KR20100055601A - Security system for fugitive prevention - Google Patents

Abstract

PURPOSE: A security system for fugitive prevention is provided to prevent a suspect or a criminal from run away by comparing an image of a person entering the gateway with an image file stored in a memory in advance. CONSTITUTION: An image gathering unit 1(11) collects a face image. A face image detection unit(12) changes a collected image into a digital image signal. The face image detection unit detects an element of the face. An image revision part(13) eliminates a noises of the face image. A characteristic detecting part(14) extracts the characteristic portion of the face image. A registration part(15) inputs the registration of an extraction image. A database(16) stores the extraction image as a digital file. An image collection unit(17) collects the image of the person entering the gateway. A comparison unit(18) compares the image with the extracted image of the person entering the gateway. An entrance control unit(19) blocks the door(D).

Description

Security system for fugitive prevention

The present invention relates to a security system for preventing escape, and more particularly, a escape that prevents a suspect, criminal or prisoner from escaping from a certain area such as a police station, a court, a prosecution, a prison, etc. It relates to a security system for prevention.

In general, a security system is a security system that restricts access to a specific area or restricted area other than a specific authorized person.

The conventional technology of such a security system includes a security system using human recognition, which includes a sample data such as a fingerprint, a voice, an iris, a face, a hand vessel, and a body provided by a user. It is a security system that allows access by opening and closing the door when some data is compared and matched through a recognition window (recognition sensor).

This conventional security system is used to prevent unauthorized outsiders from entering into a specific area or restricted area.

The conventional security system as described above can prevent unauthorized intruders from entering unauthorized areas or restricted areas, but there is a problem that prevents unauthorized escape.

For example, when a police officer interrogates a suspect (suspect) at a police station, if a suspect escapes through a neglected official, there is a problem to allow escape without defense, which is the same in courts, prosecutors, prisons and the like.

As described above, the conventional security system is secured to a certain extent to prevent intrusion, but a security system capable of preventing unauthorized escape, that is, escapes because it is a half-security system that is defenseless to an unauthorized departure.

The present invention provides an image collection unit 1 for collecting a face image; A face image detection unit converting the collected image into a digital image signal and detecting components of a face through a preset algorithm; An image compensator for removing and correcting noise of the detected face image; A feature extractor for extracting feature parts from the corrected face image; A registration unit for inputting an identity inquiry and blood content, and inputting registration of extracted images; A database for storing the extracted image input from the registration unit as a digital file; Doors; An image collecting unit 2 installed at the door to collect an image of the person in real time; A comparison unit comparing the collected image of the accessor with the extracted image previously stored in the database through a preset program; An access control unit to block the door if there is a matched image through the comparison; As shown in FIG.

The present invention compares the image of the person collected from the door and the pre-stored registered image file, if it is matched with the access control needs to block the door, it is possible to prevent the suspect or criminal to escape.

In addition, the present invention can be made natural and quick access control even while walking without frequent authentication procedures through human recognition or input at the entrance door frequently.

In addition, the present invention can keep the personal hygiene clean access control is made without contacting the human body information collection device for human body recognition.

1 is a system block diagram showing a system configuration according to an embodiment of the present invention, and FIG. 2 is a system configuration according to an embodiment of the present invention. More specifically, it is a system block diagram, and FIG. 3 is a sequence block diagram showing a procedure of a system according to an embodiment of the present invention, and FIGS. 7 is an operational state diagram showing an allowable state, and FIG. 7 is a sectional view showing a configuration of another embodiment related to a door of the present invention.

