JP2001117877A - Image collation log-in system - Google Patents

Abstract

PROBLEM TO BE SOLVED: To facilitate a recovery processing when a collation object comes out of continuous monitoring after log-in to an object device. SOLUTION: A log-in collation processing is executed, by which image information obtained by picking-up the image of collation object by a CMOS camera is collated with a previously registered dictionary to decide whether log-in to PC is permitted or not (S11). After the log-in to PC is permitted, a continuous collation processing without collation with the dictionary is executed in order to continuously monitor that the collation object still exists within the image pickup range of the CMOS camera (S14). When the collation object comes out of the image pickup range of the camera, the continuous collation means executes a register collation processing (S17) for deciding whether the collation object is collated with the dictionary to keep a log-in state or a log-in collation processing (S11) is executed again. The register collation processing (S17) is more simplified than the log-in collation processing (S11).

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention recognizes / collates a face or the like of a person registered in advance with image information in order to prevent unauthorized login to a target device such as a personal computer, and rejects login to the target device. It relates to an image collation login system.

[0002]

2. Description of the Related Art In recent years, outflow of secret information has become a social problem with the spread of personal computers and networks. One way to restrict unauthorized login to your computer is to
Most typically, a method using a several-digit password is known. However, there is a problem that the password is easily stolen, unauthorized access is relatively easily performed, and the confidentiality is insufficient. Such a problem similarly occurs in other target devices such as industrial equipment and safes in which login of unauthorized persons is restricted.

Accordingly, an image collation login system has been developed in which the physical characteristics of a user are determined based on image information to determine whether the user is an authorized user or an unauthorized user, and login is restricted. As for the recognition of image information,
Known are those that determine fingerprints and those that determine facial features. In the fingerprint discrimination method, there is a problem that the contact surface is soiled and maintenance is required because a contact-type fingerprint sensor is used. In addition, maintenance of the face characteristics is easy because it is a non-contact type,
In order to enable high-accuracy and high-speed recognition processing, it is important to position the camera at a predetermined position to a certain extent and to correctly project one's face in a predetermined size. It is important how to correctly and efficiently recognize a face image of a person whose position, angle, etc. change variously.

[0004]

In the above-described image collation login system, the target device is locked even when the user is replaced after the login collation (a state in which substantial operation is impossible, for example, mouse and keyboard operations are invalid). It is desired to perform continuous monitoring even after log-in so that the state becomes as follows. In this case, when the user to be verified disappears from the imaging range of the camera, a verification process for logging in to the personal computer again is performed. However, when assuming daily work, the user often leaves the computer in front of the personal computer, and performing the log-in process each time has a problem that the work efficiency is extremely low.

SUMMARY OF THE INVENTION The present invention has been made in view of the above-described problems, and it is easy to perform a return process when a continuous monitoring after logging in to a target device is not performed, and an image collation that does not cause a reduction in work efficiency. The purpose is to provide a login system.

[0006]

An image collation log-in system according to the present invention collates image information obtained by imaging an object to be collated by an imaging means with a dictionary registered in advance and transmits the collated information to the target device. Login collation means for determining whether to permit login, and collation with the dictionary for continuously monitoring that the collation target continues to be within the imaging range of the imaging means after login authorization by the login collation means. And the collation target is out of the imaging range of the imaging unit, and the collation target is compared with the dictionary to maintain the login state, or the login collation is performed. A registration collation unit that determines whether or not to perform the login collation by the unit again, the registration collation unit having a process simplified than the login collation unit.

According to this invention, when the collation target goes out of the imaging range of the imaging means at the time of the continuous collation by the continuous collation means, the registered collation means is not temporarily switched to the logoff or locked state, but is temporarily set to the collation target. It is determined whether the log-in state is maintained by performing the collation with the dictionary or the log-in collation by the login collation unit is executed again. Since the registration and collation means is simpler in processing than the login collation means, it is easy to return to the continuous collation if the user is an authorized user, and if an unauthorized user, the processing shifts to the login collation again. Unauthorized use is reliably prevented.

