEL adult index determination unit according to an embodiment of the present invention.

FIG. 9 is a diagram illustrating a result of determining an adult index of a host to be inspected by an adult index determination unit according to an embodiment of the present invention; FIG.

10 is a flowchart illustrating a method of determining an adult host according to an embodiment of the present invention.

11 is a flowchart specifically showing a determination process for a skin index in the method of determining an adult host according to an embodiment of the present invention.

12 is a flow chart showing in detail the determination process for the text adult index in the adult host determination method according to an embodiment of the present invention.

FIG. 13 is a flowchart specifically illustrating a determination process of a webpage adult index in a method of determining an adult host according to an embodiment of the present invention. FIG.

14 is a flow chart showing in detail the determination process for the UEL adult index in the adult host determination method according to an embodiment of the present invention.

<Description of the symbols for the main parts of the drawings>

102: adult host judgment system 104: skin index determination unit

106: text adult index determination section 108: adult index determination section

110: adult host judgment

Claims (26)

Determining a skin index indicating an adult image relevance of the input image collected from the object; Determining a textual adult index indicating an adult term relevance of the input text collected from the object; Determining an adult index using the skin index and the text adult index; And Determining whether the object is an adult object according to the adult index; Determining the text adult index, Determining a webpage adult index indicating an adult term relevance of text extracted from a webpage related to the object among the input texts; Determining a UELEL adult index indicating an adult term relevance of a URL related to the object in the input text; And Determining the text adult index using the webpage adult index and the UEL adult index; Adult content determination method comprising a. The method of claim 1, wherein determining the skin index, Acquiring a color gamut image of the skin color line from the input image when the input image includes a color area of the skin color line; Determining a skin region using the color region image; And Determining a skin index using the area of the skin region; Adult content determination method comprising a. The method of claim 2, wherein the skin index is, Adult content determining method according to the ratio of the area of the skin area to the area of the input image. The method of claim 2, wherein determining the skin index, The method may further include determining whether the input image is comprised of a gray area of a color area. And determining the skin index as a reference value when the input image is determined to be a gray region of a color region. The method of claim 2, wherein the determining of the skin region comprises: Filtering the color gamut image to remove noise; And Determining the skin region using the noise removed image; Adult content determination method comprising a. The method of claim 2, wherein the determining of the skin region comprises: Obtaining an image from which a connected component having an area less than or equal to a predetermined value of the color area image is removed; And Determining the skin region using the image from which the connection region has been removed; Adult content determination method comprising a. The method of claim 2, wherein the determination on whether the color region is included, And determining whether the relative RGB of the input image falls within a range of the relative RGB corresponding to the flesh color line. The method of claim 1, wherein the object, Adult content determination method comprising at least one of the image content, video content, host, blog, web page, website, cafe, mini homepage and BBS (Bulletin Board System). delete The method of claim 1, wherein the text extracted from the web page, Adult content determination method, characterized in that extracted from at least one of the web page document, BBS, image posts, and video posts of the web page. The method of claim 1, wherein the determining of the webpage adult index comprises: Selecting a first candidate term from a second training set consisting of an adult webpage and an adult webpage; Generating mapping information of a first frequency score determined according to a frequency of the first candidate term and the first candidate term appearing on the adult webpage and a frequency of appearance on the non-adult webpage; And Determining the webpage adult index by using a first frequency score mapped to a first candidate term corresponding to text extracted from the webpage with reference to the mapping information; Adult content determination method comprising a. The method of claim 1, wherein the determining of the UEL Adult Index, Selecting a second candidate term from a third training set consisting of the adult UEL and the adult adult UEL; Generating mapping information of a second frequency score determined according to a frequency of the second candidate term and the second candidate term appearing in the UEL of the adult and a frequency of appearing in the UEL of the non-adult; And Determining the UEL adult index by using a second frequency score mapped to a second candidate term corresponding to the UEL related to the object by referring to the mapping information; Adult content determination method comprising a. The method of claim 1, wherein the text adult index is Adult content determination method, characterized in that determined by the maximum value of the web page adult index and the UEL adult index. A computer-readable recording medium having recorded thereon a program for performing the method according to any one of claims 1 to 8 and 10 to 13. A skin index determiner for determining a skin index indicating an adult image relevance of an input image collected from an object; A text adult index determination unit for determining a text adult index indicating an adult term relevance of the input text collected from the object; An adult index determiner that determines an adult index using the skin index and the text adult index; And And an adult material determining unit determining whether the object is adult according to the adult index. The text adult index determination unit, A web page adult index determination unit for determining a web page adult index indicating an adult term relevance of text extracted from a web page related to the object among the input texts; A UEL adult index determination unit which determines a UEL adult index indicating an adult term relation of UAl related to the object in the input text; And A text adult index calculator configured to determine the text adult index using the webpage adult index and the UEL adult index; Adult content determination system comprising a. The method of claim 15, wherein the skin index determiner, A skin color region determiner which obtains a color region image of the skin color system from the input image when the input image includes a color region of a skin color system; A skin region determiner configured to determine a skin region using the color region image; And A skin index calculator configured to determine a skin index using an area of the skin region; Adult content determination system comprising a. The method of claim 16, wherein the skin index calculator, And determine the skin index according to a ratio of the area of the skin region to the area of the input image. The method of claim 16, wherein the skin index calculator, And determining the skin index as a reference value when it is determined that the input image is formed of a gray region of a color region. The method of claim 16, wherein the skin region determiner, A filter unit filtering the color gamut image to remove noise; And A skin region determiner configured to determine the skin region using the image from which the noise is removed; Adult content determination system comprising a. The method of claim 16, wherein the skin region determiner, A connection area removal unit for removing a connection area having an area less than or equal to a set value of the color area image; And A skin region determiner configured to determine the skin region using the image from which the connection region is removed; Adult content determination system comprising a. The method of claim 16, wherein the skin region determiner, And determining whether the color gamut is included or not based on whether the relative RGB of the input image is included in a range of relative RGB corresponding to the flesh color line. The method of claim 16, wherein the skin index determiner, Further comprising a training unit configured to determine a skin color range using a first training set, and to define the skin color line according to the skin color range, The flesh area determining unit determines the inclusion of the color area by the flesh color system defined by the training unit. delete The method of claim 15, wherein the web page adult index determination unit, A first candidate term selecting unit selecting a first candidate term from a second training set consisting of an adult web page and an adult web page; A first dictionary generator configured to generate mapping information of a first frequency score determined according to a frequency of the first candidate term and the first candidate term appearing on the adult webpage and a frequency of appearing on the non-adult webpage; And A webpage adult index calculator for determining the webpage adult index by using a first frequency score mapped to a first candidate term corresponding to text extracted from the webpage, with reference to the mapping information; Adult content determination system comprising a. The method of claim 15, wherein the UEL adult index determination unit, A second candidate term selecting unit for selecting a second candidate term from a third training set consisting of a UEL of adult and a UEL of non-adult; A second dictionary generator for generating mapping information of a second frequency score determined according to a frequency of the second candidate term and the second candidate term appearing in the UEL of the adult and a frequency of appearing in the UEL of the non-adult; And A UEL adult index calculator configured to determine the UEL adult index by using a second frequency score mapped to a second candidate term corresponding to the UAL related to the object by referring to the mapping information; Adult content determination system comprising a. The method of claim 15, wherein the text adult index calculator, And the text adult index is determined according to a maximum value of the webpage adult index and the UEL adult index.

Images