Personal identification device

Description: TECHNICAL FIELD [Detailed description of the invention]

【0001】

[Industrial application field]

The present invention relates to a personal identification device that performs personal identification processing as to whether or not a person is a person by using a face image such as room entry management, and sequentially searches and outputs a person close to the input face image from a data file.

【0002】

[Conventional technology]

In the conventional personal identification process using a face image, for example, there is one described in Japanese Patent Application Laid-Open No. 63-177273. The configuration will be described below.

【0003】

FIG. 9 is a block diagram showing a configuration example of a conventional personal identification device using a face image. In the figure, 11 is an image input unit for inputting a face image, 12 is a feature point extraction unit composed of an A / D converter, a central processing unit, a memory, etc., and 13 is a feature point of the person's face image on a card or the like. The recorded personal database unit 14 uses the feature point information from the feature point extraction unit 12 and the personal database unit 13 to detect the difference, and whether the obtained difference is larger than a certain threshold or not is the person himself / herself. It is a judgment unit that performs personal identification processing.

【0004】

[Problems to be Solved by the Invention]

When the personal identification process is performed in the above-mentioned conventional technique, the feature points of the face image input from the image input unit 11 are extracted by the feature point extraction unit 12, and the feature points and the feature points are registered in the personal database unit 13. The determination unit 14 detects the difference from the feature points of the face image, and performs personal identification processing as to whether or not the person is the person according to the magnitude of the difference. At that time, if there is a difference in image size and shooting angle between the two face images, the difference between the feature points becomes large, and even the same person may be regarded as a different person.

【0005】

Further, in order to prevent such a situation, it is necessary to fix the distance between the image input unit and the photographed person and have the image input facing the front. However, such an input method imposes a burden on the person to be photographed, and even if the person is facing the front, the face is slightly rotated or tilted, so the image is accurately faced to the front. There was a problem that it was difficult to do.

【0006】

The present invention is for solving the above-mentioned problems, and an object of the present invention is to correct the rotation angle of a face image having a rotation angle in the left-right direction and perform personal identification processing with high accuracy.

【0007】

[Means for solving problems]

In order to solve the above problems, the present invention has a reference face model section and a correction section, and is in the left-right direction based on the feature points obtained from the feature point extraction section and the three-dimensional structure from the reference face model section. It is characterized in that the rotation angle of the above is corrected, and personal identification processing is also performed on a face image having a rotation angle in the left-right direction.

【0008】

[Action]

According to the present invention, the position information of the feature points can be obtained from the feature point extraction unit. Based on these position information and the depth information of the reference face model part, the rotation angle in the left-right direction is obtained. From the obtained rotation angle and the depth information of the reference face model unit, the position information of the feature point is corrected by the correction unit so as to eliminate the rotation angle, and the personal identification process is performed using the corrected feature point.

【0009】

[Example]

FIG. 1 is a block diagram showing a configuration of an embodiment of the invention according to claim 1. In the figure, 1 is an image input section for inputting a face image, 2 is a feature point extraction section for extracting feature points of a face image, 3 is a reference face model section for describing a three-dimensional structure of a face, and 4 is the feature point extraction section. The correction unit that corrects the rotation of the face in the left-right direction based on the feature points extracted by the unit and the three-dimensional structure of the reference face model unit, and 5 is the feature points of the face image of the person to be personally identified. The recorded personal database unit 6 is a determination unit that calculates the distance between the feature points corrected by the correction unit 4 and the feature points in the personal database unit 5 and performs personal identification processing as to whether or not the person is the person.

【0010】

Specifically, the image input unit 1 captures a color face image of the identification target person using a television camera and inputs it to the feature point extraction unit 2. In this feature point extraction unit 2, the input image is A / D converted and converted into a digital image, and then contour extraction, binarization processing, etc. are performed based on brightness information, hue information, etc. to perform eyebrows, eyes, nose, etc. Extract the position information of each feature point such as the mouth. For this, for example, the apparatus described in "Face image feature point extraction apparatus" for which a patent application has already been filed (Japanese Patent Application No. 4-9753) is used. The position information of the obtained feature points is converted into (x, y) two-dimensional coordinates with an arbitrary point, for example, the apex of the nose as the origin.

【0011】

Figure 2 shows an example of the feature points of a person's face image. In this embodiment, the number of feature points P is 30 (indicated by the numbers 0 to 29). A set of each xy coordinate value of the feature points extracted by the feature point extraction unit 2 P0 ~ P29: Pn = (xn, yn) And the depth information of each feature point in the reference face model part 3 Z0 ~ Z29 And.

【0012】

Next, the correction unit 4 obtains the rotation angle θ in the left-right direction of the face. For this, feature points that pass through the center line of the face are used. In this embodiment, the feature points P24, P25, P26, and P28 shown in FIG. 2 pass through the center line 30 of the face. Here, P24 (x24, y24) will be used for explanation. First, find the x-coordinate Cx of the center line 30 of the face. There are methods such as using the x-coordinate of the apex P0 and using the average of the feature points of the facial contour. Then, the rotation angle θ in the left-right direction at the feature point P24 is based on the equation of Equation 1. [0013]

[Number 1]

θ = Sin ~¹((x24-Cx) / Z24) Is sought after.