[0008] In order to make the registration and collation means simpler than the login collation means, the login collation means and the registration collation means are configured, for example, as follows. In order to confirm whether or not the imaging unit is imaging the verification target at an appropriate position, the login verification unit executes a process of transferring image information captured by the imaging unit to the target device and displaying the image information. In some cases, the registration matching unit does not execute the transfer processing of the image information to the target device. This simplifies the processing required for image transfer and the time required, so that it is possible to quickly return to the continuous collation. In this case, it is not possible to confirm the position of the object to be checked with respect to the imaging means at the time of registration and matching, and therefore, for example, a reflecting mirror for positioning the object to be checked at a position suitable for the matching processing with respect to the imaging means near the imaging means It is convenient to have

In the case where the login collating means collates image information obtained by imaging the collation target with the imaging means with all the contents of the dictionary registered in advance,
The registration and collation unit performs collation between the image information obtained by imaging the collation target by the imaging unit and only the dictionary related to the currently logged-in collation target among the pre-registered dictionaries. As a result, the number of targets of the matching process can be reduced, so that it is possible to quickly return to the continuous matching.

[0010] As the continuous collation means, for example, a specific area in the image information to be collated with little change in color information is set as a tracking frame, and a pixel having the same or similar color information as the color information of the tracking frame is set. Can be used to determine whether or not the collation target exists in the imaging range based on whether or not the set exists in the imaging range with a predetermined area or more. Since only the color information is checked, even when the collation target slightly moves and the luminance information changes, the same target can be accurately tracked, and since only the color signal is targeted, the processing amount is small,
High-speed processing suitable for continuous matching processing can be realized.

[0011]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Preferred embodiments of the present invention will be described below in detail with reference to the drawings. FIG.
1 is an external view showing a configuration of an image collation login system according to one embodiment of the present invention. This system is configured by connecting an image collating device 1 and a personal computer (PC) 2 via a communication cable 3 such as RS232C. The image matching device 1 includes a matching device main body 11 in which an electronic circuit is housed, and a stand-type camera housing portion 12 whose lower end is rotatably supported at the center of the front surface of the matching device main body 11. And the whole is designed to be small and light. The camera housing 12 has a CMO
An S camera 13 is built in, and a convex mirror 14 which is a reflecting mirror for positioning is provided immediately below the CMOS camera 13.
An LED 15 for operation confirmation is provided below the LED. The personal computer 2 includes a personal computer main body 21 and a display device 2
2 and input / output devices such as a keyboard 23 and a mouse 24.

FIG. 2 is a block diagram of the image collating system. The image matching device 1 is configured as follows. Image data captured by the CMOS camera 13 is temporarily stored in the image memory 110. The CPU 120 controls each unit of the image matching apparatus 1 according to a program stored in the program memory 130 to execute various processes. The functions of the CPU 120 include a dictionary creation / registration unit 121, a login matching unit 122, and a continuous matching unit. 123 and the registration / collation processing unit 124 are configured. The data holding memory 140, the working memory 150, and the communication interface (I / F) 160 are connected to the CPU 120. The data holding memory 140 is a battery-backup type RAM or E
A dictionary (mosaic image) created by the dictionary creation / registration unit 121 and a non-volatile memory such as an EPROM
An F (tracking frame) value, an average luminance value YA, a registrant name, a device ID number, and the like are registered. Working memory 15
0 provides a storage area for temporarily storing the result of the collation processing at the time of login collation, continuous collation, and registration collation by the CPU 120. The communication I / F 160 is a communication cable 3
For performing serial data communication with the PC 2 via, for example, RS232C.

On the other hand, the PC 2 has a communication I / F 220, an image memory 230, a data holding memory 240, a program memory 250, a work memory 2
60, an input / output I / F 270, and various input / output devices 280. The CPU 210 includes a program memory 250
Dictionary creation / registration based on the program stored in
Guide processing for login collation and registration collation, determination processing of collation rates, output of execution commands for various collation processing, and the like are performed. That is, the CPU 210 executes the dictionary creation / registration processing unit 21
1. Login collation processing unit 212 and registration collation processing unit 21
Constituting No. 3. The PC 2 receives a still image from the image matching device 1 via the communication I / F 220 and stores it in the image memory 230 only at the time of dictionary creation / registration and at the time of login matching. The data holding memory 240 is composed of a nonvolatile memory such as a battery-backup type RAM or an EEPROM, in which the ID number of the image collating apparatus 1, the registrant name, the collation rate for permitting login, and the like are registered. Working memory 260
Provides a work area for processing in the CPU 210, and various input / output devices 280 such as the display device 22, the keyboard 23, and the mouse 24 communicate with the CP via the input / output I / F 270.
The necessary information is input and output to and from U210.