The present invention will be described below with reference to the accompanying drawings presented as above. First, the present invention, as shown in Figure 1 and 2 of the accompanying drawings, the image collecting unit 1 (11) for collecting a face image; A face image detector 12 for converting the collected image into a digital image signal and detecting components of a face through a preset algorithm; An image corrector 13 for removing and correcting noise of the detected face image; A feature extractor 14 for extracting feature portions from the corrected face image; A registration unit 15 for inputting identification and blood content, and inputting registration of extracted images; A database 16 for storing the extracted image input from the registration unit 15 as a digital file; Door D; An image collecting unit 2 (17) installed at the door D to collect an image of the person in real time; A comparison unit 18 for comparing the collected image of the visitor with the extracted image previously stored in the database 16 through a preset program; If there is a matching image through the comparison, the access control unit 19 for blocking the door (D); may be configured to include.

The image collecting unit 1 (11) of the present invention may be a web camera (including a web cam) connected to the registration unit 15, or a CCTV camera (Closed-Circuit Television camera; closed circuit television camera). It may also be a CCD-charged device (Charged Coupled Device Camera).

In addition, the image collection may be image capturing, in this case, the image to be photographed may be a whole body image, a face image, or may be an image taken from multiple angles (front, side, 45 degrees). In addition, the image may be a moving image, may be a still image, or may be a combined image of a moving image and a still image.

On the other hand, the face image detection unit 12 of the present invention comprises image conversion means (12a) for converting the collected analog image to a digital method; And a face detection unit 12b for detecting a face part based on the eyes in the digitally converted image signal.

In this case, the image converting means 12a may be an A-D converter 12a, may be a digital video recorder (DVR), or may be a video encoder.

The image converting means 12a may convert the input image into NTSC (National Television System Committee) or PAL (Phaser Alternation Line). It can be converted into a composite method (a mixed method; a signal in which several signals are mixed into a single unit. A composite video signal or a composite color signal generally used in television transmission and reception or some computer monitors). Time to live method) can be converted, and according to the image collecting unit 1 (11) can be converted in any way.

In addition, the face detection unit 12b may detect a face region using a cascaded face classifier using an AdaBoost (Adaptive Boost Learning Filter) algorithm.

For reference, since the conventional face detection method uses motion, a person must move or a color camera must be used when using the skin color extraction method in the color information, and is sensitive to light change, race, and skin color.

The method developed to solve this problem is Cascaded Face Classifier using AdaBoost. The Cascaded Face Classifier algorithm using AdaBoost requires a complex decision boundary for class classification when the change in the class is very severe.

The AdaBoost algorithm is a classifier learning algorithm that is suitable for this situation and generates a strong classifier by combining several weak classifiers.

Since the Cascaded Face Classifier algorithm using AdaBoost does not use motion information and color information, a face detection function can be performed at high speed without using any restrictions on a user in a general monochrome camera.

At this time, the face image detection based on the eye is applied as a standard for normalizing the face image by finding the position of the eye in the face area detected by the face image detecting means 12b, but using conditions such as wearing glasses, face rotation, and eye closure. Multi-Block Matching (MMF) is used to accurately locate and track the eye position.

On the other hand, in the present invention, the image compensator (13) comprises: histogram smoothing means (13a) for converting the histogram of the input image into a uniform distribution; Gamma correction means (13b) for alleviating non-uniform illumination conditions of the histogram smoothed image in the histogram smoothing means (13a); And a binner filter 13c for removing noise of the gamma-corrected image from the gamma correction means 13b.

In this case, the histogram smoothing means 13a converts the histogram of the given input image into an equal distribution, and uses the cumulative distribution function of the given histogram as the conversion function.

In general, the histogram smoother 13a is used to improve the contrast of the image, but since the histogram smoothing does not reflect important local facial features in face recognition, uneven illumination, especially in one direction In the case of an illuminated image, the image is damaged enough to be very fatal for face recognition in a dark area.

Further, the gamma correction means 13b normalizes all pixel values of the histogram smoothed image to [0,1], converts them, and then restores them to the range of the original pixel values.

The gamma correction of the image by the gamma correction means 13b plays an important role in mitigating non-uniform lighting conditions.

