JP2003242487A - Image recognition device - Google Patents

Abstract

[PROBLEMS] To prevent image signal shading due to uneven brightness of a light source, an illumination environment, and the shape and position of a subject in an image recognition device provided with a light emitting means in a peripheral portion of an imaging area. . SOLUTION: The light emission amount of a peripheral light emitting means is changed,
Obtain image information of a plurality of frames by means such as changing the exposure time of the sensor, changing the threshold voltage when binarizing the image signal, and correcting the image signal shading by synthesizing them.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image recognition device, and can be applied to, for example, a fingerprint collation device. The present invention corrects the peculiar exposure unevenness that occurs in an image recognition device having a light emitting means in the periphery of the image pickup area by means such as changing the amount of light emitted by the light emitting means and changing the exposure time of the optical sensor to obtain an image. Make it possible to improve recognition accuracy.

[0002]

2. Description of the Related Art In a conventional image recognition apparatus of this type, an optical system such as a lens, a prism, and a fiber and a light emitting means are provided, and light is emitted from a predetermined position and the reflected light is two-dimensionally imaged by the optical system. Mostly, it is configured to output an image to a sensor. However, in this configuration, there is a problem that the system becomes large depending on the size of the optical member itself such as a lens and the arrangement position of the light emitting means.

On the other hand, Japanese Unexamined Patent Publication No. 2000-217803
In order to obtain an image without distortion, a thin device, a small size, and a low price have been devised.

[0004]

However, in the case of the method devised by the above-mentioned Japanese Patent Laid-Open No. 2000-217803, the exposure irregularity peculiar to the subject is caused by the shape and position of the object, the luminance irregularity of the light source and the illumination environment. Therefore, it is an object of the present invention to provide an image recognition apparatus capable of appropriately correcting this exposure unevenness.

[0005]

In order to solve the above problems, an image detecting means for optically detecting a subject, and a light emitting means provided on the periphery of the image detecting means for irradiating the subject with light. A driving unit that acquires a plurality of frame images under different exposure conditions in the image detection unit, a storage unit that holds the frame images detected by the image detection unit, and a frame image of each frame image of the plurality of frame images. An image recognition device having an image composition means for composing a partial image which is a part.

Further, an image detecting means for optically detecting the subject, a light emitting means provided on the periphery of the image detecting means for irradiating the subject with light, and the image detecting means for acquiring a plurality of frame images. The driving unit, the binarizing unit that binarizes the plurality of frame images based on different thresholds, and the partial image that is a part of each frame image of the binarized frame images are combined. And an image synthesizing unit for performing the image recognizing process.

[0007]

BEST MODE FOR CARRYING OUT THE INVENTION The present invention will be described below with reference to the drawings, using a fingerprint input device as an example.

FIG. 1 is a block diagram showing the overall configuration of a fingerprint collation apparatus 1 which is one of the embodiments of the present invention. In the fingerprint collation device 1, the system control unit 3 responds to a user's request input from the key operation unit 2.
Controls the whole operation by. In response to a user's fingerprint registration request, the registered fingerprint data D1 input from the fingerprint data input unit 4 is recorded in the memory together with other user data, and the database 5 is constructed.

On the other hand, when the user inputs a request for fingerprint collation, the verification fingerprint data D2 is fetched through the fingerprint data input unit 4 and temporarily stored in the verification fingerprint memory 6. The verification fingerprint data D2 and the registered fingerprint data D1 registered in the database 5 are input to the fingerprint collation unit 7, where the degree of coincidence between these data is detected. The system control unit 3 determines whether or not the fingerprints match, depending on the degree of matching, and outputs the result to the display unit 8.

When a plurality of data are registered in the database 5, the fingerprint collation device 1 sequentially collates from the first registered data to determine whether the user who requested the fingerprint collation is the corresponding person. As the number of registered data in the database 5 increases, the collation takes more time. Therefore, when registering a fingerprint, the key operation unit 2
It is also possible to greatly shorten the matching time by prompting to input other information such as a personal identification number in advance through the above, and by inputting this information also when a matching request is made.