Next, the operation of the thus configured image collation log-in system will be described. FIG. 3 is a main flowchart showing the operation of the image matching login system. First, after the image collating apparatus 1 and the PC 2 are connected via the communication cable 3 and the PC 2 is started, an initialization process is executed (S1). FIG. 4 is a flowchart of the initialization process. First, a menu screen for the initialization process is displayed (S11), and selection of any of reception of a login request, environment setting, dictionary editing, and suspension of the process is awaited (S11). Here, if the environment setting (S12) is selected,
A setting process for setting environment parameters such as a password setting, a port setting, a COM port selection, a communication speed setting, a collation rate setting, a login collation rate setting, a continuous collation rate setting, a directory setting, and various fragments is started, and various environments are set. Parameters can be set. After the setting, the display returns to the menu screen display (S11) and waits for an instruction. When dictionary editing (S14) is selected, a dictionary editing process is started. If the user selects the stop of the process (S15), after the confirmation process (S15a) as to whether the process can be stopped, a stop display is executed (S15b), and then the process is stopped.

On the other hand, when the login request is accepted (S13), the port opening process is executed at the communication speed set in the PC 2, and after a waiting time of about 10 seconds, it is checked whether or not the port has been opened (step S13). S13
a). If the port has not been opened, a message “Port cannot be opened” is displayed (S13
b), the display returns to the menu screen display (S11) and waits for an instruction. If the port is successfully opened, an ID collation process is executed (S13c). Here, the ID value is confirmed between the image collation device 1 and the PC 2 because the dictionary exists on the image collation device 1 side, and when the image collation device 1 is replaced improperly, this is detected. To do that. If the IDs do not match in the ID confirmation processing (S13d), the fact is displayed on the display device 22 (S13e), and the display returns to the menu screen (S11). I
If the D values match, check whether the dictionary has been registered (S
13f) If the dictionary has been registered, the initialization process ends and the process proceeds to the next step. If the dictionary has not been registered, the process proceeds to the dictionary editing process (S14) and the dictionary editing process is executed. Proceed to the next step.

FIG. 5 is a flowchart showing details of the dictionary editing process (S14). In the dictionary editing process, first, a dictionary is written from the image collating device 1 to the PC 2 by TE.
It is read as an MP file (S14a). Subsequently, the process waits for selection of any one of dictionary registration (S14b), attribute entry (S14c), dictionary deletion (S14d), and dictionary writing (S14j). Here, dictionary registration (S1
When 4b) is selected, a dictionary frame is selected (S14e),
The face image of the registrant is captured (S14f). The registrant confirms the captured original image (S14g), and if OK, registers it in the dictionary as a TEMP dictionary (S14h), but if OK, repeats imaging (S14g). Attribute entry (S14
When c) is selected, a process for registering attributes such as a registrant name and a posture in the dictionary is executed. When the dictionary deletion (S14d) is selected, a dictionary frame is selected (S14i), and the data of the selected dictionary frame is deleted. Dictionary writing (S1
When 4j) is selected, all dictionary data generated by the PC 2 is uploaded to the image matching device 1.

In the idle state, when a dictionary registration event is activated (S6), dictionary creation / registration processing is started (S7). In the dictionary creation / registration processing, the image of the face of the registrant to be collated is converted into a multi-mosaic and registered as a dictionary. When the registration is completed, the process returns to the idle state (S8). When the login verification process is started (S
9), the login process shown in FIG. 4 is started. That is, first, the login collation processing is executed (S11). In the login collation processing, the image of the face of the user to be collated is converted into a multi-mosaic pattern and collated with the dictionary. In order to confirm how the still image to be compared has been captured from the image matching device 1, the captured image is transferred from the image matching device 1 to P
The data is transferred to C2 and can be confirmed on the display device 22 of the PC2.

The above is the initialization process (S1) in FIG. A login request is accepted in the initialization process, and if port opening and ID collation succeed, the process proceeds to the login collation process (S2). In the login collation process, the image of the face of the user to be collated is multi-mosaiced and collated with the dictionary. In order to check how the still image to be compared was captured from the image matching device 1, the captured image is transferred from the image matching device 1 to the PC2, and can be confirmed on the display device 22 of the PC2. ing.

