JP5782613B2 - User authentication device - Google Patents

Description

  The present invention relates to a occupant authentication device used for a door that is opened and closed by a handle and incorporates an electric lock for locking and unlocking.

  The entrance doors provided at the entrances of general houses, condominiums, offices, etc. have an entrance frame fixed to the housing, a door fixed inside the entrance frame so that it can be opened and closed, and the door is locked and unlocked. A lock case including a handle is mainly configured.

The front door is generally one in which a key is inserted into a key hole of a lock case provided in the vicinity of the handle and is manually rotated to lock or unlock the door.
On the other hand, in recent years, entrance doors that open and close doors with electric locks have also become widespread. As one of the means for controlling the electric lock, biometric authentication using biometric information such as fingerprints and palm prints has been put into practical use, and has come to be used in applications for opening and closing the door of the entrance door (for example, (See Patent Document 1 (FIG. 1)).

Patent document 1 is demonstrated based on the following figure.
FIG. 10 is a block diagram showing the configuration of the occupant authentication apparatus 100 that detects patterns (fingerprints and palm prints). The lever handle 103 of the door 102 incorporating the electric lock 101 has fingers and palms touched by the hand when opening and closing. A pattern reader (in the pattern detection window 104) that reads the pattern and outputs a detection signal S1 is embedded at a position such as.
The pattern read by the pattern reader is collated with a pattern registered in advance by the pattern collator 105. The control unit 106 outputs the unlock signal S3 and controls the unlocking of the electric lock 101 when the two patterns match and the match signal S2 is input from the pattern collator 105. In this state, the occupant can turn the lever handle 103 to open and close the door 102.

  However, according to the technique disclosed in Patent Document 1, the pattern detection window 104 provided at the position of the finger or palm of the hand touching the lever handle 103 is small, and thus the finger or palm of the hand touching the lever handle 103 is small. The position is limited. Therefore, depending on the place touched, the pattern cannot be read well, and an extra action such as a re-reading operation occurs. In this case, when the person entering the room opens the door 102 and enters the room, the pattern cannot be automatically acquired in a natural act of touching the lever handle 103.

  In addition, since the pattern detection window 104 is allocated only an area that can be touched by the thumb, the pattern that can be read by the pattern reader is only a fingerprint or information on a part of the palm. Therefore, the amount of pattern information obtained for authenticating a room occupant is small, and therefore, the authentication rate is lowered and the room occupant authentication with high accuracy cannot be performed.

JP-A-8-28109

  It is an object of the present invention to provide a room occupant authentication device that performs room occupant authentication with high accuracy in a natural act of a room occupant opening a door and touching a handle when entering a room.

The invention according to claim 1 is a occupant authentication device used for a door that is opened and closed by a handle and includes an electric lock for locking and unlocking, and is embedded in the handle, and triggers to start shooting and end shooting An image photographing unit that senses a trigger and captures a plurality of palm print images in between, and converts the plurality of photographed palm print images into a panoramic image and simultaneously samples them into a plurality of palm print images that are not distorted in shape obtaining a plurality of palmprint images converted performs the plurality of palmprint images between the trigger of the trigger and the imaging end of the previous SL photographing start, the collation with the plurality of palmprint images registered in advance sequentially, the collation And a control unit for controlling the unlocking of the electric lock according to the result.

  According to a second aspect of the present invention, in the occupant authentication device according to the first aspect, the trigger for starting the photographing is the beginning of gripping the handle, and the trigger for ending the photographing is the end of gripping the handle. And

  According to a third aspect of the present invention, in the occupant authentication device according to the first aspect, the shooting start trigger is the beginning of gripping of the handle, and the shooting end trigger is the end of capturing of the plurality of palm print images. It is characterized by that.

According to a fourth aspect of the present invention, in the occupant authentication device according to the first aspect, the shooting start trigger is the beginning of gripping of the handle, and the shooting end trigger is the end of the collation. To do.

  According to a fifth aspect of the present invention, in the occupant authentication device according to the first aspect, the image photographing unit includes an omnidirectional camera, and transmits a palmprint image photographed by the omnidirectional camera to the control unit. Features.

  According to a sixth aspect of the present invention, in the occupant authentication device according to the first aspect, the image photographing unit includes a digital camera from which an infrared light shielding filter is removed and an infrared light passing filter is attached, and is photographed by the digital camera. The palm print image is transmitted to the control unit.

  According to a seventh aspect of the present invention, in the occupant authentication device according to the first aspect, the control unit includes a predetermined range of pixels having a base of the little finger as a reference point for each palm print image photographed by the image photographing unit. Cut out, generate a feature vector in units of the cut out pixels, obtain a degree of coincidence with the feature vector of the previously registered palmprint image, and solve the electric lock according to a comparison result between the degree of coincidence and a threshold value. An unlocking signal to be locked is generated and output.