In addition, the wiener filter 13c removes noise from the gamma corrected image. There is a lot of noise in the face image, especially in the dark area, the effect of noise is large, which acts as a barrier to face recognition.

Therefore, the step of removing noise is essential, and it is very important to select a suitable filter for this.

The binner filter 13c is an adaptive filter that reflects local characteristics of a given image as a type of linear filter. The binner filter 13c may recognize noise statistically and experimentally for an image and adaptively design an appropriate filter accordingly.

On the other hand, in the present invention, the feature extraction unit 14 includes a log filter (14a) for extracting the facial features of the image from which the noise is removed from the binner filter (13c); And a contrast correction means 14b for improving the contrast of the facial features extracted by the log filter 14a.

At this time, the log filter 14a calculates LoG mask and the image passing through the binner filter 13c and extracts facial features from the face image from which the noise is removed through the binner filter 13c. In addition, the contrast correction means 14b extends and maps the histogram of the original image over the entire range.

The histogram of the original image in which the illumination is biased in one direction tends to be biased to one side rather than distributed over the entire range, and the contrast correction means 14b extends and maps the histogram of the original image to the full range.

On the other hand, in the present invention, the registration unit 15 is a personal computer connected to the LAN via the database 16 and the management server 20, the document creation and editing means for providing a document creation sheet (15a); And storage form means 15b for storing the image input from the image collecting unit 1 11 according to a standardized form.

At this time, the document creation and editing means 15a may be a word processor.

In addition, the storage form means 15b may be a program for storing and storing the registration unit 15 itself or providing a standardized storage form and a storage form which are set in advance to be stored in the database 16 through a LAN.

In addition, the database 16 of the present invention may be a lower storage server connected to the management server. On the other hand, the door D in the present invention is a rotating door (Da) of the "X" shape; It may be configured to include; a blocking space (Db) for providing a space as much as the radius of rotation on both sides of the rotary door (Da).

The rotary door Da and the blocking space Db may be formed in the form of "(X)". At this time, the rotary door Da is normally positioned as "X" and stops by rotating 180 degrees when entering and exiting, but when the access control is made, the rotating door Da may be rotated by only 90 degrees to confine the person in the blocking space Db.

In addition, the door (D) may be a sliding automatic door. In addition, the door (D) may be an E.M lock (Electro Magnetic Lock; electric lock), it may be a door including a door lock (number key), a card reader (RFID card or smart card), a human body reader.

On the other hand, the image acquisition unit 2 (17) in the present invention may be a CCTV camera, a CCD camera, may be an infrared camera (separate use at night time off).

In addition, the image collection may be image capture. In this case, the image collecting unit 2 (17) may further include the above-described face image detector 12, the image correction unit 13 and the feature extractor 14, the components are as described above specific Description is omitted.

On the other hand, the comparison unit 18 of the present invention by sequentially loading the extracted image stored in the database 16 to compare with the image of the accessor image input in real time from the image acquisition unit 2 (17) to compare It can be an analysis algorithm or a comparison circuit. In this case, the comparison analysis algorithm may compare the effective feature number for recognizing about 200 to 3000 features of the facial feature extracted by the feature extractor 14.

Effective features for the recognition may be, for example, changes in the contours of the face, shadows and distances and lengths around the eyes and eyebrows, nose, mouth, and the like.

In addition, the comparison circuit may be a logic circuit or an OP-AMP that compares the degree of matching by inputting the difference of the shades by 0 and 1 for each pixel.

The comparison unit 18 as described above sequentially loads and compares an image of an accessor input in real time from the image collecting unit 2 17 and an extracted image previously stored in the database 16 according to an output signal. If 1 (HIGH signal) does not match, 0 (LOW signal) is output, the drive drive (not shown) of the opening / closing means 19a is started, and the opening / closing means 19a is supplied by supplying the necessary power to the opening / closing means 19a. Can be controlled.