If the collation result is OK (S3), the process proceeds to the continuous collation processing (S4). In this continuous matching process, it is continuously monitored whether or not there is a region having a predetermined area or more displaying color information of a part of the user's face in the imaging region. The reason why the continuous collation processing is performed is to prevent the PC 2 from being operated by another person when the registrant stands after the login. On the PC 2 side, during the continuous matching process, only a predetermined decision branch process is performed on the matching result sent periodically (for example, twice a second), and no other special process is performed.

When the collation result is NG in the continuous collation (S5), a registration collation process is executed (S6). In the registration collation, similarly to the login collation, the image of the face of the user to be collated is converted into a multi-mosaic pattern and collated with the dictionary.
However, in the registration verification, in order to reduce the burden on the PC2 side,
The captured image is not transferred to PC2. Also,
The dictionary to be collated is also limited to that of the registrant who is currently logged in. If the result of the registration verification is NG, (S
7) Execute the PC lock process (S8) to lock the PC 2 and then execute the registration verification process (S6) again. If the collation result is OK in the registration collation process (S7), if it is currently locked (S9), the lock is released (S10), and the process returns to the continuous collation (S4). Next, more detailed contents of each processing will be described.

(1) Dictionary Creation / Registration Processing As shown in FIG.
An image is taken at 3. Two types of frames are set in the face image. The first frame is set to a size and a position where the face part is fully filled in the mosaic frame MF. By confirming the state of one's own face reflected on the convex mirror 14 of the image matching device 1, the mirror image is converted into a mosaic frame MF.
Can be adjusted to a predetermined position so as to be captured. This convex mirror 14 is such that when the registrant faces this convex mirror 14 facing the front, the image range reflected there is:
In order to substantially match the imaging range of the CMOS camera 13,
In the mechanism design stage, the degree of the engineering convex surface is adjusted in consideration of the positional relationship with the lens surface of the CMOS camera 13 and the imaging range characteristics. When the imaging state is confirmed on the display device 22 of the PC 2, the still image data is transferred from the image collating device 1 to the PC 2, but the still image captured by releasing the shutter is checked later. Therefore, also in this case, the alignment by the convex mirror 14 becomes important. The second frame is a tracking frame TF, which is set in a range of a small variation of the UV signal in the mosaic frame MF, for example, a cheek portion below the eyes.

The image data in the mosaic frame MF is
The luminance signal Y is converted into a mosaic by the multi-mosaic processing described later, and the amount of information is reduced. Here, two methods are exemplified as the multi-mosaic processing. In the first method, in consideration of the fact that the characteristic portion is concentrated in the central portion of the face, as shown in FIG. 7A, the mosaic frame MF is mosaiced so as to become gradually smaller from the periphery to the central portion. . For example, the outermost part is a large mosaic of [8 × 8] blocks, the inner part is a middle mosaic of [16 × 16] blocks, and the inner side is a small mosaic of [32 × 32] blocks. To go. That is, when the entire area of 352 pixels (pixels) and 288 pixels in width is used as the mosaic frame MF, the mosaic of the [8 × 8] block is configured such that each mosaic unit is 44 pixels in length × 36 pixels in width. Similarly, the middle mosaic is 22 pixels by 18 pixels horizontally, and the small mosaic is 1 pixel vertically.
It is composed of 1 pixel x 9 horizontal pixels. FIG. 7B shows an example of the mosaiced image in this manner. For example, three mosaic images with different angles are created for one registrant, and these are registered as a dictionary of the registrant. Dictionaries allow registration for multiple registrants,
The information is registered in the data holding memory 140 together with the information of the registrant's name.

The second method is a method of finely mosaicing a portion having a particularly large change in brightness, based on the fact that characteristic portions are concentrated in a portion where the change in brightness is sharp.
As shown in FIG. 8A, the mosaic frame MF is first made into a large mosaic with the coarsest [8 × 8] block. Next, the luminance of one block of interest and the eight blocks adjacent thereto are compared, and if there is a change equal to or more than a predetermined value, as shown in FIG. The block is placed in the same [16 ×
[16]. This operation is sequentially repeated to generate a multi-mosaic composed of a plurality of layers composed of mosaic blocks having different roughnesses, as shown in FIG. This multi-mosaic image may be registered in the data holding memory 140 as a dictionary as it is, or information on its position and density may be registered in the memory 140 in order from the finest mosaic block. In the latter case, in particular, it is possible to perform the matching process first from the important data registered first, and to end the matching process when the matching value of the important data exceeds a predetermined value. Processing can be performed.

