KR102569379B1 - De-identification processing system and method of data based on specific image search - Google Patents

Abstract

The present invention relates to a specific image search-based de-identification processing system and method, and prior to de-identification of image data, a specific image in image data input through a specific image search technique to which an AI image analysis model is applied. It relates to a technique capable of greatly improving processing performance by extracting only image data including , and performing de-identification processing thereon.

Description

De-identification processing system and method of data based on specific image search

The present invention relates to a specific image search-based de-identification processing system and method, and more particularly, to de-identification processing only for image data including a specific image by combining a specific image search technique to which an AI image analysis model is applied. It relates to a specific image search-based de-identification processing system and method capable of performing

As artificial intelligence (AI) technology develops, various services through detection/analysis models using it are increasing.

In particular, it is widely used as a service that detects a specific object based on image recognition of a person's face image or an object such as a license plate.

Of course, it is also used in non-identification technology to protect unique data including personal information such as human face images and license plates.

As an example of a method for detecting a specific object that is widely used in general, first, the characteristics of each object that exists in various forms are analyzed using an AI analysis model suitable for each object to be detected, and the analyzed characteristics are converted into metadata It is converted into a DB in the storage format and stored. Thereafter, the characteristic data of the specific object to be found is compared with the characteristic data of the objects stored in the database, the similarity (concordance) between them is analyzed, and the specific object is detected based on the high similarity.

Non-identification technology is used for data security and personal information protection, and includes deletion, totalization, pseudonymization, categorization, and masking, and a masking method is widely used for image data.

These AI-based specific image recognition or non-identification technologies for specific objects are constantly being developed to improve performance in various service fields.

Using image data such as CCTV, specific image recognition such as suspect tracking, missing person tracking, and hit-and-run vehicle identification is processed, while non-identification processing is performed for data security and personal information protection for other objects. is being demanded However, in the de-identification process for image data so far, a desired specific face or specific object is detected for all image data, processing is performed so that only the detected object can be identified, and the remaining objects are de-identified to make it difficult to identify. Fire processing is being performed.

That is, since analysis and identification/de-identification are individually processed for numerous objects included in all image data, not only is the efficiency of de-identification processing low, but the larger the size of the image data, the greater the de-identification processing. There is a problem that takes a long time.

Domestic Patent Publication No. 10-2021-011299 (published date 2021.09.15.)

The present invention has been devised to solve the problems of the prior art as described above, and an object of the present invention is not to perform de-identification processing on the entire input image data, but to search for a specific image by applying an AI image analysis model. Provides a specific image search-based de-identification processing system and method capable of improving de-identification processing performance by extracting only image data containing a specific image and performing de-identification processing thereon by combining the techniques will be.

In the specific image search based de-identification processing system according to an embodiment of the present invention, an image input unit 100 receiving image data to be subjected to de-identification processing, and transmitting the image data of the image input unit 100 received and extracted from the object extractor 200 using the object extractor 200 that extracts the face image information of a person included in each image frame constituting the image data, and the previously stored AI image analysis model Characteristics analysis unit 300 that analyzes the characteristics of the face image information and specific image information to be searched received from the outside, and the characteristics of the face image information using the characteristics analysis result of the characteristic analysis unit 300 A comparison search unit 400 that performs a comparative search between data and characteristic data of the specific image information, receives video data from the video input unit 100, and applies the result of the comparison search unit 400 to perform de-identification. The image processing unit 500 extracts and stores effective image frames that can be processed, and performs de-identification processing for the image frames extracted by the image processing unit 500 to perform de-identification processing on objects other than the specific image information. It is preferred to include section 600 .

Furthermore, the comparison search unit 400 compares the characteristic data of the face image information with the characteristic data of the specific image information by applying a pre-stored comparison technique, and compares the characteristic data of the specific image information with a predetermined criterion. It is preferable to search for the face image information having .

Furthermore, the comparison search unit 400 compares the characteristic data of the face image information with the characteristic data of the specific image information by applying a plurality of pre-stored comparison techniques, and compares the characteristic data of the specific image information with the predetermined characteristic data. It is preferable to search for each of the face image information having a criterion matching degree.