According to an eighth aspect of the present invention, in the occupant authentication device according to the seventh aspect, the control unit unlocks the electric lock when there are a predetermined number or more of palmprint images having a degree of coincidence equal to or greater than the threshold. A lock signal is generated and output.

In the invention according to claim 1, the image photographing unit embedded in the handle detects the photographing start trigger and the photographing end trigger and photographs the palm print image therebetween, and the control unit captures the plurality of photographed palm print images. was converted to a panoramic image, a plurality of get a palm print image, a plurality of palm print image between the trigger end shooting trigger start shooting a distortion in the shape is converted into a plurality of palmprint images no sampling simultaneously, The electronic lock is controlled to be unlocked by sequentially collating with a plurality of palmprint images registered in advance . In this way, since the palm print image is collated based on the video shot between the detection of the shooting start trigger and the detection of the shooting end trigger, the amount of palm print information used for authentication is large. It is possible to perform a highly accurate occupant authentication in a natural action such as when the occupant opens the door and enters the room and touches the handle.

  The invention according to claim 2 detects the trigger at the beginning and end of gripping, performs collation based on the palm print image taken during that time, and controls the unlocking of the electric lock. It is possible to perform a highly accurate occupant authentication in a natural action such as touching the handle when performing the operation.

In the invention according to claim 3, since the trigger for starting the photographing is set to the beginning of gripping the handle and the trigger for the end of photographing is set to the end of capturing a plurality of palmprint images necessary for collation, the handle is gripped for a relatively long time when the door is opened. Even if there is a tendency in the room, the photographing ends without depending on the time for which the handle is held, and therefore, the chance of photographing a useless palm print image is reduced.

In the invention according to claim 4, since the trigger for starting the shooting is the beginning of gripping the handle and the trigger for ending the shooting is the end of collation, it is a resident who tends to grip the handle for a relatively long time when opening the door. However, since the shooting is completed without depending on the time when the handle is held, the chances of shooting a useless palm print image are further reduced.

In the invention according to claim 5, the image photographing unit transmits the palm print image photographed by the omnidirectional camera to the control unit. Since palm prints are taken with an omnidirectional camera, the entire image of the palm prints can be taken without depending on the position of the palm holding the handle. Accordingly, it becomes possible to authenticate a room occupant with high accuracy.

  In the invention according to claim 6, the palm print image taken by the digital camera to which the infrared pass filter is attached is transmitted to the control unit. Since a palmprint image can be taken with a widely used digital camera, a occupant recognition device can be provided at low cost.

  In the invention according to claim 7, the controller cuts out a predetermined range of pixels with the base of the little finger as a reference point for each captured palmprint image and performs collation. Further, at the time of collation, a feature vector of the cut out pixel is generated, and the electric lock is unlocked according to the degree of coincidence with the feature vector of the palmprint image registered in advance. In this way, the matching between the pixel clipped by the feature vector and the registered palmprint image is performed only for the extracted palmprint image, so that the accuracy is higher than the matching by the template matching by the classic still image. Entering user authentication is possible.

In the invention according to claim 8, the control unit generates a feature vector in units of the cut out pixels, obtains a degree of coincidence with a feature vector of a palmprint image registered in advance, and further has a degree of coincidence equal to or greater than a threshold value. Since the control for unlocking the electric lock is performed when the number of palm print images exceeds a predetermined number, it is possible to authenticate the occupant with high accuracy.

It is a figure which shows the system configuration | structure of the occupant authentication apparatus which concerns on Example 1 of this invention. It is a figure which shows the cross-section of the door handle shown in FIG. It is a block diagram which shows the internal structure of the image imaging | photography part and control part which are shown in FIG. It is a flowchart which shows operation | movement of the image imaging | photography part shown in FIG. It is a flowchart which shows operation | movement of the control part shown in FIG. It is a figure which shows the room entrance structure at the time of applying the room occupant authentication apparatus which concerns on Example 1 of this invention to a doorknob. It is a flowchart which shows operation | movement of the occupant authentication apparatus which concerns on Example 2 of this invention. It is a flowchart which shows the continuation of FIG. It is a figure which shows the cross-section of the doorknob shown in FIG. It is a figure which shows the system configuration | structure of the conventional occupant authentication apparatus.

  Embodiments of the present invention will be described below with reference to the accompanying drawings.

(Configuration of Example 1)
First, Embodiment 1 of the present invention will be described with reference to the drawings.
As shown in FIG. 1, the occupant authentication device 1 is attached to the entrance door 2. In the door handle 3a of the door 2 in which the electric lock 20 is incorporated, an image photographing unit 10 for photographing a palm print is embedded in a palm part (a block indicated by a broken line in the drawing) gripped when opening and closing. Here, an omnidirectional camera is used as the image capturing unit 10.

  A lock case 5 in which an electric lock 20 is incorporated is provided at the end of the door 2 to which the door handle 3 a is coupled via the connecting member 4. The lock case 5 includes a keyhole 6 and a motor or solenoid for locking or unlocking by moving a dead bolt by an electric signal supplied from the outside (a control unit 30 described later).