On the other hand, in the present invention, the access control unit 19 is an opening and closing means (19a) made of a motor including a motor or a reduction gear is limited to the operation only when the comparison object is matched as a result of the comparison of the comparison unit 18; And a message output means 19b for outputting a warning message if the comparison result of the comparison unit 18 matches.

In this case, the opening and closing means (19a) is usually rotated 180 degrees and stopped, but when the operation is limited, it can be rotated and stopped only by 90 degrees. In this way, the operation of the rotating door Da, which is the door D, can be restricted.

In addition, the message output means 19b may be a speaker that provides / outputs a warning message or may be a screen output device such as a monitor including a liquid crystal display.

On the other hand, the management server 20 of the present invention includes the components other than the image collecting unit 1, 2 (11) 17 and the registration unit 15 and the door (D) of the components of the present invention, It may be a server capable of wired / wireless communication including a program for controlling the operation of the components, a server and a computer system of a security system service company operating the server, and an access controller of a computer central control center at an installation site. .

At this time, the program for controlling the operation is a series of operation steps for pre-image collection-image detection-image correction-feature extraction-storage-real-time image collection-comparison-access control and communication between the registration unit 15 and the management server 20 It can be a program that controls.

The wired and wireless communication may be RF transmission and reception, or may be a local area network (LAN).

When the present invention configured as described above is applied as a security system for preventing suspect escape from the police station in one embodiment as shown in Figure 3 attached.

First of all, the prerequisites for the police station are for the police officers to be authenticated by a normal security system, and the system is installed in the criminal and / or other departments, corridors such as corridors, entrances, and front door guard rooms where the suspect is questioned. It is desirable to be.

Under these conditions, if a police officer uses a personal computer, which is the registration unit 15, and judges that there is a suspicion in the process of preparing a report based on the suspect's interrogation content, the web-camera, which is the image collection unit 1 (11), is operated. Take a picture of the suspect (Suspect Image Shooting Step (S11)).

Then, the photographed image is first loaded into the registration unit 15 so that the police officer can visually check.

As described above, when the image of the suspect photographed by the image collecting unit 1 (11) is transmitted to the management server 20 through the communication unit through the registration unit 15, the management server 20 is an image of the face image detection unit 12 The conversion means 12a is activated to convert the analog image collected by the image acquisition unit 1 11 into a digital method, and when converted into a digital image file. Next, the face detection means 12b is activated to perform digital conversion. The face part is detected based on the eye from the received video signal (face detection step S12).

At this time, face detection detects a face region using a cascaded face classifier using AdaBoost (Adaptive Boost Learning Filter) algorithm. In this case, the cascaded face classifier algorithm has a very severe change in class. In this case, complex decision boundaries are needed for class classification.

The AdaBoost algorithm is a classifier algorithm suitable for such a situation. The AdaBoost algorithm generates a strong classifier with a combination of several weak classifiers and is suitable for classifier learning for face detection.

Since the Cascaded Face Classifier algorithm using AdaBoost does not use motion information and color information, it is possible to perform face detection at high speed without any restrictions on a user even in a general black and white camera.

At this time, what is important is the detection of the eye. Eye detection uses multi-block matching (MMF) to accurately locate and track the eyes under conditions such as wearing glasses, face rotation, and eye closure.

The MMF is an applied model of a normalized template matching technique, collects various databases (DBs) of target patterns to be detected, creates a representative image among the collected databases, and then uses image-based template matching (Image based Templage Matching).

Eye detection detects eye position by pre-calculating and holding various sizes, lighting, rotations, etc. of eye images as templates using MMF, and performing face-to-face template matching using this template.

In detail, a virtual face image is generated by using a plurality of eye candidates selected using an MMF technique, and a candidate including an actual face is selected from among them. The classifier used in the candidate selection process uses the K-Mean Clustering method for a large-scale face DB, so that low-resolution high-speed eye detection is possible.