Furthermore, the image processing unit 500 receives the video data of the image input unit 100 and extracts image frame information including the face image information searched by the comparison search unit 400. 510 and among the video frames constituting the video data of the video input unit 100, the video frame corresponding to the video frame information extracted by the video extraction unit 510 is set as a valid video frame capable of non-identification processing It is preferable to include an image storage unit 520 to store.

Furthermore, the de-identification performer 600 analyzes the valid video frame stored by the image storage unit 520, extracts included object information, and the comparison search unit among the extracted object information. It is preferable to detect and track the face image information searched in step 400, and to perform non-identification processing on object information other than the face image information to be tracked.

In the specific image search-based de-identification processing method in which each step is performed by a specific image search-based de-identification processing system implemented by a computer according to another embodiment of the present invention, in the image input unit, from the image capture module In the image input step (S100) of receiving the image data to be de-identified, the object extractor analyzes the image frames constituting the image data in the image input step (S100) and includes them for each image frame. An object extraction step (S200) of extracting face image information of a person who has been selected and assigning unique identification information to each extracted face image information (S200), and a search information input step of receiving specific image information to be searched from the outside in a characteristic analysis unit. (S300), in the characteristic analysis unit, using the pre-stored AI image analysis model, the characteristics of the face image information by the object extraction step (S200) and the specific image information by the search information input step (S300) In the characteristic analysis step (S400) of analyzing the image, a pre-stored comparison technique is applied in the comparison search unit, and based on the analysis result of the characteristic analysis step (S400), the degree of correspondence between the characteristic data of the specific image information and a predetermined criterion is determined. A search step (S500) of searching for the face image information having received the video data by the image input step (S100) in the image processing unit, and the face image information searched by the search step (S500) is included. In the image reprocessing step (S600) of extracting and storing valid image frames, and in the de-identification performing unit, the search step (S500) searches for the valid image frames stored in the image reprocessing step (S600). It is preferable to include a de-identification processing step (S700) of performing de-identification processing on the remaining objects except for the face image information.

Furthermore, the search step (S500) applies a plurality of pre-stored comparison techniques to compare the characteristic data of the face image information and the characteristic data of the specific image information, respectively, and compares the characteristic data of the specific image information with a predetermined criterion. It is preferable to search for each of the face image information having a degree of agreement of .

Furthermore, the image reprocessing step (S600) receives the image data by the image input step (S100) and extracts valid image frame information including the face image information searched by the search step (S500). a frame extraction step (S610) and a non-identification image storage step (S620) of storing image data for de-identification using only image frames corresponding to the valid image frame information extracted by the frame extraction step (S610). It is desirable to do

Furthermore, the de-identification processing step (S700) analyzes the image frames constituting the image data for de-identification, and extracts object information including human face image information included in each image frame as an effective object. Extraction step (S710), valid object tracking step (S720) of detecting and tracking the face image information searched by the search step (S500) among the object information extracted by the valid object extraction step (S710) and tracking the effective object information It is preferable to include a de-identification step (S730) of performing de-identification processing on object information other than the face image information tracked by the object tracking step (S720).

The specific image search-based de-identification processing system and method of the present invention according to the above configuration do not perform de-identification processing on the entire input image data, but prior to performing the de-identification processing, the whole image Data is processed as much as the number of image data frames used for de-identification processing is reduced by performing a search for a specific object desired for data and performing an operation for de-identification processing only for some image frames in which a specific object is searched for. It has the advantage of increasing the efficiency of

That is, since analysis for de-identification processing is performed only on image data including a desired specific object among all image data, de-identification processing can be efficiently performed.

Through this, since the capacity of image data for de-identification is processed with a small capacity and then de-identification is performed, there is an advantage in that processing time can be greatly reduced.

1 is an exemplary configuration diagram illustrating a de-identification processing system based on a specific image search according to an embodiment of the present invention.
2 is another exemplary configuration diagram illustrating a specific image search-based de-identification processing system according to an embodiment of the present invention.
3 is a flowchart illustrating a method for processing non-identification based on a specific image search according to an embodiment of the present invention.

Hereinafter, a specific image search based de-identification processing system and method of the present invention will be described in detail with reference to the accompanying drawings. The drawings introduced below are provided as examples to sufficiently convey the spirit of the present invention to those skilled in the art. Accordingly, the present invention may be embodied in other forms without being limited to the drawings presented below. Also, like reference numerals denote like elements throughout the specification.