  FIG. 2 shows a cross-sectional structure of the door handle 3a. As shown in FIG. 2, the door handle 3a is a hollow cylinder in cross-sectional shape. The part of the cylindrical portion that comes into contact with the palm 40 of the room occupant is formed of a translucent translucent acrylic material, and in this, a special lens and a mirror are combined to view 360 degrees at a time in the horizontal direction. An omnidirectional camera that can do so is embedded. In this case, since the whole image of the palm print can be taken without depending on the position of the palm holding the door handle 3a, it is possible to authenticate the occupant with high accuracy. However, it is necessary to convert the captured omnidirectional image into a panoramic image. For this reason, it is necessary to generate an image in which the shape is not distorted by a highly accurate sampling method.

  Here, an omnidirectional camera is used as the image capturing unit 10, but a commercially available digital camera may be used instead. The digital camera may be of any type for visible light photography or infrared light photography, but in the case of infrared light photography, it has a filter structure in which an infrared shielding filter is removed and an infrared transmission filter is attached instead. To do.

  Returning to FIG. The palm print image shot by the image shooting unit 10 is collated with a palm print registered in the control unit 30 in advance. The control unit 30 performs unlocking control of the electric lock 20 when both palm prints coincide. In this state, the occupant can turn the door handle 3a to open and close the door 2.

As illustrated in FIG. 3, the occupant authentication device 1 according to the present embodiment includes an image capturing unit 10, an electric lock 20, and a control unit 30.
The image capturing unit 10 includes a touch sensor 11, a light source 12, a camera body 13, a camera control unit 14, and an image transmission unit 15.

  The touch sensor 11 is a sensor that detects the presence or absence of contact of the palm with the door handle 3a, and may employ either a resistance film type or a capacitance type detection method. The touch sensor 11 detects a signal indicating the start and end of gripping of the door handle 3a by the occupant and outputs the signal to the camera control unit 14 for moving image shooting.

  Based on the signal output from the touch sensor 11, the camera control unit 14 turns on and off the light source 12 configured by an LED (Light Emitted Diode) and the like, and starts and ends the palm print shooting of the occupant by the camera body 13. Control timing. In the case of infrared light photography, irradiation with an infrared LED as the light source 12 is desirable.

  Under the control of the camera control unit 14, the image transmission unit 15 captures the palm print image of the room occupant photographed by the camera body 13 and transmits it to the control unit 30. Note that the palmprint image may be transmitted by wire or wireless.

  The electric lock 20 is constituted by a motor or a solenoid that moves a dead bolt by an electric signal supplied from the control unit 30 and performs locking and unlocking.

  The control unit 30 includes a mode control unit 31, an image reception unit 32, an image conversion unit 33, an image recognition unit 34, an image extraction unit 35, an image registration unit 36, an image collation unit 37, and an unlock control. The unit 38 and a palm print DB (Data Base) 39 are included.

  The mode control unit 31 controls switching between a palm print registration operation (hereinafter referred to as a registration mode) performed during maintenance and an authentication operation (hereinafter referred to as an authentication mode) performed during occupant authentication. It has become.

  The image receiving unit 32 receives the palmprint image of the occupant transmitted from the image transmitting unit 15 of the image capturing unit 10 and outputs the palmprint image to the image converting unit 33. The image conversion unit 33 converts the omnidirectional palm pattern image captured and transmitted by the image capturing unit 10 when performing palm pattern authentication into a panoramic image, and at the same time, converts the palm pattern image into a palm pattern image that does not cause distortion in shape by sampling. Output to the recognition unit 34.

  The image recognizing unit 34 has a function of detecting a reference point for cutting out a partial image used for collation of palm prints from the palm print image converted by the image conversion unit 33 according to a known contour detection algorithm. Here, the base of the little finger is used as the reference point.

  The image extraction unit 35 extracts a partial image used for collation from the converted palm print image output from the image conversion unit 33 based on the reference point detected by the image recognition unit 34. One palm print image is extracted from a plurality of shot palm print images each time, and a predetermined range of pixels with the base of the little finger as a reference point is cut out for each extracted palm print image to the image matching unit 37. Output.

  The image registration unit 36 operates when the mode control unit 31 indicates “registration mode”, and registers the registered palm pattern image acquired from the image photographing unit 10 in the palm pattern DB 39. Here, the registered palm print image is converted into a feature vector by the image extraction unit 35 and registered as a palm print code. In order for the mode control unit 31 to indicate the “registration mode”, a maintenance staff connects a dedicated device to the control unit 30 and communicates with the control unit 30, and the mode control unit 31 is set to “authentication mode”. To "registration mode". In addition, when the control unit 30 has this switching function, the user needs to perform an operation with a switch or the like added to the control unit 30.