When the face image is detected in this way, the management server 20 activates the histogram smoothing means 13a of the image correction unit 13 to convert the histogram of the detected face image into an equal distribution, and the gamma correction means 13b. ) To mitigate uneven illumination conditions of the histogram smoothed image in the histogram smoothing means 13a, and then start the vener filter 13c again to remove the noise of the gamma corrected image in the gamma correction means 13b. The face image is corrected, i.e., preprocessed. (Face image correction step S13).

In this case, the image of the captured image is greatly influenced by the noise in the dark area, and since this acts as an obstacle to face recognition, it is essential to remove the noise, and it is very important to select a suitable filter for this. The filter 13c is a kind of linear filter that reflects local characteristics of a given image, and statistically and experimentally recognizes noise on the image and applies an appropriate filter value accordingly.

In addition, the binner filter 13c adaptively designs a filter for removing noise from a gamma-corrected image by using a conventional filter design technique.

The binner filter 13c removes noise of an image through an adaptive filter, and this adaptive filter is more selective than a linear filter that simply preserves a high frequency region or a boundary.

Moreover, no special system variables are required, so it can be used easily. However, it requires a little more computation time than other linear filters.

In this way, when the image file is corrected, the management server 20 starts the log filter 14a of the feature extractor 14 to extract the facial features of the image from which the noise is removed from the veneer filter 13c, Improves the contrast of the facial features extracted through the contrast correction means 14b. (Face feature extraction step (S14))

In this case, the log filter 14a calculates a LoG mask and an image passing through the binner filter 13c and extracts facial features from the face image from which noise is removed through the binner filter 13c. Therefore, it is important to maintain useful facial features and to remove unnecessary pixel values caused by illumination. The Laplacian operator, which is applied to an embodiment of the present invention, is an example of an isotropic measure with respect to the second derivative of a given image. Since it plays a role in making the image appear raised, boundary detection or feature extraction is advantageous.

However, since the Laplacian operator is sensitive to noise, it is necessary to perform filtering to alleviate the noise first. In the present invention, a Gaussian filter is used, that is, the present invention is based on the log filter 14a including a Gaussian filter and a Laplacian filter. Extract facial features. The log filter mask is set to 9 * 9.

In addition, the contrast correction means 14b extends and maps the histogram of the original image to the whole range. When the histogram of the original image in which the illumination is biased in one direction is not distributed over the entire range and is biased to one side, the contrast correction means 14b maps the characteristic part by extending the histogram of the original image to the whole range.

For example, the mapping is to adjust a predetermined image size (size based on the distance between eyes and eyes) at a central position within a certain image area, and partition the image area into a predetermined pixel.

When the feature portion is extracted from the image file of the captured image in this way, the management server 20 assigns an address to the database 16 and stores it as an extracted image file. (Extracted image file input storage step (S15))

In this case, the image collecting unit 2 (17) is installed around the door (D), and collects and collects the image of the outgoing, in particular, the person going out in real time.

The image collected by the image collecting unit 2 (17), that is, the captured image is also received by the management server 20, the above-described face image detection unit 12, image correction unit 13 and feature extraction unit 14 Detects the face image, corrects the image and extracts the image into an image file.

Thereafter, the management server 20 starts the comparison unit 18 and compares them while sequentially loading the extracted image files stored in the database 16 based on the image files photographed and extracted by the image collecting unit 2 17. To determine whether the image is matched. (Shooting image comparison step (S17))

The extracted image file and the image file (subject image) input in real time are adjusted to the same image size at the same position within the same image area, and the image area is divided into preset pixels to compare and analyze each pixel. The comparative analysis algorithm is a feature of the face extracted from the feature extraction unit 14, for example 200 to 3000 effective feature number for recognition, that is, changes in the contour of the face, eyes and eyebrows, nose, mouth around Analyze by comparing the degree of matching of the shaded and the comparison template such as distance and length.