FIG. 9 is a flow chart showing the dictionary creation / registration procedure of the PC 2. First, the display device 22
Then, a guide display for capturing a face image is performed (S21).
At this time, the guide display includes the convex mirror 1 of the image matching device 1.
4 displays a message urging the user to photograph the user's face from the front. Next, a shutter-on wait state is entered by operating the keyboard 23 or the mouse 24 (S22). When the shutter is turned on, a dictionary creation / registration instruction command is transferred to the image matching device 1, and the camera shutter sound is notified by a dummy sound (S23). The captured still image data of the user's face is received from the image matching device 1 (S24), and displayed on the display device 22 as shown in FIG. 5 (S25). The user sees this screen and confirms that the face is completely filled in the mosaic frame MF, and issues a luminance correction instruction if necessary, or instructs re-acquisition of an image (S26). The luminance correction instruction is performed by, for example, evaluating the luminance of the image in the mosaic frame MF in five steps as follows.

Dark → + 20 Slightly dark → + 10 Good → 0 Slightly bright → −10 Very bright → −20

Here, when the user selects the redo button, the correction information corresponding to the above-mentioned luminance is transmitted to the image collating apparatus 1 (S28), and the process returns to the initial processing again, and the registration button is pressed. If is selected, the registrant name, collation rate, etc. are registered in the data holding memory 240 of the PC 2 (S29). Necessary registration information such as a registrant name is also transmitted to the image matching device 1 (S30). The collation rate is a threshold value for determining the degree of matching between the registered dictionary and the image captured at the time of login before permitting login, and is not transferred to the image collation apparatus 1 side.
Further, once these pieces of registration information are set, the change is set so that the change can only be performed by an authorized user.

FIG. 10 is a flowchart showing a dictionary creation / registration procedure on the image collating apparatus 1 side corresponding to the above processing. First, in response to a shutter-on instruction from the PC 2, the shutter of the CMOS camera 13 is released, and an image is taken into the image memory 110 (S31). Subsequently, a light source mask process is executed as a countermeasure when a light source enters the captured image (S32). Specifically, when the luminance value is 2
Pixels of 00 or more are replaced with 200 and an upper limit is applied. Conversely, a pixel having a luminance value of 30 or less may be set to 30 to apply a lower limit. Further, with respect to the fluctuation of the luminance, the average luminance value YAV in the mosaic frame MF is used.
Is calculated (S33), and the set luminance YA (experience value) and the luminance average value YA stored in advance in the data holding memory 140 are calculated.
The difference from V is subtracted from the luminance value of each pixel, and the luminance is corrected so that the average luminance of the image data matches the set value YA (S34). Next, the image quality is improved by executing the density conversion processing (S35). Here, the density conversion process is a process for more effectively utilizing this range when the brightness value can take 0 to 255, and is black (dark) with a brightness value of 0 and white (bright) with a brightness value of 255. ) Are linearly proportional to the center of the luminance value, that is, the value 127, and to both sides, that is, to the 0 side and the 255 side, on the assumption that the values in each neighborhood are not normally possible in actual use. Alternatively, it is a process of expanding in logarithmic proportion.

Next, the image data subjected to the luminance correction and the density correction are transferred to the PC 2 (S37). When an instruction to start over is issued from the PC 2 (S38), a correction value such as brightness information is acquired from the PC 2 (S39), and the luminance average value YA is obtained.
Is corrected by the correction value and the process returns to the brightness correction process (S34), or
Alternatively, the process returns to the image capturing process (S31). When there is a registration instruction from the PC 2, the luminance signal of the image data is multi-mosaiced by the method described above (S41), and this is registered as a dictionary in the data holding memory 140 together with the average luminance YA and the name of the registrant. (S42).