  The image collation unit 37 operates when the mode control unit 31 indicates “authentication mode”, and based on the feature vector cut out and generated by the image extraction unit 35, between the palm pattern codes registered in the palm pattern DB 39. The occupant is authenticated by checking. Specifically, the image collation unit 37 generates a SIFT feature (Scale Invariant Feature Transform) of the pixel cut out by the image extraction unit 35, and obtains the degree of coincidence with the SIFT code of the palm print image registered in advance in the palm print DB 39. Then, an unlock signal for unlocking the electric lock 20 is generated and output in accordance with the comparison result between the degree of coincidence and the threshold value.

  When the unlock signal is generated and output by the image collating unit 37, the unlock control unit 38 outputs the unlock signal to the electric lock 20 and controls the unlocking of the electric lock 20.

(Operation of Example 1)
The operation of the first embodiment will be described below.
The operation of the image photographing unit 10 in FIG. 3 will be described using the flowchart in FIG.

In FIG. 4, first, the presence / absence of a shooting start trigger is determined (step S101).
Specifically, in the image capturing unit 10 illustrated in FIG. 3, the camera control unit 14 determines the presence or absence of a capturing start trigger by monitoring a touch signal output from the touch sensor 11.

  If it is determined in step S101 that there is a shooting start trigger (YES), the process proceeds to step S102. In step S102, palm prints are photographed by light irradiation, and in step S103, a captured image is sequentially transmitted.

  Specifically, when the person in the room holds the door handle 3a in FIG. 1 and receives the signal, the touch sensor 11 detects a palm contact and outputs a touch signal in the image capturing unit 10 in FIG. Controls the light source 12 to irradiate the palm of the occupant holding the door handle with light, and at the same time, initiates palmprint photography by the camera body 13. The image transmission unit 15 sequentially transmits the captured palm print image to the control unit 30 (image reception unit 32).

  Next, in step S104, it is determined whether there is a shooting end trigger. If it is determined that there is a shooting end trigger (YES), the shooting operation is terminated. If it is determined that there is no shooting end trigger (NO), the process returns to step S102.

  Specifically, when the occupant releases his / her hand from the door handle 3a in FIG. 1, in the image photographing unit 10 in FIG. 3, the camera control unit 14 receives the release signal from the touch sensor 11 and has a photographing end trigger. judge. Then, the image transmission unit 15 sequentially transmits the palm pattern moving image photographed by the camera body 13 as 60 still images for each frame if, for example, about 2 seconds are required for one grip.

Next, the operation of the control unit 30 in FIG. 3 will be described using the flowchart in FIG. In FIG. 5, first, mode determination is performed (step S201). If the result of mode determination is “registration mode”, the process proceeds to step S202, and if it is “authentication mode”, the process proceeds to step S207. Here, the “registration mode” refers to an operation for registering the palm print of the person permitted to enter the room, and the “authentication mode” refers to performing authentication of the person entering the room and, depending on the result, performing unlock control of the electric lock. . Here, since the “authentication mode” is normally set, the captured image capturing process in step S207 is performed.

  Specifically, in the control unit 30 of FIG. 3, the mode control unit 31 determines whether the room occupant authentication device 1 is in “registration mode” or “authentication mode”. If in the “registration mode”, the mode control unit 31 validates the palm print registration operation by the image registration unit 36 and captures the registered palm print image transmitted from the image capturing unit 10 received by the image receiving unit 32.

  Next, a reference point extraction process is executed in step S203. Specifically, in the control unit 30 of FIG. 3, when the image receiving unit 32 captures the omnidirectional registered palm print image, the registered palm print image is supplied to the image conversion unit 33. In response to this, the image conversion unit 33 converts the omnidirectional palm print image into a panoramic image, performs sampling as appropriate, converts it into a palm print image with no distortion in shape, and outputs it to the image recognition unit 34. Then, in the image recognition unit 34, for example, a reference point of the little finger base is extracted and transferred to the image extraction unit 35 using a known contour extraction algorithm.

  Next, an edge extraction process using a separability filter is executed in step S204. Specifically, in the control unit 30 of FIG. 3, the image extraction unit 35 cuts out an image having a width of 360 pixels and a height of 340 pixels, for example, based on the reference point detected by the image recognition unit 34 and cut out the image. Extract edges (wrinkles) from the image. Here, the edge is divided into smaller areas, and the edge is extracted by a separability filter that takes the variance in the divided area and extracts it as an edge candidate when the variance is large. Here, the magnitude of the variance is determined by normalizing the variance of the average luminance of the divided region with the luminance variance of the entire region.

  In step S205, a shift feature (SIFT key) generation process is executed. Specifically, in the control unit 30 of FIG. 3, the image extraction unit 35 generates a SIFT key that is a feature vector from the extracted edge, and outputs the SIFT key as a palm print code to the image registration unit 36. Here, the “SIFT key” is one of well-known feature extraction approaches, and a feature vector used to obtain a singular point by paying attention to a singular luminance distribution of the minimal region divided as described above. I mean.