At this time, unless there is another definition in the technical terms and scientific terms used, they have meanings commonly understood by those of ordinary skill in the art to which this invention belongs, and the gist of the present invention in the following description and accompanying drawings Descriptions of well-known functions and configurations that may be unnecessarily obscure are omitted.

In addition, a system refers to a set of components including devices, mechanisms, and means that are organized and regularly interact to perform necessary functions.

A specific image search-based de-identification processing system and method according to an embodiment of the present invention improves processing performance by analyzing image data and performing de-identification processing on objects other than a specific object. It's about technology that can do it.

1 and 2 are configuration example diagrams showing a specific image search-based de-identification processing system according to an embodiment of the present invention, with reference to FIGS. 1 and 2, specific image search according to an embodiment of the present invention The base de-identification processing system will be described in detail.

As shown in FIG. 1, the specific image search-based de-identification processing system according to an embodiment of the present invention includes an image input unit 100, an object extraction unit 200, a characteristic analysis unit 300, a comparison search unit ( 400), an image processing unit 500, and a de-identification performing unit 600. It is preferable that each of the components perform an operation by being included individually or in combination with at least one calculation processing unit including a computer.

For a detailed look at each component,

The image input unit 100 receives image data to be subjected to non-identification processing from an associated image capture means (for example, CCTV, etc.).

At this time, the image input unit 100 may receive the image data from a single image capture unit or a plurality of image capture units.

Preferably, the object extractor 200 receives the video data of the video input unit 100 and extracts human face image information included in each video frame constituting the video data.

The object extractor 200 extracts the face image information included in each image frame constituting each image data from all the image data received from the image input unit 100, and stores it in advance for this purpose. It is preferable to use a separate extraction algorithm that is already provided.

The object extractor 200 extracts one or more face image information, and it is preferable to store and manage by assigning corresponding video frame information and unique identification information (ID, etc.) to each extracted face image information.

The specific image search-based de-identification processing system according to an embodiment of the present invention is limited to extracting human face image information from the object extractor 200 for smooth explanation, but this is only one embodiment. Depending on the settings, the object extractor 200 may extract requested object information (for example, a license plate number, etc.) instead of the face image information of a person included in each image frame constituting the image data. there is.

The characteristic analysis unit 300 analyzes the characteristics of the face image information extracted by the object extraction unit 200 using a pre-stored AI image analysis model. In addition, the characteristic analysis unit 300 analyzes the characteristics of specific image information to be searched received from the outside using a pre-stored AI image analysis model. At this time, the characteristic analysis unit 300 analyzes the characteristics of the face image information and the specific image information using the same AI image analysis model, thereby managing image data (face image information) and the image to be searched for. Data (specific image information) is analyzed through the same analysis engine.

In detail, the characteristic analysis unit 300 analyzes the characteristics of the face image information extracted by the object extractor 200, and stores and manages the characteristics by matching with the unique identification information assigned to each face image information. It is desirable to do

In addition, when the specific image information to be searched input from the outside is input, it is preferable to receive one or more specific image information, and input along with separate unique identification information for smooth management as a plurality of specific image information is input. It becomes. It is desirable to store and manage the characteristic data analyzed based on this by matching it with unique identification information assigned to each specific image information.

In addition, the AI image analysis model used in the characteristic analysis unit 300 inputs image data and outputs the characteristic is a conventional technique that is widely used, and a detailed description thereof will be omitted.

Preferably, the comparison search unit 400 compares and searches between the characteristic data of the face image information and the characteristic data of the specific image information using the characteristic analysis result of the characteristic analysis unit 300 .

In detail, the comparison search unit 400 compares the characteristic data of the face image information with the characteristic data of the specific image information by applying a pre-stored comparison technique, and compares the characteristic data of the specific image information with a predetermined criterion. It is desirable to search for the face image information having a matching degree.

That is, it is preferable that the comparison search unit 400 compares the characteristic data of the face image information with the characteristic data of the specific image information to be searched for, and calculates the degree of coincidence (similarity). At this time, although a cube extension technique is applied as an embodiment of the comparison technique, this is only an embodiment and is not limited thereto.

Through this, using the calculated matching analysis result, the face image information having a matching degree with the characteristic data of the specific image information and a predetermined criterion is searched, and the search result is determined as corresponding unique identification information together with the matching degree. do.