In this way, if the analysis result image file (entry image) input in real time and the extracted image file previously stored in the database 16, the management server 20 is deemed to attempt unauthorized escape, that is, the escape attempt door (D Control the rotation angle of the rotary door Da of () (access control step (S18)).

That is, as shown in Figs. 4 to 6, when the image file (incoming image) input in real time from the image collecting unit 2 (17) and the extracted image file previously stored in the database 16, the management server 20 restricts the operation of the opening and closing means 19a including the motor coupled to the rotating shaft of the rotating door D to rotate and stop only at an angle of 90 degrees.

In this way, the rotary door D stops at an angle of 90 degrees when the rotary door D is pushed out, and the escape suspect is trapped in the blocking space Db, and at the same time, the message output means 19b is unauthorized. A warning message about escape is issued to the police officials. At this time, the management server 20 may be blocked without operating the rotary door (D) itself when it is determined that the unauthorized departure.

Of course, if the image file (incoming image) inputted in real time from the image collecting unit 2 (17) and the extracted image file previously stored in the database 16 does not match, accompanying drawings Figure 4c) the management server 20 is a rotating door The operation of the opening / closing means 19a including the motor coupled to the rotating shaft of (D) is restricted to rotate and stop only at an angle of 180 degrees to allow outing.

In addition, the present invention may be applied to other embodiments as described below with respect to the door D. For example, the door D may be a door in which an E.M lock (electric lock) is installed. Specifically, as shown in FIG. 7, an electromagnet lock D2 is installed on the door frame D1, and the armature strike D3 corresponding to the electromagnet lock D2 is opened. Attached to (D), the door D is opened and closed by energization or interruption of electricity. In addition, the handle is provided in the outer surface, the outer surface of the door (D) is opened, the door (D) is opened by the operation of the person.

In this case, first, if the image file (incoming image) inputted in real time from the image collecting unit 2 (17) and the extracted image file previously stored in the database 16 do not match, the comparison unit of the management server 20 ( 18) the user can manually open the door D by cutting off the electric power applied to the electric magnet lock D2 to remove the magnetic force of the electric magnet lock D2.

However, if the image file inputted in real time from the image acquisition unit 2 (17) and the extracted image file previously stored in the database 16, the comparison unit 18 of the management server 20 is an electromagnet Electricity is supplied to the lock D2 to generate magnetic force, thereby preventing the door D from being opened.

In addition, the door (D) is installed in the automatic door or reinforcement door may be applied to the dead bolt lock to secure the door by falling down the lock embedded in the top of the door frame, Electric Strike Lock may be applied.

In addition, the door D may be a door including a door lock (number key), a card reader (RFID card or smart card), a human body recognizer, and the like.

At this time, the door lock is a key button security device for authenticating access by setting and storing a password in advance, the card reader and the human body recognizer may be a card reader means or a human body recognition means, respectively installed on each side of the front and rear doors, The body recognition means and the card reader means may be mixed.

In this case, the human body recognition means may be a hand vessel recognizer, a fingerprint reader, a voice recognizer, may be a iris reader, the card reader means may be an RFID card reader. In this case, it may be possible to open and close the door separately by comparing the personal body recognition data extracted in advance or the personal identification data recorded in the card in advance to determine whether the access is possible.

As such, the door D of the present invention can be selectively applied in various ways depending on the type of door and the situation.

As described above, the present invention may be used as a security system in a police station as an embodiment, but may also be applied to the same security system in courts, prosecutors, prisons, etc. It can be applied as a company security and time and attendance management system (commuting management, public office overtime check management) that can check the attendance status, and also can be applied to school life management system of students in schools.

While the present invention has been particularly shown and described with reference to exemplary embodiments thereof, it is to be understood that the invention is not limited to the disclosed exemplary embodiments.

Rather, it will be apparent to those skilled in the art that many changes and modifications to the present invention are possible without departing from the spirit and scope of the appended claims.

