JP2009244921A - Face image characteristic extraction method, its device and its program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To perform face image characteristic extraction for preventing erroneous recognition since featured values never appear as any large difference even when the direction of the face is different from that during registration. <P>SOLUTION: The positions of characteristic sections are extracted from the input face image. Position coordinates are assigned to the extracted characteristic sections, and the area of a first region where the characteristic sections are connected as vertexes is calculated. The area of a second region as a region where the characteristic sections are connected as vertexes, that is, a region different from the first region is calculated. The rate of the area of the first region to the area of the second region is calculated, and the value obtained by the pertinent calculation is obtained as face image featured values. <P>COPYRIGHT: (C)2010,JPO&INPIT

Description

  The present invention relates to feature amount extraction in an apparatus that recognizes a face by inputting a face image with a camera.

  In the fields of monitoring systems and biometric authentication of information processing devices, it is necessary to improve the accuracy of facial image analysis.

As an invention related to this point, for example, as in the technique described in Patent Document 1, there is a technique in which features of an acquired face image (such as the position and shape of eyes, nose, and lips) are directly used as feature amounts. Yes (see, for example, Patent Document 1).
In the technique described in Patent Document 1, imaging information of the image sensor of the information processing apparatus is input to the skin color detection unit and the biometric authentication unit. Then, the skin color detecting means searches the skin color part in the imaging information and detects the contour of the face. In parallel, the biometric authentication means searches for parts constituting the face in the imaging information, detects the face, determines whether the face is registered by comparing the search result with the face information, and realizes face authentication. .
JP 2006-293732 A

  As described above, in the technique described in Patent Document 1, the features of the acquired face image (such as the positions and shapes of eyes, nose, and lips) are used as feature amounts.

  However, there is a problem with the technique described in Patent Document 1. That is, since the positions of the eyes, nose, and lips are used as feature amounts, if the face direction changes greatly, the positions (feature amounts) of the eyes, nose, and lips change and cannot be recognized. Or it is the problem of misrecognizing.

  Therefore, the present invention provides a face image feature extraction method, an apparatus thereof, and a feature image that can prevent erroneous recognition even if the face orientation is different from that at the time of registration, and the feature amount does not appear as a large difference. The purpose is to provide the program.

  According to a first aspect of the present invention, in a face image feature extraction method for extracting a face image feature quantity as a method, a position extraction step of extracting a position of a characteristic part from an input face image, and the position extraction A first calculation step of assigning position coordinates to the feature parts extracted in the step, and calculating an area of a first region obtained by joining the feature parts as vertices; and connecting the feature parts as vertices; A second calculation step of calculating an area of a second region that is a combined region and a region different from the first region, an area of the first region, and an area of the second region There is provided a face image feature extraction method characterized by comprising a step of calculating a ratio, and a face image feature value calculating step using the value obtained by the calculation as a face image feature value.

  According to a second aspect of the present invention, in the face image feature extraction apparatus that extracts a face image feature amount as an apparatus, a position extraction unit that extracts a position of a characteristic part from an input face image, and the position extraction A first calculating means for assigning position coordinates to the feature parts extracted by the means and calculating the area of the first region obtained by joining the feature parts as vertices; and connecting the feature parts as vertices; A second calculating means for calculating an area of a second region that is a combined region and a region different from the first region, an area of the first region, and an area of the second region There is provided a face image feature extraction device comprising: a face image feature amount calculating means for calculating a ratio and using a value obtained by the calculation as a face image feature amount.

  According to a third aspect of the present invention, in a face image feature extraction program for causing a computer to function as a face image feature extraction device that extracts a face image feature quantity as a program, a feature portion from an input face image is obtained. A position extraction unit for extracting a position, and a first calculation for calculating an area of a first region obtained by allocating position coordinates to a feature portion extracted by the position extraction unit and connecting the feature portion as a vertex. Means, a second calculation means for calculating an area of a second region which is a region connected with the characteristic portions as apexes and is different from the first region, and the first region A face image feature amount calculating means for calculating a ratio of the area of the second area and the area of the second region, and using the value obtained by the calculation as a face image feature amount. Facial image feature extraction program for causing a computer to function as a feature extraction apparatus is provided.

  According to the present invention, since the ratio of different regions of the face is taken, the amount of change due to the orientation of the face can be reduced, and it is possible to calculate a facial feature amount that does not depend on the orientation of the face.

  First, the outline of the present invention will be described.

  In the present invention, in order to solve the problem, from the image acquired from the camera, the ratio of the area of the region surrounded by the left and right eyes and the right and left nostrils and the area of the region surrounded by the two nostrils and both ends of the lips is calculated. Used as a feature value. By performing such processing, even if the orientation of the face is different from that at the time of registration, the feature amount is the ratio of the two areas and does not appear as a big difference. It becomes possible to prevent.