Here, the comparison search unit 400 may detect as many as a preset number of matches by arranging the matches in descending order, or may set a threshold and detect all image data having a match equal to or higher than the threshold. . This may be different according to the setting of an external user (administrator, etc.), but is not limited thereto.

In this case, depending on the number of face image information extracted by the object extraction unit 200 and the number of specific image information input by the characteristic analysis unit 300, it may take a lot of time for comparative analysis. there is.

At this time, as shown in FIG. 2, the comparison search unit 400 applies a plurality of pre-stored comparison techniques to compare characteristic data of the face image information and characteristic data of the specific image information, respectively, It is also possible to search for the face image information having a degree of agreement between characteristic data of specific image information and a predetermined criterion.

Through this, it is possible to reduce the time for each comparative analysis by configuring an engine that performs the comparison technique in parallel according to the number of the specific image information to be searched for. That is, the matching calculation time can be reduced through comparison and matching with characteristic data of the face image information for specific image information to be searched for.

Preferably, the image processor 500 receives the video data of the image input unit 100, extracts and stores effective image frames capable of non-identification processing by applying the results of the comparison and search unit 400. .

To this end, the image processing unit 500 is preferably configured to include an image extraction unit 510 and an image storage unit 520 as shown in FIGS. 1 and 2 .

The image extractor 510 receives the video data (whole image data) of the image input unit 100 and extracts image frame information including the face image information searched by the comparison search unit 400. it is desirable

In detail, the image extraction unit 510 uses image frame information including the face image information having a degree of agreement with the specific image information to be searched, which is a search result of the comparison search unit 400. Accordingly, it is preferable to receive the video data of the video input unit 100 and extract corresponding video frame information.

At this time, the image extraction unit 510 uses the corresponding image frame information stored and managed for each face image information in the object extraction unit 200 to match the face in the comparison search unit 400. It is desirable to analyze video frame information of image information.

The image storage unit 520 is a valid image capable of de-identifying an image frame corresponding to the image frame information extracted by the image extractor 510 among image frames constituting the image data of the image input unit 100. It is preferable to set it as a frame and save it.

That is, it is preferable that the image storage unit 520 extracts some image frames from the entire image data input by the image input unit 100 and newly stores and manages them as image data for de-identification processing.

At this time, the image storage unit 520 derives an image frame corresponding to the image frame information extracted by the image extraction unit 510, and newly stores and manages it. Through this, the image storage unit 520 Through this, all image frames for newly stored and managed image data include face image information having a degree of agreement with the specific image information to be searched for and a predetermined criterion.

It is preferable that the de-identification performer 600 performs de-identification processing on the remaining objects except for the specific image information on the image data newly stored and managed by the image processor 500.

In other words, de-identification processing means leaving only the specific image information (or specific object) to be searched for and performing de-identification processing so that identification of all other objects is difficult or impossible.

Therefore, the de-identification performer 600 receives image data newly stored and managed through the image storage unit 520, and the matching degree of the specific image information to be searched for in all image frames and a predetermined criterion. It is preferable to perform de-identification processing on all remaining objects except for the corresponding face image information while tracking the face image information having . Typically, masking is performed to de-identify image data, but this is only an example and is performed according to input de-identification.

In detail, the de-identification performer 600 analyzes the image data stored and managed by the image storage unit 520, that is, the valid image frame, and analyzes the valid image frame included in the effective image frame. All object information is extracted.

Thereafter, among the extracted object information, the face image information (the face image information having a degree of agreement with the specific image information to be searched for and a predetermined standard) searched by the comparison search unit 400 is detected and tracked, It is desirable to perform de-identification processing on the excluded object information.

3 is a flowchart illustrating a method for processing non-identification based on a specific image search according to an embodiment of the present invention. Referring to FIG. 3 , a de-identification processing method based on a specific image search according to an embodiment of the present invention will be described in detail.

As shown in FIG. 3, the specific image search-based de-identification processing method according to an embodiment of the present invention includes an image input step (S100), an object extraction step (S200), a search information input step (S300), and characteristic analysis. It is preferable to include a step (S400), a search step (S500), an image reprocessing step (S600), and a de-identification processing step (S700). In addition, each step is performed by a specific image search-based de-identification processing system implemented by a computer.