【0014】

Further, for all the vertices Pn (xn, yn) (0 n 29) in the correction unit 4, the obtained rotation angle θ is used by the equation of Equation 2. [0015]

[Number 2]

<img file="JPH06168317A_D0001.tif" />【0016】

Β is obtained from the simultaneous equations related to α and β of, and the x-coordinate value of the feature point is newly calculated by the equation of Equation 3. [0017]

[Number 3]

xn = β + Cx By doing so, the rotation of the face in the left-right direction is corrected. The above schematic diagram is shown in FIG. 3, and an example of a face image rotated to the left when facing the front is shown in FIG.

【0018】

Next, the set of x and y coordinate values of the feature points of the face image of a certain person in the personal database section 5 Pdt0 ~ Pdt29 Then, for example, the total sum S of the distances of all the feature points is calculated by the determination unit 6 by the formula of equation 4. [0019]

[Number 4]

<img file="JPH06168317A_D0002.tif" />【0020】

If the sum of the distances to a certain person S is equal to or less than a certain threshold value Th, it is determined that the input face image is the person himself / herself, and the result is output. It is also possible to calculate the sum S of the distances to all the people and output the results in ascending order of the sum S.

【0021】

Here, the rotation angle θ shown in FIG. 3 was obtained from one feature point P24, but it is more accurate to obtain the rotation angle θ from a plurality of feature points and take the average. Further, by having the depth information of the center line 30 of the face in the reference face model unit 3, it is possible to obtain the angle of rotation by using points such as the centers of both eyebrows and the centers of both eyes.

【0022】

Further, in the case of the invention of claim 2 of the present invention, by assuming that the three-dimensional shape of the face is a cylinder, the storage capacity of the reference face model unit 3 and the calculation amount of the correction unit 4 can be reduced.

【0023】

FIG. 4 is a block diagram showing a configuration of an embodiment of the invention according to claim 3. In the figure, 7 is an enlargement / reduction section, and other blocks having the same function as those in FIG. 1 are indicated by the same reference numerals. The feature point Pn obtained by the correction unit 4 in the previous stage is subjected to enlargement / reduction processing in the enlargement / reduction unit 7. Specifically, this process involves the scaling factor a and the y-coordinate scaling factor a._yFor all feature points Pn (xn, yn) (0 n 29) using [0024]

[Number 5]

xn = xn × a yn = yn × a × a_yIs processed. The enlargement / reduction coefficient a is used to absorb the error in the distance from the image input unit 1 to the photographed person, and the y-coordinate compression coefficient a is used._yIs used to absorb the difference between the input device used when creating the personal database unit 5 and the device of the image input unit 1. Here, for example, (0.80 a 1.20, 0.01 step: a_y= 1.00) is set, and in this example, a total of 41 total distances S obtained by the determination unit 6 are obtained for each step of the enlargement / reduction coefficient a, and the minimum value Smin among them is used as the input image. The distance to the data in the personal database section 5 is used, and the result is output based on this distance.

【0025】

5 and 6 are block diagrams showing the configuration of each one embodiment of the inventions according to claims 4 and 5 of the present invention. In the figure, 8 is a data rotation unit, FIG. 5 shows an example in which the data rotation unit 8 is added after the feature point extraction unit 2, and FIG. 6 shows the data rotation unit 8 and the enlargement / reduction unit 7 shown in FIG. This is an example. In addition, the same functional blocks as those in FIGS. 1 and 4 are indicated by the same reference numerals. The data rotation unit 8 corrects the inclination of the face in the face image that is inclined with respect to the front surface of the face illustrated in FIG. 8, and specifically, the inclination angle is set to φ, and the feature point extraction unit 2 obtains the data. All features For Pn (xn, yn) (0 n 29), by the equation of equation 6 [0026]

[Number 6]

<img file="JPH06168317A_D0003.tif" />【0027】

The feature point whose inclination is corrected by performing Pn (xn , yn ) (0 n 29) To generate. Here, the inclination angle φ is generated using the feature point Pn obtained in the feature point extraction unit 2. For example, as in the invention according to claim 6 of the present invention, using the feature points P17 (x17, y17) and P23 (x23, y23) of the left and right outer corners of the eye, the equation of Equation 7 is used. [0028]

[Number 7]

φ = tan ~¹((y23-y17) / (x23-x17)) Ask.

【0029】

[Effect of the invention]

As described above, the personal identification device of the present invention has a reference face model unit even when there is a left-right rotation angle between the input face image and the face image registered in the personal database unit. By having a reference face model in the face, detecting the rotation angle of the face image in which the face is tilted with respect to the front of the face, and correcting this, it is possible to prevent a decrease in the recognition rate of the face image.

[Simple explanation of drawings]

[Figure 1]

It is a block diagram which shows the structure of one Example of the invention of claim 1 of this invention.

[Figure 2]

It is a figure which shows an example of the feature point P of the face image of a person.

[Fig. 3]

It is a schematic diagram of the rotation correction in the left-right direction of a face.

[Fig. 4]

It is a block diagram which shows the structure of one Example of the invention of claim 3 of this invention.

[Fig. 5]

It is a block diagram which shows the structure of one Example of the invention of claim 4 of this invention.

[Fig. 6]

It is a block diagram which shows the structure of one Example of the invention of claim 5 of this invention.

[Fig. 7]

It is a figure which shows the example of the face image rotated to the left toward the front.

[Fig. 8]

This is an example of a face image in which the front of the face is tilted.

[Fig. 9]

It is a block diagram which shows one configuration example of the personal identification device by a conventional face image.

[Explanation of symbols]

1 ... image input section, 2 ... feature point extraction section, 3 ... reference face model section, 4 ... correction section, 5 ... personal database section, 6 ... judgment section, 7. .. Enlargement / reduction part, 8 ... Data rotation part.