JP2011054056A - Face position detector, face position detection method, and program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To detect the face of a driver who drives a vehicle. <P>SOLUTION: Pixel groups consisting of pixels arranged to be aligned with a contour curve determined in advance to meet the contour of the face are sequentially selected from among pixels constituting an image obtained by imaging the face of the driver. Likelihood of the selected pixel groups constituting the contour of the face of the driver is computed. The position of the face of the driver is detected, on condition that the pixels constituting the pixel groups determined based on the likelihood constitute the contour of the face. Since linear edges which do not constitute the contour of the face are not thereby extracted as the contour of the face, the contour of the face can be accurately detected. <P>COPYRIGHT: (C)2011,JPO&INPIT

Description

  The present invention relates to a face position detection device, a face position detection method, and a program, and more particularly, a face position detection device that detects the position of a face that appears in an image, and a face position detection method that detects the position of a face that appears in an image. And the program.

  There are various causes for inducing traffic accidents, but driving a vehicle in a vague state by a driver is one of the causes for inducing an accident. The ridiculous state means that the driver's attention is distracted due to the driver's actions other than driving, such as conversation while driving or the use of a mobile phone, or the driver's attention due to fatigue or sleepiness. A state of reduced attention. In view of this, a system has been proposed in which the state of the driver who drives the vehicle is observed at any time, and a warning is given to the driver when the driver reaches an ambiguous state.

  In this type of system, various methods are used for accurately recognizing a driver in the field of view such as a monitor camera (see, for example, Patent Documents 1 and 2).

JP 2005-011097 A JP 2007-72628 A

  The apparatus described in Patent Document 1 detects a plurality of straight lines that are candidates for a center line passing through the center of a face, based on a change in luminance of each pixel, from an image obtained by imaging a human face. . Then, the most probable straight line is selected from the candidate straight lines using the target of the pixel representing the contour of the face, and the presence or absence of the driver's face is determined based on the probability related to the selected straight line. Determine.

  However, in this apparatus, if a shadow is formed on the driver's face due to movement of the vehicle or the like, pixels matching the face outline may not be identified well. For this reason, it may be difficult to accurately determine the presence or absence of a face.

  Moreover, the apparatus described in Patent Document 2 performs image processing using a Sobel filter on an image obtained by imaging a predetermined area including a human face. Next, a histogram based on the edge values of the pixels constituting the processed image is generated. Then, the position corresponding to the peak of the histogram is detected as the face outline. However, the position where the peak of the histogram appears fluctuates due to the influence of the shadow formed on the driver's face. For this reason, in this apparatus, it may be difficult to accurately detect the position of the face.

  The present invention has been made under the above-described circumstances, and an object thereof is to accurately detect both ends of a face without being affected by shadows appearing on the face.

  In order to achieve the above object, a face position detection device according to the present invention has an imaging means for imaging a human face and a contour curve shape that follows the outline of the face among pixels constituting an image captured by the imaging means. A selection unit that sequentially selects a pixel group including arranged pixels, an edge direction detection unit that detects an orientation of an edge included in an image in each pixel for each pixel group selected by the selection unit, and a pixel group Computing means for calculating a likelihood that the pixels constituting the pixel group constitute the face contour based on the positional relationship between the pixels and the direction of the edge of each pixel.

  The calculation means may calculate the likelihood based on a distribution of points defined by the coordinates of each pixel and the direction of the edge.

  The computing means may calculate the likelihood based on the distribution of points with respect to a reference curve indicating the coordinates of each pixel constituting the pixel group and the direction of the normal line of the contour curve at the coordinates.

  Further, the calculation means may calculate the likelihood based on the number of points within a range set based on statistics.

  The face position detection device identifies a pixel group based on likelihood, and detects the position of the face from the image, assuming that the pixel of the identified pixel group is a pixel constituting at least a part of the contour of the face Means may be further provided.

  The pixel group is a pixel group composed of pixels arranged in a first contour curve shape along the right contour of the face and a pixel group composed of pixels arranged in a second contour curve shape along the left contour of the face. And the face position detection means extracts one of the pixel groups as a first pixel group based on the likelihood from a pixel group consisting of pixels arranged in a first contour curve shape, One pixel group is extracted as a second pixel group based on likelihood from the pixel group composed of pixels arranged in the second contour curve, and the first pixel group and the second pixel group are configured. The position of the face included in the image may be detected by assuming that the pixel is a pixel constituting at least a part of the contour of the face.