For a detailed look at each step,

In the image input step (S100), the image input unit 100 receives image data to be subjected to de-identification processing from an associated image photographing means. In this case, the image data may be received from a single image capture unit or a plurality of image capture units.

In the object extracting step (S200), the object extracting unit 200 extracts human face image information included in each image frame constituting the image data by the image input step (S100).

At this time, the object extraction step (S200) extracts the face image information included in each image frame constituting each image data with respect to all the image data transmitted by the image input step (S100). For this purpose, a separate pre-stored extraction algorithm is used.

In addition, in the object extraction step (S200), one or more face image information is extracted, and corresponding image frame information and unique identification information (ID, etc.) are assigned to each extracted face image information to be stored and managed.

In the search information input step (S300), the characteristic analysis unit 300 receives input of specific image information to be searched for from the outside. Here, when the specific image information to be searched input from the outside is input, it is preferable to receive one or more specific image information, and as much as a plurality of specific image information is input, together with separate unique identification information for smooth management. will be entered

The characteristic analysis step (S400) uses the AI image analysis model stored in advance in the characteristic analysis unit 300, the face image information by the object extraction step (S200), and the search information input step (S300). ) to analyze the characteristics of specific image information. In the feature analysis step (S400), the face image information by the object extraction step (S200) and the specific image information by the search information input step (S300) use the same AI image analysis model to analyze the characteristics , The image data being managed (face image information) and the image data to be searched for (specific image information) go through the same analysis engine to perform characteristic analysis. In addition, the AI image analysis model used is a conventional technique in which image data is input and characteristics are output, and a detailed description thereof is omitted.

In detail, the characteristic analysis step (S400) analyzes the characteristics of the face image information by the object extraction step (S200), and then stores and manages matching with unique identification information assigned to each face image information. it is desirable

Of course, the characteristic data analyzed for the specific image information in the search information input step (S300) is matched with the unique identification information assigned to each specific image information to be stored and managed.

In the search step (S500), the pre-stored comparison technique is applied in the comparison search unit 400, and based on the analysis result of the feature analysis step (S400), the degree of correspondence between the characteristic data of the specific image information and a predetermined criterion. The face image information having a is searched for.

In detail, the search step (S500) compares the characteristic data of the face image information with the characteristic data of the specific image information to be searched to calculate the degree of correspondence (similarity). At this time, one of the comparison techniques As an embodiment, a cube extension technique is applied, but this is only an embodiment and is not limited thereto.

Through this, using the calculated matching analysis result, the face image information having a matching degree with the characteristic data of the specific image information and a predetermined criterion is searched, and the search result is determined as corresponding unique identification information together with the matching degree. do.

Here, as the degree of agreement of the predetermined criterion, a predetermined number of points of agreement may be detected by arranging the degree of agreement in descending order, or a threshold value may be set to detect all image data having a degree of agreement greater than or equal to the threshold value. This may be different according to the setting of an external user (administrator, etc.), but is not limited thereto.

At this time, the search step (S500) applies a plurality of pre-stored comparison techniques to compare the feature data of the face image information and the feature data of the specific image information, respectively, and compares the feature data of the specific image information with a predetermined criterion. It is also possible to search for the face image information having a degree of agreement of .

Through this, it is possible to reduce the time for each comparative analysis by configuring an engine that performs the comparison technique in parallel according to the number of the specific image information to be searched for. That is, the matching calculation time can be reduced through comparison and matching with characteristic data of the face image information for specific image information to be searched for.

In the image reprocessing step (S600), the image processing unit 500 receives the image data by the image input step (S100), and the face image information searched by the search step (S500) is included. Preferably, the image frame is extracted and stored.

In detail, as shown in FIG. 3, the image reprocessing step (S600) includes a frame extraction step (S610) and an image storage step (S620).

In the frame extraction step (S610), the image processing unit receives the image data from the image input step (S100) and extracts a valid image frame including the face image information searched by the search step (S500). will do

In other words, the frame extraction step (S610) uses the image frame information including the face image information having a matching degree with the specific image information to be searched, which is a search result by the search step (S500). , the corresponding image frame information is extracted. At this time, the image frame information of the face image information matched by the search step (S500) is analyzed using the corresponding image frame information stored and managed for each face image information.