(2) Login Verification Processing FIG. 11 is a flowchart showing a log-in verification procedure of the PC 2. First, a guide display for capturing a face image is performed on the display device 22 (S51). At this time, on the guide display, a message is displayed to prompt the convex mirror 14 of the image matching device 1 to photograph the user's face from the front. Next, a shutter-on wait state is entered by operating the keyboard 23 or the mouse 24 (S52). When the shutter is turned on, the command of the login collation instruction is transferred to the image collation device 1 and the camera shutter sound is notified by a dummy sound (S
53). The captured still image data of the user's face is received from the image matching device 1 (S54), and is displayed on the display device 22 as shown in FIG. 5 (S55).
The user looks at this screen and confirms that the face is completely filled in the mosaic frame MF, and if necessary, issues a luminance correction instruction or instructs re-acquisition of an image (S56).
The luminance correction instruction is performed by, for example, evaluating the luminance of the image in the mosaic frame MF in five steps as follows.

Dark → + 10 Slightly dark → + 5 Good → 0 Slightly bright → −5 Very bright → −10

Here, when the user selects the redo button, the correction information corresponding to the above-mentioned luminance is transmitted to the image collating apparatus 1 (S58), and the process returns to the initial processing again, and the registration button is pressed. Is selected, the maximum collation rate and the registrant name described later are received from the image collation apparatus 1 (S
59), and compares the previously registered collation rate with the received maximum collation rate to determine the collation result (S60). If the result of the collation is NG, the process returns to the idle state, but if it is determined to be OK, a continuous collation start command is output (S62).

FIGS. 12 and 13 are flowcharts showing a login collation procedure on the image collation apparatus 1 side. First, in response to a shutter-on instruction from the PC 2 side, the CMOS
The shutter of the camera 13 is released, and the image is stored in the image memory 110.
(S71). Subsequently, the above-described light source masking process is executed as a countermeasure when a light source enters the captured image (S72). Also, with respect to fluctuations in luminance,
The above-described average luminance value YAV in the mosaic frame MF is calculated (S73), and the difference between the set luminance YA (experience value) previously stored in the data holding memory 140 and the average luminance value YAV is calculated as the luminance value of each pixel. , And correct the luminance so that the average luminance of the image data matches the set value YA (S74). Next, the image quality is improved by performing the above-described density conversion processing (S75).

Next, a TF value is calculated (S76). That is, as shown in FIG. 5, the average value of the color signals of the tracking frame TF set in advance in the mosaic frame MF is calculated. This is stored in the working memory 150 as a TF value. Then, the image data subjected to the luminance correction and the density correction is transferred to the PC 2 (S77). When an instruction to start over is issued from the PC 2 (S78), a correction value such as brightness information is acquired from the PC 2 (S79), and the average luminance value YA is obtained.
Is corrected by the correction value and the process returns to the brightness correction process (S74), or
Alternatively, the process returns to the image capturing process (S71). When the collation instruction is received from the PC 2 (S80), the luminance signal of the image data is multi-mosaiced by the above-described method and stored in the working memory 150 (S81), which is registered in the data holding memory 140. Check with the dictionary (S82).
The collation processing is executed by reading out the multi-mosaic images for three angles of each registrant registered as a dictionary one by one, collating them with the created mosaic, and storing each collation rate in the working memory 150. Is done. When all the collation rates are obtained, the maximum collation rate is found (S83). Then, the maximum matching rate and the registrant name of the dictionary from which it was obtained are transferred to the PC 2 and the registrant name is stored in the working memory 150 (S84). In response,
The collation determination result sent from C2 is received (S8).
5). If the collation result is NG, the process returns to the idling state, but if the collation result is OK and a continuous collation start command is received, the process proceeds to continuous collation (S86).