FIG. 2 is a block diagram showing the configuration of the fingerprint data input unit 4 in the fingerprint collation device 1. First,
The light receiving process of the two-dimensional image sensor will be described. The finger 11, which is a subject, is placed at a predetermined position on the two-dimensional image sensor 12 via the protective cover 13. As shown in an enlarged view of the contact surface in FIG. 3, at this time, since the ridge line portion of the fingerprint contacts the protective cover 13, the transmitted light passing through the inside of the finger 11 directly passes through the protective cover 13 and the two-dimensional image sensor 12 is passed.
Is received by. On the other hand, since a space is created between the valley of the fingerprint and the surface of the cover 13, the finger 11 released from the valley
The transmitted light from the inside is the skin and the space, and the space and the cover 13.
Light that is refracted and scattered at the interface between the surfaces and attenuated is received by the two-dimensional image sensor 12. Therefore, in the two-dimensional image sensor 12, the ridge line portion that is in contact with the surface of the cover 13 is a bright area where light is clearly detected, while light is not detected so much in the valley area and is a dark area. To be done.

The transmitted light from the inside of the finger 11 is generated by the outside light around the finger 11. If this outside light is not sufficient for fingerprint image pickup, the light emitting means 1 such as LED provided around the image pickup area.
4 is made to emit light.

By the way, the light incident from the upper part of the finger is scattered and attenuated inside the finger before being transmitted to the ball of the finger. Therefore, even if the amount of light incident on the upper part of the finger is the same, the amount of transmitted light emitted from the contact surface with the protective cover 13 varies depending on the thickness of the individual finger. Further, since the fingers are usually cylindrical, the thickness is not uniform from the contact surface with the protective cover 13. As a result, uneven exposure occurs in which the transmitted light is largely attenuated in the central portion having a large thickness and the exposure is insufficient, but the peripheral portion having a relatively small thickness is appropriately exposed.

Even when the peripheral light emitting means is used, this uneven exposure cannot be suppressed because the light from the light emitting means does not easily reach the central portion. Further, when the light emitting means 14 is used, by disposing the finger not on the center of the imaging region but on a certain side, the distance from the light emitting means becomes non-uniform, which also causes uneven exposure.
In other words, in the image recognition device that includes the light emitting means in the peripheral part of the image pickup area and picks up the transmitted light, the degree and distribution of the exposure unevenness that occurs varies depending on the individual situation, and an appropriate correction is required each time.

Specifically, FIG. 4 shows an image taken by turning on the light emitting means 14 and a luminance level in one horizontal scanning period in which the image is present. In this way, the peripheral portion 15 of the finger near the light emitting means 14 has a sufficient amount of transmitted light and is imaged with proper exposure, whereas the central portion 16 of the finger has a small amount of light and is imaged darkly. Therefore, it becomes difficult to extract the fingerprint feature points in the central portion, and the fingerprint recognition rate decreases.

(Embodiment 1) In the first embodiment of the present invention, the exposure unevenness is changed by changing the light emission amount of the light emitting means 14, and a plurality of frames having different exposure amounts are imaged. Correct by combining the images.

Specifically, in the image pickup apparatus in which three light emitting means 14 (LEDs) are arranged in the periphery of FIG. 5, first, the light emitting means 14 in the middle of each piece is turned on to capture the first frame image D11. In the first frame image D11, the peripheral portion 15 of the finger is imaged within the proper exposure range and the fingerprint pattern can be extracted, but the central portion 16 is an image in which the fingerprint pattern cannot be extracted due to insufficient exposure. (Fig. 6). This first frame image D11 is an 8-bit AD
After being converted into 256 gradations by the converter 17, it is held in the inside frame memory 18.

Subsequently, all the light emitting means are turned on to capture the second frame image D12, which is converted into a digital value by an 8-bit AD converter like the first frame image, and then stored in the frame memory 18. In the second frame image D12, the central portion 16 is imaged within the proper exposure range, but the peripheral portion 15 is exposed beyond the saturation level of the sensor, and as a result, a uniformly white image is obtained and the fingerprint pattern is extracted. Cannot be done (Fig. 7).