In the image storage step (S620), the image processing unit 500 receives the image data by the image input step (S100) and receives the image frame corresponding to the image frame information extracted by the frame extraction step (S610). is set as an effective image frame capable of non-identification processing and stored.

Through this, some image frames are extracted from the entire image data received in the image input step (S100), and are newly stored and managed as image data for de-identification processing. In addition, all image frames of the image data that are newly stored and managed include face image information having a degree of agreement between the specific image information to be searched for and a predetermined criterion.

In the de-identification processing step (S700), in the de-identification performing unit 600, the face searched by the search step (S500) for the valid image frame stored by the image reprocessing step (S600). De-identification processing is performed on the remaining objects except for the image information.

In other words, de-identification processing means leaving only the specific image information (or specific object) to be searched for and performing de-identification processing so that identification of all other objects is difficult or impossible.

The de-identification processing step (S700) receives the image data newly stored and managed by the image storage step (S620), and the face having the specific image information to be searched for in all image frames and the degree of agreement with a predetermined criterion. While tracking the image information, it is desirable to perform de-identification processing on all objects other than the corresponding face image information. Typically, masking is performed to de-identify image data, but this is only an example and is performed according to input de-identification.

To this end, the de-identification processing step (S700) includes a valid object extraction step (S710), a valid object tracking step (S720), and a de-identification step (S730), as shown in FIG.

The effective object extraction step (S710) analyzes the image data newly stored and managed by the image storage step (S620), that is, the effective image frame, and extracts all object information included in the effective image frame. will do

The valid object tracking step (S720) includes the face image information (the specific image information to be searched for and a predetermined criterion) searched by the search step (S500) among the object information extracted by the valid object extraction step (S710). The face image information having a matching degree) is detected and tracked.

In the de-identification step (S730), de-identification processing is performed on object information other than the face image information tracked by the effective object tracking step (S720).

Through this, the specific image search-based de-identification processing system and method according to an embodiment of the present invention does not perform de-identification processing on the entire input image data, but prior to performing the de-identification processing , The number of image data frames used for de-identification processing is reduced by performing a search for a desired specific object on all image data and performing an operation for de-identification processing only on some image frames in which the specific object is searched for. It has the advantage of increasing the efficiency of data processing as much as possible.

As described above, the present invention has been described with specific details such as specific components and limited embodiment drawings, but this is only provided to help a more general understanding of the present invention, and the present invention is not limited to the above embodiment. No, and those skilled in the art to which the present invention pertains can make various modifications and variations from these descriptions.

Therefore, the spirit of the present invention should not be limited to the described embodiments, and not only the scope of the claims described later, but also all modifications equivalent or equivalent to the scope of the claims belong to the scope of the scope of the present invention. .

100: Video input view TT
200: object extraction unit
300: characteristic analysis unit
400: comparison search unit
500: image processing unit
510: image extraction unit 520: image storage unit
600: de-identification unit

Claims (9)