  Next, embodiments of the present invention will be described with reference to the drawings.

  An embodiment of the present invention is shown in the block diagram of FIG. The embodiment of the present invention includes a camera 11, a central processing unit 12, a face image storage device 13, an eye position extraction device 14, a nostril position extraction device 15, a lip position extraction device 16, a face upper area storage device 17, and a lower face area. A storage device 18 is included.

  The central processing unit 12 includes a RAM (Random Access Memory), a ROM (Read Only Memory), and a CPU. The face image storage device 13 stores a face image. The eye position extraction device 14 extracts the eye position from the input face image. The nostril position extraction device 15 extracts the position of the nostril. The lip end position extraction device 16 extracts the positions of both ends of the lips. The upper face area storage device 17 is a device for storing the area of a quadrangular region having both eyes and two nostrils as apexes. The lower face area storage device 18 is a device for storing the area of a quadrangular region having two nostrils and lip end points as apexes.

  The face image storage device 13, the face upper area storage device 17, and the face lower area storage device 18 all have information storage media such as a RAM (Random Access Memory) and a hard disk. The face image storage device 13, the upper face area storage device 17 and the lower face area storage device 18 are connected to the central processing device 12. Examples of the connection method include LAN (Local Area Network), RS232C serial communication, connection with a bus signal line, and the like.

  The storage capacity of the information storage medium of the face image storage device 13 is a capacity that can store an image obtained by the camera 11. Further, the storage capacity of the information storage media of the upper face area storage device 17 and the lower face area storage device 18 is a capacity that can store the area value calculated by the central processing unit 12.

  In order to input a face image to the apparatus, stand in front of the camera as shown in FIG. 2 and start image acquisition of the camera by an image acquisition start switch (a foot switch is shown as an example in the figure).

  Thereby, a face image as shown in FIG. 3 is acquired. The image acquisition start switch 19 may be a foot switch as shown in FIG. 2, or may be an arbitrary switch such as a switch by hand operation or a switch by a timer or a sensor.

[Description of Operation of Embodiment]
The operation of the embodiment will be described with reference to the block diagram of FIG. 1 and FIG. 4 which is a flowchart showing preprocessing.
When the image acquisition start switch 19 is pressed (step S101), the camera 11 is instructed to start face image acquisition, and face image data is transmitted from the image sensor of the camera 11 to the face image storage device 13. The face image data is stored in the face image storage device 13 (step S102).

  For the face image data acquired in the face image storage device 13, the eye position extraction device 14, the nostril position extraction device 15, and the lip end position extraction device 16 respectively use the object detection method such as the pattern matching method to detect the left eye. Center coordinates (coordinates (Ex1, Ey1) with the lower left corner of the face image as the origin (0, 0)), center coordinates of the right eye (coordinates (Ex2, Ey2) with the lower left corner of the face image as the origin (0, 0)) Is calculated (step S103).

  Also, the left nostril center coordinates (coordinates (Nx1, Ny1) with the lower left corner of the face image as the origin (0,0)) and the right nostril center coordinates (coordinates with the lower left corner of the face image as the origin (0,0)) ( Nx2, Ny2)) is calculated (step S104).

  Also, the left lip coordinate (coordinate (Mx1, My1) with the lower left of the face image as the origin (0,0)) and the right lip coordinate (coordinate with the lower left of the face image as the origin (0,0) ( Mx2, My2)) is calculated (step S105).

  Next, four points of right eye center coordinates (Ex1, Ey1), left eye center coordinates (Ex2, Ey2), right nostril center coordinates (Nx1, Ny1), and left nostril center coordinates (Nx2, Ny2) obtained by the above-described processing. The area A1 (upper face area) shown in FIG.

  The straight line L1 passing through the left and right eye center points (Ex1, Ey1), (Ex2, Ey2) and the straight line L2 passing through the left and right nostril center points (Nx1, Ny1), (Nx2, Ny2) are the faces of both eyes and both nostrils. As shown in the following mathematical formula 1, a square shape having four points (Ex1, Ey1), (Ex2, Ey2), (Nx1, Ny1), (Nx2, Ny2) as vertices is assumed. The area S1 can be calculated (step S106).

  These calculations are performed by the central processing unit 11 and the calculation results are stored in the upper face area storage device 17.