Next, in the frame image data held in these frame memories, the effective segment is detected in the system control unit 3 constituted by the microcomputer according to the processing flow shown in FIG.

First, in step 1, the frame image D
11 and D12 are each divided into three in the horizontal direction and four in the vertical direction to form a 3 × 4 segment structure (FIG. 10).
(A) and FIG. 11 (a)). In step 2, in a certain segment, the output level of each pixel within one horizontal scanning period is monitored, and the number of output level differences from the previous pixel exceeding a certain threshold value is counted. Subsequently, the system control unit 3 determines in step 3 whether or not this count number is equal to or larger than a predetermined value. In the case of affirmative determination, the process proceeds to step 4 and the corresponding segment is determined to be a valid segment exposed within the proper exposure range. In the case of negative determination, the process proceeds to step 5 and it is determined that the segment is an invalid segment. In the case of a pattern with a high contrast such as a fingerprint or a bar code, when monitoring each pixel within a certain horizontal scanning period, if the output level is uniform over several tens of pixels, or if the output level changes gently, exposure is performed. Since it can be said that is not appropriate, the effective segment can be effectively detected by the above-mentioned method.

Subsequently, in step 6, it is judged whether the valid / invalid judgment has been made for all the segments. In the case of negative determination, the processing returns to step 2 and the same processing is repeated for the next segment. When the determination is affirmative, the process proceeds to step 8, and the image synthesizing process is subsequently performed.

The image synthesizing unit 19 is controlled by the system control unit 3 and forms one fingerprint image data only by the selected effective segments D11 and D12. Since the effective segment positions are different between D11 and D12 having different exposure amounts, the fingerprint collation device 1 is configured to obtain a proper exposure image in a wider area by synthesizing the two frame image data. Is being improved.

FIG. 8 shows a processing flow of the system controller 3 in the image synthesizer 19. In the image synthesis processing, the system control unit 3 moves from step 8 to step 9,
Frame memory 1 for the effective segment of the first frame image
8 is read (FIG. 10 (b)). D1 in step 10
The number of valid segments of 1 is counted, and in step 11, D
Only 11 effective segments are rearranged on the frame memory 18. Subsequently, in steps 12 and 13, the system control unit 3 performs the same processing on D12 as that on D11 (FIG. 11B).

In step 14, the system controller 3 compares the number of valid segments of D11 counted in step 9 with the number of valid segments of D12 counted in step 13.

When it is determined that the number of valid segments of D11 is larger, the system control unit 3 proceeds to step 15 and advances the processing. In step 15, the effective segment of D12 is rearranged in the same frame memory area as the area in which the effective segment of D11 is rearranged, and image composition is performed. Do not replace. on the other hand,
If it is determined that the number of effective segments in D12 is larger, the process proceeds to step 16. In step 16, D
When rearranging the 12 valid segments in the same frame memory area as the location of the valid segment data of D11, a process for replacing the data of the segment in which the segment data of D11 is already arranged is performed.

In this example, the image information is held in both D11 and D12 for the (2,1) segment.
The image information of D11 is adopted for the (2,1) segment because the number of effective segments of is larger.

In this way, when both the segments at the same position of D11 and D12 are determined to be effective segments, a frame image having a large number of segments to be used as the entire image is adopted, so that the image data between the segments can be changed. It is possible to synthesize frame images with less tearing. The term "disruption" as used herein means that the sensor output level fluctuates due to different exposures on the boundary line of the combined segment, and the system misrecognizes this output fluctuation as a fingerprint pattern and deteriorates the recognition accuracy. In order to prevent this erroneous recognition, the image information of the segment boundary line is not used as information for matching. Further, in the above method, a part of the image information captured at a certain angle is rejected, and as a result, the fingerprint matching accuracy is also deteriorated. It is also possible to generate a composite image with a small number and improve the accuracy of fingerprint collation.