(3) Continuous collation processing After the PC 2 outputs a continuous collation command, the image collating apparatus 1
The side performs the continuous collation processing, and does not involve PC2 at this time. However, when performing OPEN, COPY, and PRINT of a special file for security specified on the PC 2 side, the PC 2 needs to perform a more strict check.
May output a command to stop continuous collation, and may output a command to shift to registration collation described later. FIG.
5 is a flowchart showing a continuous collation procedure in the image collation device 1. Upon receiving the continuous collation command from the PC 2, the image collation device 1 turns on the shutter and captures a UV signal into the image memory 110 (S91). The entire screen is corrected with the correction value YA (S92), and a 32 × 32 color mosaic is created (S93). Next, the color of each mosaic is compared with the TF value stored at the time of login matching. If the color of each mosaic is within the range of TF value ± α, it is set to 1; A collation process is performed (S94). As a result, as shown in FIG.
Several islands (pixel sets) Ia and Ib appear in which mosaics of colors close to the F value are connected, and the maximum area of the islands Ia and Ib is determined (S95), and the maximum area is determined by a predetermined value (for example, For [32 × 32] blocks, ie, 11
If it is (× 9 × 4 pixels) or more (S96), it is determined that the collation target exists in the field of view of the camera, and the collation OK is transferred to PC2 (S97), and the disappearance counter is reset (S98). If the maximum area is smaller than the predetermined value, it is determined that the collation target has disappeared from the camera field of view, and the disappearance counter is incremented by one (S99). Then, when the disappearance counter exceeds 8 (S100), the PC 2 is notified of the transfer of registration and collation (S102). Here, the reason why the disappearance counter is used is to prevent frequent shifts to the registered collation when the collation target is instantaneously lost from the camera field of view. Further, the continuous collation is executed at regular intervals, for example, at intervals of two times per second. Therefore, except for the case of shifting to registration verification, a predetermined time elapse waiting step S101
The above processing is repeated via. If the maximum area suddenly increases, this may also be abnormal, but for example, it is assumed that the palm of the operator overlaps in the camera field of view and the TF value is similar. Therefore, in this case as well, it is also possible to look at the state for a predetermined time as in the case of the above-described disappearance counter, and then make a determination based on the state.

(4) Registration / collation processing FIG. 16 is a flowchart showing the registration / collation processing in the PC 2. When the process shifts to registration verification, a guide display for registration verification is displayed on the display device 22 of the PC 2 at the same time as a warning sound in the window (S111). At this time, on the guide display, a message is displayed to prompt the convex mirror 14 of the image matching device 1 to photograph the user's face from the front. Next, a shutter-on wait state is entered by operating the keyboard 23 or the mouse 24 (S112). When the shutter ON is instructed (S112), the image collating apparatus 1 is instructed to execute the registration collation (S113). When the maximum collation rate transferred from the image collation device 1 is received (S114), the collation result is determined (S115). If the collation result is NG, the PC 2 is locked and the process returns to the login collation (S11).
6) If the collation result is OK, the guide display is stopped (S117), and a command to return to the continuous collation is issued (S11).
8).

FIG. 17 is a flowchart showing the registration and collation processing on the image collating apparatus 1 side. First, in response to a shutter-on instruction from the PC 2, the shutter of the CMOS camera 13 is released, and an image is taken into the image memory 110 (S1).
21). Subsequently, similarly to the dictionary creation / registration and login collation, the captured image is subjected to light source mask processing (S122), average luminance value YAN calculation (S123), and luminance correction (S12).
4) The density conversion processing (S125) is sequentially performed, and the image data is multi-mosaiced and stored in the working memory 150 (S126). Next, the dictionary of the registrant who is currently logged in is read out three times at a time, and the work memory 1
The collation is performed with the mosaic image in 50, and each obtained collation rate is stored in the working memory 150 (S127). Then, the maximum value of the obtained collation rate is found (S128), and the maximum collation rate is transferred to PC2 (S129).

As described above, in the registration collation processing, the log collation processing is performed in that the image data is not transferred from the image collation apparatus 1 to the PC 2 and the collation is performed only with the dictionary of the registrant who is currently logged in. Has been simplified. In this way, when the registered user who is currently working on the PC 2 comes back up for a short period of time, it is possible to quickly return to the continuous collation as far as possible without hindering the work. .

In the registration / collation processing, the image collation device 1
, The image data is not transferred to the PC 2 side, so that the positioning by the convex mirror 14 of the image matching device 1 becomes more important. In addition to the above-described embodiment, a large convex mirror 16 disposed on the front of the CMOS camera 13 as shown in FIG. You can also.
In this case, a pinhole 17 may be provided at the center of the convex mirror 16, and light may be guided to the CMOS camera 13 through the pinhole 17. Further, as shown in FIG. 19, a convex mirror 18 made of a magic mirror may be arranged in front of the CMOS camera 13. According to these aspects, there is an advantage that positioning can be confirmed in a larger area.