前述の処理で得られた右鼻孔中心座標（Ｎｘ１，Ｎｙ１）、左鼻孔中心座標（Ｎｘ２，Ｎｙ２）、右唇端座標（Ｍｘ１，Ｍｙ１）、左唇端座標（Ｍｘ２，Ｍｙ２）の４点で囲まれる図６に示した領域Ａ２（顔下部領域）を定義することができる。また、顔上部領域のときと同様に、左右鼻孔中心点（Ｎｘ１，Ｎｙ１），（Ｎｘ２，Ｎｙ２）を通る直線Ｌ２と、左右唇端座標（Ｍｘ１，Ｍｙ１），（Ｍｘ２，Ｍｙ２）を通る直線Ｌ３は、左右鼻孔と左右唇端の顔の位置関係から平行であると仮定し、以下に記載の数式２で（Ｎｘ１，Ｎｙ１），（Ｎｘ２，Ｎｙ２），（Ｍｘ１，Ｍｙ１），（Ｍｘ２，Ｍｙ２）の４点を頂点とする四角形の面積Ｓ２が算出できる。この計算は、中央処理装置１１が行い、計算結果を顔下部面積記憶装置１８に保存する（ステップＳ１０７）。

Four points of right nostril center coordinates (Nx1, Ny1), left nostril center coordinates (Nx2, Ny2), right lip end coordinates (Mx1, My1), and left lip end coordinates (Mx2, My2) obtained by the above-described processing. An area A2 (lower face area) shown in FIG. 6 can be defined. Similarly to the upper face region, a straight line L2 passing through the left and right nostril center points (Nx1, Ny1), (Nx2, Ny2) and a straight line passing through the left and right lip coordinates (Mx1, My1), (Mx2, My2) L3 is assumed to be parallel based on the positional relationship between the left and right nostrils and the left and right lip edges, and is expressed as (Nx1, Ny1), (Nx2, Ny2), (Mx1, My1), (Mx2, A square area S2 having apexes at four points of My2) can be calculated. This calculation is performed by the central processing unit 11, and the calculation result is stored in the lower face area storage device 18 (step S107).

最後にＣＰＵ１１は、上記で算出された顔上部面積Ｓ１と顔下部面積Ｓ２を顔上部面積記憶装置１７と顔下部面積記憶装置１８から取得し、以下に記載の数式３で、Ｓ１とＳ２の比Ｈを求め、顔認識のための特徴量として使用する（ステップＳ１０８）。顔の向きにより、顔上部面積と顔下部面積の変化は、同じような傾向で変化するため、顔の向きによるＨの変化は、小さい。このことから、Ｈは、顔の向きに影響されにくい特徴量である。

Finally, the CPU 11 obtains the upper face area S1 and the lower face area S2 calculated above from the upper face area storage device 17 and the lower face area storage device 18, and the ratio of S1 and S2 is expressed by the following Equation 3. H is obtained and used as a feature amount for face recognition (step S108). The change in the upper face area and the lower face area changes with the same tendency depending on the orientation of the face, so the change in H due to the orientation of the face is small. Therefore, H is a feature quantity that is not easily affected by the face orientation.

［発明の他の実施形態］
本発明の他の実施形態について説明する。本発明では、左右の目と左右鼻孔、左右鼻孔と左右唇端の面積比を特徴量として算出したが、両眉毛位置と左右の目の位置で構成される四角形の面積を用いてもよい。一般的な人の顔の左右に存在する眉毛、目、耳、鼻、唇などの位置座標を用いて領域を決定し、それぞれの領域の面積比を計算することによって、顔特徴量とすることも可能である。

[Other Embodiments of the Invention]
Another embodiment of the present invention will be described. In the present invention, the area ratio between the left and right eyes and the left and right nostrils, and the left and right nostrils and the left and right lips is calculated as the feature amount. However, a square area formed by the positions of both eyebrows and the left and right eyes may be used. Determine the area using the position coordinates of eyebrows, eyes, ears, nose, lips, etc. that exist on the left and right sides of a typical human face, and calculate the area ratio of each area to make it a facial feature value Is also possible.

  These eyebrows, eyes, ears, nose, lips and the like correspond to “characteristic portions” in the present invention.

  In the above-described embodiment, the ratio of the areas at two locations is calculated as the feature amount. However, the ratio of the areas at three or more locations may be calculated as the feature amount.

  In the present invention, the coordinate origin is set to the lower left of the face image acquired from the camera, but any point may be used as long as one point on the face image can be designated.

  The embodiment of the present invention described above has the following effects.

  The first effect is that the amount of change due to the orientation of the face can be reduced by taking the ratio of the different areas of the face, and the facial feature amount that does not depend on the orientation of the face can be calculated.

  The second effect is that the facial feature amount can be obtained with a small amount of calculation because only the area of the trapezoid is obtained.

  Note that the face image feature extraction apparatus according to the embodiment of the present invention can be realized by hardware, software, or a combination thereof.