The fingerprint image data synthesized here is stored in the database 5 as registered fingerprint data D1 when the user's request input from the key operation unit is a fingerprint registration request.
Written in. On the other hand, if the fingerprint collation request is issued, the collation fingerprint data D2 is temporarily stored in the collation fingerprint memory.

The fingerprint matching unit 7 checks the correlation between the registered fingerprint data D1 registered in the database and the verification fingerprint data D2 to perform fingerprint matching. As the fingerprint collating unit 7, a generally used one can be used.

The system control unit 3 receives the result of the fingerprint collation unit 7, determines whether the fingerprints match or not, and displays the result to the user via the display unit 8.

Regarding the adjustment of the amount of light emission, it is considered that more flexible exposure adjustment can be performed by controlling by pulse driving instead of changing the number of light emission of the light emitting means.

(Embodiment 2) The fingerprint collation apparatus according to the first embodiment described above acquires a plurality of frame images having different exposure amounts by changing the light emission amount of the light emitting means, and combines them to correct the exposure unevenness. It is designed to do
In the present embodiment, a plurality of frame images with different exposures are obtained by changing the light accumulation time of the two-dimensional image sensor.

That is, using the two-dimensional image sensor having the electronic shutter function, the time from resetting the photoelectric conversion unit such as the photodiode and the diode in the two-dimensional image sensor to reading the signal from the photoelectric conversion unit. Is changed for each frame image, and the light accumulation time is changed.

This embodiment has the same structure as that of the above-mentioned first embodiment. In the present embodiment, in the image pickup apparatus having the light emitting means in the periphery of the image pickup area of FIG. 4, first, all the light emitting means of each piece are turned on, and the light accumulation time of the two-dimensional image sensor 12 is set to 100 ms in the first frame. The image D13 is captured. For the sake of convenience, it is assumed that D13 captured here is the same as the first frame image D11 in the first embodiment shown in FIG. That is, the peripheral portion 15 of the finger is imaged within the proper exposure range and the fingerprint pattern can be extracted, but the central portion 16 is an image in which the fingerprint pattern cannot be extracted due to insufficient exposure. The first frame image D13 captured here is digitized into 256 gradations by the 8-bit AD converter 17, and then is temporarily held in the frame memory 18.

Subsequently, the light emitting means of each piece are all turned on as in the case of the first frame image pickup, and the system control unit 3 turns on the light.
The second frame image D14 is captured so that the light accumulation time of the three-dimensional image sensor is doubled to 200 ms. Also for this D14, for convenience, it is assumed that the same image as the second frame image in the first embodiment shown in FIG. 6 is captured.
That is, the central portion 16 is imaged within the proper exposure range, but the peripheral portion 15 is exposed beyond the saturation level of the sensor, and as a result, a uniform white image is obtained and the fingerprint pattern cannot be extracted. is there. In addition, the second frame image D14 captured here is the same as the first frame image D13.
After being digitized into 256 gradations by the bit AD converter 17, it is once held in the frame memory 18.

D13 stored in these frame memories
The D14 frame image and the D14 frame image are detected by the system control unit 3 as in the first embodiment (FIG. 8).
Process flow shown in FIG. 9) and image synthesis process (process flow shown in FIG. 9) are performed to form synthetic fingerprint image data imaged within the appropriate exposure range.

The fingerprint image data obtained by the image synthesizing process is stored in the database 5 or subjected to the fingerprint collating process according to the user's request.

(Third Embodiment) In the third embodiment, the fingerprint data input unit 4 has the structure shown in FIG. The image signal output from the two-dimensional image sensor 12 is converted into digital data by the AD converter 23 and then temporarily held in the image memory 22. This digital image data is binarized in the comparison circuit 21 based on the threshold value output from the threshold value generator 20. In the present embodiment, the exposure unevenness peculiar to the transmitted light imaging system is corrected by this threshold value.