  The face position detecting means may extract a pixel group when the likelihood is the highest.

  In addition, the face position detection means makes the first pixel group and the second pixel group symmetrical with respect to the edge direction in the pixels constituting the first pixel group and the edge direction in the pixels constituting the second pixel group. It is good also as extracting based on.

  In addition, the face position detection unit determines the first pixel group and the second pixel group based on a relative positional relationship in the vertical direction of the face between the pixels constituting the first pixel group and the pixels constituting the second pixel group. It is good also as extracting.

  In order to achieve the above object, the face position detection method of the present invention sequentially selects a pixel group composed of pixels arranged in a contour curve along the face contour from among pixels constituting an image showing a human face. For each selected pixel group, detecting the direction of the edge included in the image in each pixel, and for each pixel group, the distribution of points defined by the coordinates of each pixel and the edge direction Based on the step of calculating the likelihood that the pixels constituting the pixel group constitute the contour of the face, identifying the pixel group based on the likelihood, and replacing the pixel of the identified pixel group with at least a part of the contour of the face And detecting a position of a face included in the image as a constituent pixel.

  In order to achieve the above object, the program of the present invention sequentially selects a pixel group consisting of pixels arranged in a contour curve along the outline of a face from among pixels constituting an image showing a human face. For each selected pixel group, a means for detecting the direction of the edge included in the image in each pixel, and for each pixel group, a distribution of points defined by the coordinates of each pixel and the edge direction. Based on the means for calculating the likelihood that the pixels constituting the pixel group constitute the face contour, the pixel group is identified based on the likelihood, and the pixel of the identified pixel group is determined as at least a part of the face contour. It is assumed that the pixel is a constituent pixel and functions as means for detecting the position of the face included in the image.

  According to the present invention, both ends of the face can be accurately detected.

It is a block diagram of the face position detection apparatus concerning a 1st embodiment. It is a figure which shows the image imaged by the imaging device. It is a figure for demonstrating a contour curve. It is a figure which shows a 1st template. It is a figure which shows a 2nd template. It is a figure for demonstrating operation | movement of a selection part. FIG. 3 is a diagram (part 1) illustrating points plotted in a y-θ coordinate system. It is a figure which shows the normal vector of the outline curve prescribed | regulated to the 1st template. It is a figure which shows the position of the outline curve on an image. FIG. 6 is a diagram (part 2) illustrating points plotted in the y-θ coordinate system. FIG. 10 is a diagram (part 3) illustrating points plotted in the y-θ coordinate system. It is a figure which shows the normal vector of the contour curve prescribed | regulated to the 2nd template. It is a block diagram of the face position detection apparatus which concerns on 2nd Embodiment. It is a flowchart for demonstrating operation | movement of an image processing apparatus.

<< First Embodiment >>
Hereinafter, a first embodiment of the present invention will be described with reference to FIGS. FIG. 1 is a block diagram showing a schematic configuration of a face position detection apparatus 10 according to the present embodiment. The face position detection device 10 is a device that detects a human face in an image. As illustrated in FIG. 1, the face position detection device 10 includes an imaging device 20 and an image processing device 30 that detects a human face from an image captured by the imaging device 20.

  The imaging device 20 is a device that converts an image acquired by imaging a subject into an electrical signal and outputs the electrical signal. For example, as can be seen by referring to the image IM shown in FIG. 2, the imaging device 20 is configured so that the face F of the driver 100 when seated in the driver's seat of the vehicle is imaged, for example, the windshield of the vehicle, Attached to the dashboard. Then, the driver 100 is imaged at a predetermined cycle, and information related to the image obtained by imaging is output to the image processing device 30.

  Returning to FIG. 1, the image processing apparatus 30 includes a selection unit 31, an edge angle detection unit 32, a calculation unit 33, and a face position detection unit 34.

  The selection unit 31 sequentially selects and outputs a pixel group composed of a plurality of pixels arranged along a preset contour curve from the image captured by the imaging device 20.

  FIG. 3 shows a contour curve L1 and a contour curve L2 along the contour of the face F of the driver 100. The contour curve L <b> 1 and the contour curve L <b> 2 are curves defined based on image data obtained by imaging a plurality of subjects seated in the driver's seat of the vehicle in advance by the imaging device 20. Each of the contour curve L1 and the contour curve L2 has a smoothly curved shape along the right (+ X side) and left (−X side) contours of the face F of the driver 100.