  Since the SIFT key is a feature extraction approach that makes it possible to authenticate more strongly against fluctuations in illumination and changes in the appearance of an object than in comparison with classic template matching, the camera is connected to a door as in this embodiment. It is considered suitable for authentication of images taken in the environment embedded in the handle 3a.

  In step S206, palmprint registration processing is executed. Specifically, in the control unit 30 of FIG. 3, the image registration unit 36 registers the palm pattern code acquired from the image extraction unit 35 in the palm pattern DB 39.

  On the other hand, if the result of mode determination in step S201 is “authentication mode”, it is determined whether or not a captured image is received (step S207). Specifically, if the mode control unit 31 is in the “authentication mode”, the palm print image authentication operation by the image collation unit 37 is validated, and the image collation unit 37 that has received this clears the built-in counter i to 0. (Step S207). The counter i is used for storing the number of coincident palm print images by one grip. Details will be described later.

  In the image receiving unit 32, the palm print image sequentially transmitted from the image capturing unit 10 (image transmitting unit 15) is received in step S 208 and output to the image converting unit 33. Thereafter, the reference point (vertex) is extracted in step S209, the edge extraction using the separability filter is performed in step S210, and the shift feature is generated in step S211. The processes performed in the “registration mode” are performed in steps S203, S204, This is the same as each processing in S205. The processing for extracting the reference point (vertex) is performed by the image recognition unit 34 in the control unit 30 of FIG. 3, and the processing for generating edge extraction and shift features by the separability filter is performed by the image extraction unit 35.

  Next, a shift feature matching process is executed in step S212, and then a matching degree calculation process is executed in step S213. Specifically, in the control unit 30 of FIG. 3, edge extraction is performed by the image extraction unit 35, and the generated SIFT feature is output to the image matching unit 37. The image collation unit 37 collates the SIFT key output from the image extraction unit 35 with the SIFT code registered in the palm print DB 39, and calculates the degree of coincidence.

  Next, in step S214, the comparison degree α is compared with the threshold value β. Here, as the threshold β, for example, a matching rate of 80% is assumed. When the coincidence rate is 80% or more, the counter i is updated by +1 (step S215), and then the value of the counter i is compared with the threshold γ (step S216). The threshold value γ to be compared here is, for example, 48 (60% of 60) of the 60 palm print images sequentially transmitted from the image capturing unit 10.

  When it is determined in step S216 that there are 48 (80% of 60) palm print images having a degree of coincidence of 80% or more, unlock control is performed (step S217), and 80% or more coincidence is determined. If it is determined that the number of palm print images is less than 48 or that the degree of coincidence is less than 80% in step S214, the process returns to step S208 to wait for reception of the next palm print image to be transmitted. Repeat the same process as above.

  Specifically, in the control unit 30 of FIG. 3, the image matching unit 37 compares the calculated coincidence α with the threshold β, and further indicates a counter indicating the number of frames of the palm print image having a coincidence of 80% or more. The value of i is compared with the threshold value γ, and the result is output to the unlock control unit 38. In other words, if there are 48 frames or more of palmprint images having 80% or more matching degree between the SIFT key of the imported occupant and the SIFT code registered in the palmprint DB 39, the image collating unit 37 is authenticated. At this time, the unlock control unit 38 generates an unlock signal and controls the unlocking of the electric lock 20. If the palmprint image having a degree of coincidence of 80% or more is less than 48 frames, the image collating unit 37 regards that the authentication has failed, and at this time, the unlock control unit 38 locks the electric lock 20 without generating a release signal. Maintain the state. That is, the control part 30 produces | generates and outputs the unlocking signal which unlocks the electric lock 20, when the ratio of the coincidence degree which is more than a threshold value and the coincidence degree which is less than a threshold value is more than predetermined value.

(Effect of Example 1)
As described above, according to the occupant authentication device 1 according to the first embodiment, the image photographing unit 10 embedded in the handle (door handle 3a) detects the beginning and end of the grip and detects the palm print between them. The control unit 30 is configured to control the unlocking of the electric lock 20 by capturing an image and comparing the captured palm print image with a previously registered palm print image. In this way, the palm print image is collated based on the video shot from the beginning to the end of the grip, so the authentication rate is improved compared to still image authentication such as fingerprints, and the occupant opens the door 2. It is possible to unlock the door in the natural action of touching the handle when entering the room.

  Moreover, according to the occupant authentication apparatus 1 according to the first embodiment, the photographing unit 10 includes a digital camera or an omnidirectional camera with an infrared transmission filter attached, and a palmprint image of the occupant photographed with these cameras. Is output to the control unit 30 to perform room entry authentication. For this reason, in the former, a digital camera can be used only by exchanging the filter, and thus the photographing unit 10 can be realized with a low-cost configuration. In the latter case, the whole image of the palm print can be taken without depending on the position of the palm holding the handle. Therefore, the palm print of the room occupant is obtained with higher accuracy than when shooting using a general digital camera. be able to.