That is, in the case of an image recognition apparatus in which a subject is brought into close contact with the image pickup surface and the transmitted light from the subject is picked up, it has already been described that a specific exposure unevenness occurs as shown in FIG. If binarized with a fixed threshold value, it is difficult to obtain a binarized image signal accurately in the central portion where the amount of light is small, and therefore the image recognition rate is significantly reduced.

Therefore, in this embodiment, a plurality of binarized frame images obtained by changing the threshold value are combined to suppress a decrease in the image recognition rate caused by uneven exposure.

FIG. 13 is a flow chart showing a process for obtaining binarized frame image data in this embodiment.

First, in step 30, the finger 1 which is the subject
The fingerprint image of 1 is taken. This image data is step 3
In step 1, the signal is digitized by the AD converter 23 and is once held in the image memory 22 (step 32). Next, at step 34, the comparison circuit 21 generates a binarized frame image D15 based on the threshold value Ref1 output from the threshold value generating section 20. The D15 generated here is temporarily held in the frame memory 18 in step 35.

Subsequently, in step 36, it is determined whether the number of frame images acquired by changing the threshold value has reached the number required by the system. If the determination is affirmative, the process proceeds to the effective segment detection processing of step 38, but if the determination is negative, the threshold value is changed in step 37 to generate a different binarized frame image. Here, assuming that the system requires two frame images, the threshold level is lowered and a different binarized frame image D16 is generated and held in the frame memory 18.

D15 and D16 held in the frame memory 18 are subjected to effective segment detection processing and image composition processing as in the first and second embodiments. This effective segment detection processing and image composition processing are basically shown in FIG.
Since it is the same as the flow shown in FIG. 9 and FIG. 9, detailed description will be omitted here. The different point is that the frame image is binarized in the third embodiment, so in the effective segment detection processing step 2 of FIG. 8, the number of switching from low level to high level is counted, and the effective / ineffective is determined based on the result. The point is to make a determination.

In this embodiment, a plurality of frame images D
15, 15 may be acquired under the same exposure condition, or may be acquired under different exposure conditions as described in Embodiments 1 and 2 above.

Incidentally, the first and second embodiments shown so far.
Further, in the third embodiment, when image data with less exposure unevenness is required for collation with a high security level, the number of frame images used for composition is increased. In other words, in order to improve the matching accuracy and secure the security level, it is necessary to set the judgment conditions for judging proper exposure to be strict and generate a clear composite image. However, by tightening the proper exposure conditions, it is effective within one frame. Since the number of segments decreases and image composition using only two frames cannot secure a sufficient number of effective segments, the collation accuracy deteriorates. Therefore, by increasing the number of frame images to be combined A plurality of pieces of information with different exposure conditions are held so that the proper exposure condition required by the system is satisfied.

As described above, according to the present invention, it is possible to improve the matching accuracy by making a trade-off with the time required for the matching.

[0048]

As described above, according to the present invention, the peripheral portion of the image recognition apparatus is provided with the light emitting means as the auxiliary light in the peripheral portion of the image pickup area, and the subject is brought into close contact with the image pickup element to pick up the transmitted light from the subject. By capturing a plurality of frame images by means such as changing the light emission amount of the light emitting means, changing the exposure time of the image pickup element, changing the threshold value when binarizing the image, and combining them. The exposure unevenness peculiar to the image recognition device that images the transmitted light is suppressed. Therefore, according to the present invention, in an image recognition device such as a fingerprint sensor or a bar code, the image recognition range can be greatly expanded, and the image recognition rate can be improved. In addition, it is not necessary for the user to change the positional relationship between the image recognition target and the image recognition device when acquiring a plurality of frames, and convenience can be improved. In addition, it is possible to increase the number of frames to be acquired and combined and improve the image recognition accuracy by making a trade-off with the time required for recognition.

[Brief description of drawings]

FIG. 1 is a fingerprint collation device 1 according to an embodiment of the present invention.
3 is a block diagram showing the overall configuration of FIG.