[0039]

As described above, according to the present invention, when the collation object goes out of the imaging range of the imaging means at the time of the continuous collation by the continuous collation means, the log-off is not performed immediately, but the log-in collation means. It also determines whether to maintain the login state by collating the object to be collated with the dictionary by the simplified registration collation means, or to execute the login collation again by the login collation means. The return to the collation is easily performed, and if the user is an unauthorized user, the process shifts to the login collation again, so that there is an effect that the unauthorized use is reliably prevented.

[Brief description of the drawings]

FIG. 1 is a perspective view showing an overall configuration of an image collation login system according to an embodiment of the present invention.

FIG. 2 is a block diagram showing a configuration of the drawing system.

FIG. 3 is a flowchart showing main processing of the system.

FIG. 4 is a flowchart showing an initialization process of the main processing.

FIG. 5 is a flowchart showing a dictionary editing process of the initialization process.

FIG. 6 is a diagram showing a relationship between a captured image and a mosaic frame and a tracking frame in the same system.

FIG. 7A is a diagram showing a multi-mosaic used in a first multi-mosaic process in the same system.

FIG. 7B is a diagram showing an example of an image mosaiced by the multi-mosaic.

FIG. 8 is a diagram for explaining a second multi-mosaic process in the same system.

FIG. 9 is a flowchart of a dictionary creation / registration process on a PC in the same system.

FIG. 10 is a flowchart of a dictionary creation / registration process in the image matching device in the same system.

FIG. 11 is a flowchart of a login collation process on a PC in the same system.

FIG. 12 is a flowchart of a login collation process in the image collation device in the same system.

FIG. 13 is a flowchart showing a process subsequent to FIG. 12;

FIG. 14 is a flowchart of a continuous matching process in the image matching device in the same system.

FIG. 15 is a diagram illustrating a continuous matching process in the system.

FIG. 16 is a flowchart of a registration / collation process on a PC in the same system.

FIG. 17 is a flowchart of a continuous matching process in the image matching device in the same system.

FIG. 18 is a front view of an image matching device according to another embodiment of the present invention.

FIG. 19 is a front view of an image matching device according to still another embodiment of the present invention.

[Explanation of symbols]

DESCRIPTION OF SYMBOLS 1 ... Image collation device, 2 ... PC, 3 ... Communication cable, 11
… Verification device body, 12… camera housing, 13… CMOS
Camera, 14, 16, 18 ... convex mirror.

 ────────────────────────────────────────────────── ─── Continuing on the front page (72) Inventor Masajun Kawahara 4-7-16 Kudanminami, Chiyoda-ku, Tokyo Ichigaya KT Building F-term in Wise Corporation Tokyo Branch F-term (reference) 5B043 AA04 BA04 DA05 DA09 EA01 FA07 GA03 GA11 HA02 HA05 5B085 AE06 AE25 BC01

Claims (5)

[Claims] A login collation unit configured to collate image information obtained by imaging an object to be collated by an imaging unit with a pre-registered dictionary to determine whether to permit login to the target device; A continuous collation unit that does not involve collation with the dictionary, continuously monitors that the collation target is continuously present in the imaging range of the imaging unit after the login is permitted by the login collation unit; After the collation target is out of the imaging range of the imaging unit, determine whether to perform collation between the collation target and the dictionary and maintain the login state, or execute login collation again by the login collation unit. An image collation login system comprising: a registration collation unit whose processing is more simplified than a login collation unit. 2. The login verification unit transfers image information captured by the imaging unit to the target device in order to check whether the imaging unit has captured the verification target at an appropriate position. 2. The image matching login system according to claim 1, wherein the registration / collating unit does not execute a process of transferring the image information to the target device. . 3. The apparatus according to claim 1, further comprising a reflector near the imaging unit for positioning the object to be compared with the imaging unit at a position suitable for a matching process.
Or the image collation login system according to 2. 4. The log-in collating means collates image information obtained by imaging the collation target with the imaging means with all contents of a dictionary registered in advance. Out of image information obtained by imaging the object to be checked by the imaging means and a dictionary registered in advance,
The image collation login system according to any one of claims 1 to 3, wherein collation is performed only with a dictionary related to the collation target that is currently logged in. 5. The continuous collation unit sets a specific area in the collation target image information where color information has little change as a tracking frame, and sets a pixel having color information identical or similar to the color information of the tracking frame. The determination of whether or not the collation target exists in the imaging range is based on whether the set of images exists in the imaging range with a predetermined area or more.
The image collation login system according to any one of the above.