  Further, according to the occupant authentication device 1 according to the first embodiment, the control unit 30 cuts out a predetermined range of pixels with the base of the little finger as a reference point for each acquired palmprint image, and performs matching. At the time of collation, a feature vector of the clipped pixel is generated, and the electric lock is unlocked according to the degree of coincidence with the feature vector of the palmprint image registered in advance. Further, as the feature vector, for example, by using the SIFT key, the extracted pixel and the registered palm print image are collated, so that the variation in illumination and the appearance of the object can be compared with the collation by the classic template matching. It is possible to authenticate resident who is resistant to fluctuations. Accordingly, the authentication rate is improved. At this time, the electric lock is unlocked when there are a predetermined number or more of palm print images having a degree of coincidence equal to or higher than the threshold value, so that it is possible to authenticate the occupant with high accuracy.

  Next, Example 2 will be described.

  In the first embodiment described above, the camera control unit 14 monitors the touch signal output from the touch sensor 11, and uses the timing at which the start of gripping is detected as a trigger to start shooting, and the timing at which the end of gripping is detected. Shooting of palmprint images was terminated as a trigger for the end of shooting. On the other hand, the trigger for the end of shooting is not based on the touch sensor 11 but monitors the number of palm print images transmitted by the image transmission unit 15 to capture a predetermined number (for example, 60) necessary for collation. It may be the time when the collation of the number of palm print images (for example, 48) necessary for the image collation unit 37 to determine that the authentication has been established by monitoring the output of the image collation unit 37 at the end timing. . In the second embodiment described below, the latter case where the trigger for the end of photographing is the end of collation will be described. According to the occupant authentication apparatus 1 according to the second embodiment, even if the occupant has a tendency to hold the handle for a relatively long time when opening the door, the photographing is finished without depending on the time of holding the handle. As a result, there are fewer opportunities to take useless palm print images.

(Configuration of Example 2)
In the second embodiment described below, the same configuration as that of the first embodiment shown in FIG. 3 is used, and only the operation of the control unit 30 is slightly different. 7 and 8 are flowcharts showing the operation of the control unit 30 of the occupant authentication device 1 according to the second embodiment. Hereinafter, the operation of the control unit 30 of the occupant authentication device 1 according to the second embodiment will be described with particular attention to differences from the first embodiment with reference to the flowcharts of FIGS. 7 and 8.

(Operation of Example 2)
In FIG. 7, first, mode determination is performed in step S301. As in the first embodiment, if the result of mode determination is “registration mode”, the process proceeds to step S302, and if it is “authentication mode”, the process proceeds to step S310 shown in FIG. If in the “registration mode”, the mode control unit 31 validates the palm print registration operation by the image registration unit 36 and initializes a built-in counter j that counts the number of palm print images necessary for authentication (step S302). . Then, the image receiving unit 32 captures a registered palm print image for occupant authentication transmitted from the image capturing unit 10 (image transmitting unit 15) (step S303). Here, the image transmission unit 15 of the image capturing unit 10 captures a predetermined number of palm print images that are captured after the occupant grasps the handle under the control of the camera control unit 14.

  Next, when the image receiving unit 32 captures the omnidirectional registered palm print image, the control unit 30 supplies the registered palm print image to the image conversion unit 33. In response to this, the image conversion unit 33 converts the omnidirectional palm print image into a panoramic image, performs sampling as appropriate, converts it into a palm print image with no distortion in shape, and outputs it to the image recognition unit 34. Then, the image recognition unit 34 extracts, for example, a reference point of the little finger base using a well-known contour extraction algorithm and passes it to the image extraction unit 35 (step S304).

Subsequently, in the control unit 30, the image extraction unit 35 cuts out an image having a width of 360 pixels and a height of 340 pixels, for example, based on the reference point detected by the image recognition unit 34, and an edge ( Extract wrinkles). Here, the edge is further divided into small areas, and the edges are extracted by a separability filter that takes the variance in the divided areas and extracts as an edge candidate when the variance is large (step S305). Here, the magnitude of the variance is determined by normalizing the variance of the average luminance of the divided region with the luminance variance of the entire region.

  Next, in the control unit 30, the image extraction unit 35 generates a SIFT key that is a feature vector from the extracted edge, and outputs the SIFT key as a palm print code to the image registration unit 36 (step S306). Then, the control unit 30 updates the counter j by +1 (step S307), and determines whether or not the shift feature of n (for example, 60) palmprint images necessary for authentication has been generated (step S308). When n palm print images have been received (step S308 “YES”), palm print registration processing is executed. Specifically, in the control unit 30 of FIG. 3, the image registration unit 36 registers the palm pattern code acquired from the image extraction unit 35 in the palm pattern DB 39. If n palm print images have not been received yet (step S308 “NO”), the process returns to the captured image capturing process of step S303.