Accordingly, all such suitable changes and modifications and equivalents should be considered to be within the scope of the present invention.

1 is a system block diagram showing a system configuration according to an embodiment of the present invention;

2 is a system block diagram showing in more detail a system configuration according to an embodiment of the present invention;

3 is a sequence block diagram showing a procedure of a system according to an embodiment of the present invention;

4 to 6 is an operating state diagram showing a blocking or allowing state when the escape situation occurs (going out) of the present invention,

7 is a cross-sectional view showing another embodiment related to the door of the present invention.

* Explanation of symbols on the main parts of the drawings *

10: security system 11: image acquisition unit 1

12: face image detection unit 12a: image conversion means

12b: face image detection means 13: image correction

13a: histogram smoothing means 13b: gamma correction means

13c: binner filter 14: feature extraction unit

14a: log filter 14b: contrast correction means

15: Register 16: Database

17: image acquisition unit 2 18: comparison unit

19: access control unit 19a: opening and closing means

19b: message output means 20: management server

D: Door Da: Revolving door Db: Blocking space

Claims (3)

An image collector 1 (11) for collecting a face image; A face image detector 12 for converting the collected image into a digital image signal and detecting components of a face through a preset algorithm; An image corrector 13 for removing and correcting noise of the detected face image; A feature extractor 14 for extracting feature portions from the corrected face image; A registration unit 15 for inputting identification and blood content, and inputting registration of extracted images; A database 16 for storing the extracted image input from the registration unit 15 as a digital file; Door D; An image collecting unit 2 (17) installed at the door D to collect an image of a person in real time; A comparison unit 18 for comparing the collected image of the visitor with the extracted image previously stored in the database 16 through a preset program; If there is a matching image through the comparison, the access control unit (19) for closing the door (D); security system for escaping prevention comprising a. The apparatus of claim 1, wherein the image collecting unit 1 (11) is a web camera connected to the registration unit (15), and the face image detecting unit (12) converts the collected analog image into a digital method. ); And a face detection means (12b) for detecting a face portion on the basis of eyes in the digitally converted video signal, wherein the image compensator (13) comprises histogram smoothing means for converting the histogram of the input image into an equal distribution ( 13a); Gamma correction means (13b) for mitigating uneven illumination conditions of the histogram smoothed image in the histogram smoothing means (13a); And a binner filter 13c for removing noise of the gamma-corrected image from the gamma correction means 13b, wherein the feature extractor 14 is configured to remove the noise of the image from which the noise is removed from the binner filter 13c. A log filter 14a for extracting facial features; Contrast correction means 14b for improving the contrast of the facial features extracted from the log filter 14a, wherein the register 15 is connected to the LAN through the database 16 and the management server 20 A personal computer comprising: document creation and editing means (15a) for providing a document creation sheet; And a storage form means (15b) for storing the image input from the image collecting unit (1) according to a standardized form. The door (D) comprises a rotating door (Da) having an "X" shape; Blocking space (Db) for providing a space as much as the radius of rotation on both sides of the rotary door (Da); and comprises, the image acquisition unit 2 (17) is a CCTV camera, the comparison unit 18 It is a comparison analysis algorithm that sequentially loads the extracted images stored in the database 16 and compares them with an image of an accessor input in real time from the image collecting unit 2 (17), and the access control unit 19 compares Opening and closing means (19a) consisting of a motor including a motor or a reduction gear is limited to the operation only when the comparison results match the comparison result of the unit 18; And a message output means (19b) outputting a warning message when the comparison result is matched with the comparison result of the comparison unit (18). 2. The door D according to claim 1, wherein an electric magnet lock D2 is provided on the door frame D1, which is a fixed frame of the door D, and the armature stripes D3 corresponding to the electric magnet lock D2 are mounted. Security system for preventing escape, characterized in that the door (D) by opening and closing the door by the energization or interruption of electricity.

Images