It is a block diagram showing the basic composition of the embodiment of the present invention. It is the figure shown about the input of the face image in embodiment of this invention. It is the figure shown about the face image acquired in embodiment of this invention. It is the figure shown about operation | movement of embodiment of this invention. It is the figure shown about face upper area | region A1. It is the figure shown about the lower face area | region A2.

Explanation of symbols

DESCRIPTION OF SYMBOLS 11 Camera 12 Central processing unit 13 Face image storage device 14 Eye position extraction device 15 Nostril position extraction device 16 Lip position extraction device 17 Upper face area storage device 18 Lower face area storage device

Claims (15)

In the face image feature extraction method for extracting the face image feature amount,
A position extracting step for extracting the position of a characteristic part from the input face image;
A first calculation step of assigning position coordinates to the feature portion extracted in the position extraction step, and calculating an area of a first region obtained by connecting the feature portions as vertices;
A second calculation step of calculating an area of a second region which is a region connected with the characteristic parts as vertices and is different from the first region;
Calculating a ratio between the area of the first region and the area of the second region, and using the value obtained by the calculation as a face image feature amount;
A face image feature extraction method comprising: The face image feature extraction method according to claim 1,
The face image feature extraction method, wherein each of the first area and the second area is a quadrangular area having apexes of the four characteristic portions. In the face image feature extraction method according to claim 1 or 2,
2. The face image feature extraction method according to claim 1, wherein the feature portions are portions that exist in pairs on the left and right sides of the face. In the face image feature extraction method according to any one of claims 1 to 3,
A face image feature extraction method comprising comparing an input face image with a registered face image using the face image feature amount. The face image feature extraction method according to any one of claims 1 to 4,
The vertex of the first region is the center of the right eye, the center of the left eye, the center of the right nostril and the center of the left nostril, the vertex of the second region is the center of the right nostril, the center of the left nostril, A face image feature extraction method, characterized by being the right lip end and the left lip end. In the face image feature extraction device for extracting the face image feature amount,
Position extracting means for extracting the position of a characteristic part from the input face image;
First calculation means for assigning position coordinates to the feature portion extracted by the position extraction means, and calculating an area of a first region obtained by connecting the feature portions as vertices;
A second calculation means for calculating an area of a second region which is a region connected with the characteristic parts as vertices and is different from the first region;
A face image feature amount calculating means for calculating a ratio between the area of the first region and the area of the second region and using the value obtained by the calculation as a face image feature amount;
A face image feature extraction apparatus comprising: In the face image feature extraction device according to claim 6,
Each of the first area and the second area is a quadrilateral area having apexes of the four characteristic portions. In the face image feature extraction device according to claim 6 or 7,
The face image feature extraction apparatus, wherein the feature portions are portions that exist in pairs on the left and right sides of the face. In the face image feature extraction device according to any one of claims 6 to 8,
A face image feature extraction apparatus that compares an input face image with a registered face image by using the face image feature amount. In the face image feature extraction device according to any one of claims 6 to 9,
The vertex of the first region is the center of the right eye, the center of the left eye, the center of the right nostril and the center of the left nostril, the vertex of the second region is the center of the right nostril, the center of the left nostril, A face image feature extraction device characterized by the right lip end and the left lip end. In a face image feature extraction program for causing a computer to function as a face image feature extraction device that extracts a face image feature amount,
Position extracting means for extracting the position of a characteristic part from the input face image;
First calculation means for assigning position coordinates to the feature portion extracted by the position extraction means, and calculating an area of a first region obtained by connecting the feature portions as vertices;
A second calculation means for calculating an area of a second region which is a region connected with the characteristic parts as vertices and is different from the first region;
A face image feature amount calculating means for calculating a ratio between the area of the first region and the area of the second region and using the value obtained by the calculation as a face image feature amount;
A face image feature extraction program for causing a computer to function as a face image feature extraction device. In the face image feature extraction program according to claim 11,
A face image feature extraction program characterized in that each of the first region and the second region is a quadrangular region having apexes of the four characteristic portions. In the face image feature extraction program according to claim 11 or 12,
A face image feature extraction program characterized in that the feature portions are portions that exist in pairs on the left and right sides of the face. In the face image feature extraction program according to any one of claims 11 to 13,
A face image feature extraction program that compares an input face image with a registered face image by using the face image feature amount. In the face image feature extraction program according to any one of claims 11 to 14,
The vertex of the first region is the center of the right eye, the center of the left eye, the center of the right nostril and the center of the left nostril, the vertex of the second region is the center of the right nostril, the center of the left nostril, A face image feature extraction program characterized by a right lip end and a left lip end.

Images