  On the other hand, if it is “authentication mode” in the registration / authentication determination process in step S301, the mode control unit 31 validates the palmprint image authentication operation by the image collation unit 37, and the image collation unit 37 that has received this operation incorporates it. Counters i and k are cleared to 0 (step S310 in FIG. 8). Here, the counter i indicates the number of palm print images whose matching degree (α) is equal to or greater than the threshold value (β) as a result of collation to be described later, and the counter k is the received palm print image necessary for authentication, similar to the counter j described above. Used to count the number of each.

  When receiving the palm print image sequentially transmitted from the image capturing unit 10 (image transmission unit 15) (step S311 “YES”), the image reception unit 32 outputs the palm print image to the image conversion unit 33 each time (step S312). ). Next, the image receiving unit 32 updates the counter k by +1 (step S312), and determines whether or not the value of the counter k has reached the number required for authentication (step S313). If the number required for authentication has not yet been reached (“NO” in step S313), the image recognition unit 34 extracts a reference point (vertex) (step S314), and the image extraction unit 35, as in the case of palmprint image registration. Then, edge extraction by the separability filter is performed (step S315), and a shift feature is generated (step S316).

  Next, in the control unit 30, edge extraction is performed by the image extraction unit 35, and the generated SIFT feature is output to the image matching unit 37. The image collation unit 37 collates the SIFT key output from the image extraction unit 35 with the SIFT code registered in the palm print DB 39 (step S317), and calculates the degree of coincidence (step S318). In step S319, the degree of coincidence α is compared with the threshold value β. Here, as the threshold β, for example, a matching rate of 80% is assumed. If the coincidence rate is 80% or more (step S319 “YES”), the counter i is updated by +1 (step S320), and then the value of the counter i is compared with the threshold value γ (step S321). The threshold value γ to be compared here is, for example, 48 out of 60 palm print images sequentially transmitted from the image capturing unit 10 (80% of 60 images).

  When it is determined in step S322 that there are 48 (80% of 60) palm print images having a degree of coincidence of 80% or more, door unlock control is performed (step S322), and 80% or more coincidence is determined. If there are less than 48 palmprint images (step S321 “NO”), or if the degree of coincidence is determined to be less than 80% in the coincidence degree determination process in step S320 (step S319 “NO”) Returning to the captured image reception processing in step S311, the reception of the palmprint image to be transmitted next is awaited, and the same processing as described above is repeated. Specifically, in the control unit 30, the image collating unit 37 compares the calculated matching degree α with the threshold value β, and further, the value of the counter i indicating the number of frames of the palm print image having the matching degree of 80% or more. And the threshold value γ are compared, and the result is output to the unlock control unit 38. That is, the image collating unit 37 authenticates when there are 48 or more frames of palmprint images having a coincidence of 80% or more between the SIFT key of the captured palmprint image of the room occupant and the SIFT code registered in the palmprint DB 39. At this time, the unlock control unit 38 generates an unlock signal and controls the unlocking of the electric lock 20. If the palmprint image having a degree of coincidence of 80% or more is less than 48 frames, it is regarded as an authentication failure, and at this time, the unlocking control unit 38 maintains the locked state of the electric lock 20 without generating a release signal. .

  The above operation is performed until the value of the counter k becomes n or more in step S313, that is, 48 or more palm print images with 80% or more matching degree (80% of 60 sheets) in the image matching unit 37. It is repeatedly executed until it is determined that there is. That is, if there are 48 palm print images having a degree of coincidence of 80% or more, it is not necessary to capture 60 palm print images.

(Effect of Example 2)
As described above, according to the occupant authentication device 1 according to the second embodiment, the image photographing unit 10 embedded in the handle (door handle 3a) captures the number of sheets necessary for authentication at the beginning of gripping and authentication. A configuration in which the end is detected and a palm print image is transmitted between them, and the control unit 30 collates the captured palm print image with a plurality of pre-registered palm print images and controls the unlocking of the electric lock 20. did. In this way, since the palm print image is collated based on the moving image taken from the beginning of gripping to the time when the control device 30 captures the number of palm print images necessary for authentication, it is compared with still image authentication such as fingerprints. The authentication rate can be improved, and unlocking can be performed in a natural act such that the user touches the handle when opening the door 2 and entering the room.

In addition, according to the occupant authentication apparatus 1 according to the second embodiment, the electric lock is unlocked when there are a predetermined number or more of palmprint images having a degree of coincidence equal to or greater than the threshold value, so that it is possible to authenticate the occupant with high accuracy. Become.