FIG. 2 is a fingerprint data input unit 4 of the fingerprint matching device 1 of FIG.
It is the block diagram which showed.

FIG. 3 is an enlarged view of a contact surface between a finger 11 and a protective cover 13 in FIG.

FIG. 4 is a diagram for explaining exposure unevenness of a fingerprint image by the fingerprint data input unit of FIG.

5 is a diagram showing a specific configuration for correcting the exposure unevenness of FIG.

FIG. 6 is a diagram showing a first frame image captured with the configuration of FIG.

FIG. 7 is a diagram showing a second frame image captured with the configuration of FIG.

FIG. 8 is a flowchart showing an effective area selection process by the system control unit 3.

FIG. 9 is a flowchart showing an image synthesizing process by the system control unit 3.

FIG. 10 is a diagram for explaining the effective segment detection processing of the first frame image.

FIG. 11 is a diagram for explaining an effective segment detection process of the second frame image.

FIG. 12 is a block diagram showing a fingerprint data input unit 4 of the fingerprint collation device 1 of FIG. 1 in a third embodiment.

FIG. 13 is a flowchart showing a process of acquiring a binarized frame image in the third embodiment.

[Explanation of symbols]

1 Fingerprint matching device 2 key operation part 3 System control unit 4 Fingerprint data input section 5 database 6 Referral fingerprint memory 7 Fingerprint matching section 8 Display 11 fingers 12 Two-dimensional image sensor 13 Protective cover 14 Light emitting means 15 finger area 16 finger center 17 AD converter 18 frame memory 19 Image synthesizer 20 Threshold generator 21 Comparison circuit 22 Image memory 23 AD converter D1 registered fingerprint data D2 collation fingerprint data D11 First frame image (Example 1) D12 Second frame image (Example 1) D13 First frame image (Example 2) D14 Second frame image (Example 2) D15 First binarized frame image (Example 3) D16 Second binarized frame image (Example 3)

   ─────────────────────────────────────────────────── ─── Continued front page    F-term (reference) 4C038 FF01 FF05 FG01                 5B047 AA25 AB02 BA02 BB04 CA05                       CA19 DB06 DC20                 5B057 BA02 BA12 BA30 CA08 CA12                       CA16 CB06 CB12 CB16 CC03                       CE10 CE12

Claims (7)

[Claims] 1. An image detecting means for optically detecting a subject, a light emitting means provided in a peripheral portion of the image detecting means for emitting light toward the subject, and a plurality of frames under different exposure conditions in the image detecting means. A driving unit that acquires an image; a storage unit that holds the frame image detected by the image detecting unit; and an image synthesizing unit that synthesizes a partial image that is a part of each frame image of the plurality of frame images. Image recognition device having. 2. The image recognition device according to claim 1,
An image recognition apparatus having a driving means for changing the number of light emission of the light emitting means to acquire a plurality of images under different exposure conditions. 3. The image recognition device according to claim 1,
An image recognition apparatus having a drive unit that changes a light emission amount of a light emitting unit to acquire a plurality of images under different exposure conditions. 4. The image recognition device according to claim 1,
An image recognition apparatus having a driving unit that acquires a plurality of images with different exposure conditions by changing the exposure time of the image detection unit. 5. The image recognition device according to claim 1,
An image recognition apparatus having a driving unit that changes a threshold value and acquires a plurality of different binarized images when the image obtained by the image detection unit is binarized based on a predetermined threshold value. 6. An image detecting means for optically detecting a subject, a light emitting means provided on the periphery of the image detecting means for irradiating the subject with light, and a plurality of frame images are acquired by the image detecting means. The driving unit, the binarizing unit that binarizes the plurality of frame images based on different thresholds, and the partial image that is a part of each frame image of the binarized frame images are combined. Image recognizing device having image synthesizing means for performing the image synthesizing. 7. The image recognition apparatus according to claim 1, wherein the frame image is divided into a plurality of the partial images, and the appropriate exposure determination means determines the appropriateness of exposure for each partial image. An image recognition device having.