an image input unit 100 that receives image data to be de-identified;
an object extraction unit 200 that receives the video data of the video input unit 100 and extracts face image information of a person included in each video frame constituting the video data;
A characteristic analysis unit 300 that analyzes the characteristics of each of the facial image information extracted by the object extraction unit 200 and the specific image information to be searched received from the outside using a pre-stored AI image analysis model ;
a comparison search unit 400 that performs a comparative search between characteristic data of the face image information and characteristic data of the specific image information using the characteristic analysis result of the characteristic analyzer 300;
Image frame information including face image information having a degree of agreement with characteristic data of the specific image information according to a result of performing the comparison search unit 400 after receiving the video data of the video input unit 100. an image processing unit 500 for extracting and storing valid image frames capable of de-identification processing only for ; and
All object information included in the valid image frame extracted by the image processing unit 500 is extracted, an object corresponding to the specific image information is detected from all extracted object information, and the remaining objects are de-identified. Non-identification performing unit 600 to perform;
Including,
The characteristic analysis unit 300
At least one or more specific image information is input,
The comparison search unit 400
Depending on the number of specific image information to be compared,
In the case of one specific image information, a pre-stored comparison technique is applied to compare the characteristic data of the face image information with the characteristic data of the specific image information, and the characteristic data of the specific image information matches the predetermined criterion. search face image information,
In the case of two or more specific image information, a plurality of pre-stored comparison techniques are applied in parallel according to the number of specific image information to compare the characteristic data of the face image information with the characteristic data of the specific image information, respectively. A specific image search-based de-identification processing system that searches for the face image information having a matching degree of characteristic data of the specific image information and a predetermined criterion.
delete delete According to claim 1,
The image processing unit 500
Receives video data from the video input unit 100 and extracts video frame information including the face image information having a degree of agreement with the characteristic data of the specific image information searched by the comparison search unit 400 and a predetermined standard. An image extraction unit 510 to do; and
Image storage that sets and stores an image frame corresponding to the image frame information extracted by the image extractor 510 among image frames constituting the image data of the image input unit 100 as an effective image frame capable of non-identification processing. section 520;
Including, a specific image search-based de-identification processing system.
According to claim 4,
The de-identification performing unit 600
The effective image frame stored by the image storage unit 520 is analyzed, included object information is extracted, and among the extracted object information, the face image information searched for by the comparison search unit 400 is detected. A specific image search-based de-identification processing system that tracks and performs de-identification processing on object information other than the face image information to be tracked.
In the specific image search-based de-identification processing method in which each step is performed by a specific image search-based de-identification processing system implemented by a computer,
An image input step of receiving image data to be subjected to non-identification processing from an image capturing module in an image input unit (S100);
The object extraction unit analyzes the image frames constituting the image data in the image input step (S100), extracts the face image information of a person included in each image frame, and unique identification information for each extracted face image information. Object extraction step of giving (S200);
Search information input step of receiving at least one or more specific image information to be searched from the outside in the characteristic analysis unit (S300);
In the characteristic analysis unit, the characteristics of each of the face image information by the object extraction step (S200) and the specific image information by the search information input step (S300) are analyzed using a pre-stored AI image analysis model. Characteristics analysis step (S400);
In the comparison search unit, according to the number of specific image information to be compared and searched, an engine configuration for performing a pre-stored comparison technique is applied differently, and based on the analysis result of the characteristic analysis step (S400), the specific A search step (S500) of searching for the face image information having a degree of correspondence between characteristic data of the image information and a predetermined criterion;
In the image processing unit, the image data received by the image input step (S100) is received, and the face image information having the characteristic data of the specific image information searched by the search step (S500) and a predetermined criterion is included. An image reprocessing step of extracting and storing only valid image frames capable of non-identification processing for image frame information (S600); and
In the de-identification unit, all object information included in the valid image frame stored by the image reprocessing step (S600) is extracted, and an object corresponding to the specific image information is detected from all extracted object information. , a de-identification processing step (S700) of performing de-identification processing on objects other than the face image information searched by the search step (S500);
Including,
The search step (S500)
In the case of one specific image information, a pre-stored comparison technique is applied to compare the characteristic data of the face image information with the characteristic data of the specific image information, and the characteristic data of the specific image information matches the predetermined criterion. search face image information,
In the case of two or more specific image information, a plurality of pre-stored comparison techniques are applied in parallel according to the number of specific image information to compare the characteristic data of the face image information with the characteristic data of the specific image information, respectively. A specific image search-based de-identification processing method, wherein each of the face image information having a matching degree of characteristic data of the specific image information and a predetermined criterion is searched for.
delete According to claim 6,
The image reprocessing step (S600)
Effective video frame information including the face image information that receives the video data by the video input step (S100) and matches the characteristic data of the specific image information searched by the search step (S500) with a predetermined criterion. Frame extraction step of extracting (S610); and
a de-identification image storage step (S620) of storing image data for de-identification using only the image frames corresponding to the valid image frame information extracted by the frame extraction step (S610);
Including, a specific image search-based de-identification processing method.
According to claim 8,
The de-identification processing step (S700)
a valid object extraction step (S710) of analyzing image frames constituting the image data for de-identification and extracting object information including human face image information included in each image frame;
a valid object tracking step (S720) of detecting and tracking the face image information searched by the search step (S500) among the object information extracted by the valid object extracting step (S710); and
a de-identification step (S730) of performing de-identification processing on object information other than the face image information tracked by the valid object tracking step (S720);
Including, a specific image search-based de-identification processing method.

Images