Further, according to the occupant authentication apparatus 1 according to the second embodiment, since the trigger for starting photographing is set to the beginning of gripping the handle and the trigger for ending photographing is set to end of collation, the handle is gripped for a relatively long time when the door is opened. Even for a resident who has a tendency, since the photographing is completed without depending on the time for which the grip is continued, the photographing opportunity of a useless palm print image is reduced. In the second embodiment, the collation end is determined when it is determined that there are 48 (80% of 60) palmprint images having a matching degree of 80% (γ) or more among 60 palmprint images. In this case, it has been described that it is not necessary to capture 60 palm print images, but the degree of coincidence α and the threshold value γ are arbitrary. It should be noted that by monitoring the number of palm print images transmitted by the image transmission unit 15, a timing at which a predetermined number (for example, 60) required for collation is photographed and the capture of the palm print image is taken as a trigger for photographing termination. Alternatively, the timing at which a time-out is detected by monitoring the time (for example, 2 seconds) required to capture a predetermined number of palmprint images necessary for collation may be used as a trigger for ending the image capturing. Compared with, there are fewer opportunities to take useless palmprint images. .

  In the first and second embodiments, the SIFT key is used for the authentication. However, instead of using the SIFT key, the authentication may be replaced by a radial basis function (RBF network: Radial Basis Function) using the Euclidean distance. Highly accurate occupant authentication is possible.

  Although the occupant authentication device according to the first and second embodiments is applied to the door handle 3a, it can also be applied to, for example, a rotary door knob 3b. The relationship between the occupant authentication device 1 and the entrance door 2 in this case is as shown in FIG. As is clear from FIG. 6, this embodiment is the same as the first and second embodiments shown in FIG. 1 except that the occupant authentication device 1 is incorporated in the door knob 3b of the door 2 in which the electric lock 20 is incorporated.

  Further, as shown in the sectional structure of the door knob 3b in FIG. 9, the image pickup unit 10 is embedded as in the door handle 3a in FIG. However, in the case of the door knob 3b, since it is not possible to photograph with visible light, the omnidirectional camera or the commercially available digital camera used as the photographing unit 10 needs to remove the infrared shielding filter and newly install an infrared transmission filter.

  The occupant authentication device of the present invention is not limited to the door handle 3a and door knob 3b described above, but can be applied to lever handles, slide doors, and other handles used when the occupant opens the door to enter the room. It is.

  The room occupant authentication device of the present invention is a natural house that allows a room occupant to be authenticated and unlocked with high accuracy in a natural act of touching a handle when the room occupant opens the door. It is suitable for use in entrance doors and vehicle doors provided at entrances of buildings, condominiums, offices and the like.

  DESCRIPTION OF SYMBOLS 1 ... Room occupant authentication apparatus, 10 ... Image pick-up part, 11 ... Touch sensor, 12 ... Light source, 13 ... Camera body, 14 ... Camera control part, 15 ... Image transmission part, 20 ... Electric lock, 30 ... Control part, 31 ... mode control unit, 32 ... image receiving unit, 33 ... image conversion unit, 34 ... image recognition unit, 35 ... image extraction unit, 36 ... image registration unit, 37 ... image verification unit, 38 ... unlocking control unit, 39 ... Palmprint DB

Claims (8)

A occupant authentication device used for a door that is opened and closed by a handle and incorporates an electric lock for locking and unlocking,
An image capturing unit that is embedded in the handle and detects a shooting start trigger and a shooting end trigger, and captures a plurality of palm print images therebetween,
Said captured plurality of palm print images are converted into a panoramic image, to obtain a plurality of palmprint images distorted sampling and shape is converted to a plurality of palmprint images do not occur simultaneously, the imaging end trigger before Symbol shooting start A plurality of palm print images during the trigger and a plurality of pre-registered palm print images are sequentially collated, and a control unit that controls the unlocking of the electric lock according to the collation result;
A room occupant authentication device comprising:   2. The occupant authentication device according to claim 1, wherein the trigger for starting photographing is the beginning of gripping the handle, and the trigger for ending photographing is the end of gripping the handle.   2. The occupant authentication apparatus according to claim 1, wherein the shooting start trigger is the beginning of gripping of the handle, and the shooting end trigger is an end of capturing of the plurality of palm print images.   The occupant authentication device according to claim 1, wherein the trigger for starting photographing is the beginning of gripping the handle, and the trigger for ending photographing is the end of the collation. The image photographing unit
The room occupant authentication apparatus according to claim 1, further comprising an omnidirectional camera and transmitting a palmprint image captured by the omnidirectional camera to the control unit. The image photographing unit
The room occupant authentication device according to claim 1, further comprising: a digital camera having an infrared light shielding filter removed and an infrared transmission filter attached thereto, and transmitting a palmprint image photographed by the digital camera to the control unit. The controller is
For each palm print image photographed by the image photographing unit, a predetermined range of pixels with the little finger base as a reference point is cut out, a feature vector is generated for each cut out pixel, and the pre-registered palm print image The room occupant authentication device according to claim 1, wherein a degree of coincidence with the feature vector is obtained, and an unlocking signal for unlocking the electric lock is generated and output according to a result of comparison between the degree of coincidence and a threshold value. The controller is
8. The room occupant authentication device according to claim 7, wherein when the number of palm print images having a degree of coincidence equal to or greater than the threshold is equal to or greater than a predetermined number, an unlock signal for unlocking the electric lock is generated and output.

Images