As shown in FIG. 4, the selection unit 31 includes a template T1 in which 40 pixels matching the contour curve L1 are colored, and 40 pixels matching the contour curve L2 as shown in FIG. Has a template T2 colored. Each of these templates T1 and T2 is composed of 400 pixels a (m, n) arranged in a matrix of 40 rows and 10 columns, and a template coordinate system (xy coordinate system) with the lower left corner as the origin is defined. Has been. Further, the template T1 and template T2, for convenience of explanation, among the pixels that match the contour curve L1 or profile curve L2, what Y coordinate smallest displayed as pixels a _1, every time the Y coordinate is increased by one pixel Are displayed as pixel a ₂ , pixel a ₃ ,..., Pixel a ₄₀ .

  As shown in FIG. 6 as an example, the selection unit 31 moves the template T1 in the X-axis direction or the Y-axis direction on the image IM, and the template T1 is colored from the pixels constituting the image IM. A pixel group consisting of 40 pixels (corresponding pixels) overlapping the pixels is sequentially selected. Then, information regarding the selected pixel group is sequentially output to the edge angle detection unit 32.

  Next, the selection unit 31 moves the template T2 on the image IM in the X-axis direction or the Y-axis direction, and selects 40 correspondences that overlap the colored pixels of the template T2 from the pixels that form the image IM. A pixel group consisting of pixels is sequentially selected. Then, information regarding the selected pixel group is sequentially output to the edge angle detection unit 32.

  The edge angle detection unit 32 detects the direction of the edge of the image IM in each of the corresponding pixels that form the pixel group output from the selection unit 31.

Specifically, the edge angle detection unit 32 uses the operator represented by the following expression (1), and the X axis of the image IM in each of the 40 corresponding pixels constituting the pixel group output from the selection unit 31. Magnitude gradient magnitudes AX ₁ to AX _{40 in the direction} are detected.

Next, the edge angle detection unit 32 uses the operator expressed by the following expression (2), and the Y-axis direction of the image IM in each of the 40 corresponding pixels that form the pixel group output from the selection unit 31. The brightness gradient magnitudes AY ₁ to AY ₄₀ are detected.

Next, the edge angle detection unit 32 sets the direction of the edge in each of the 40 corresponding pixels constituting the pixel group as an angle θ _N (N = 1, 2,..., 40) represented by the following expression (3). To detect. The edge angle detecting unit 32, the information about the detected angle theta _N, and outputs it to the arithmetic unit 33.

θ _N = tan ⁻¹ (AY _N / AX _N ) (3)

The computing unit 33 determines the point defined by the y coordinate in the template coordinate system of the corresponding pixel for the pixels a _{1 to} a ₄₀ of the template T1 and the angle θ _N as the edge direction of the corresponding pixel as the y-θ coordinate. Plot into the system.

As an example, FIG. 7 shows points plotted in the y-θ coordinate system together with the reference curve S1. FIG. 8 shows a normal vector N _N of the curve L1 starting from the pixels a _{1 to} a ₄₀ that overlap the curve L1 among the pixels a (m, n) constituting the template T1. The reference curve S1 in FIG. 7 is on the y-θ coordinate system defined by the y coordinate in the template coordinate system of the pixels a _{1 to} a ₄₀ of the template T1 and the angle formed between each normal vector N _N and the x axis. It is a curve that passes through the points. Based on the positional relationship between the calculation unit 33 and the points plotted in the y-θ coordinate system with respect to the reference curve S1, the pixel group selected from the image IM by the selection unit 31 is, for example, the driver shown in FIG. The likelihood that is the contour OP of 100 faces F is calculated.

As described above, the points plotted in the y-θ coordinate system are such that the contour curve L1 defined by the pixels a _{1 to} a ₄₀ of the template T1 is almost the contour OP of the face F as shown in FIG. when in matching position P ₂ to, as indicated by the white squares in FIG. 7, almost all the points are distributed along a reference curve S1.

Meanwhile, the point to be plotted on y-theta coordinate system, for example as shown in FIG. 9, the contour curve L1 is, when in the position P ₁ of the inner bit face F from the outline OP of the face F is in FIG. 10 As indicated by the colored squares, most of the points are distributed along the reference curve S1, but some points are distributed away from the reference curve S1. A point appearing in a region C1 spaced upward from the reference curve S1 is a point corresponding to a pixel near the eyebrows of the image IM. A point appearing in a region C2 spaced downward from the reference curve S1 is a point corresponding to the vicinity of the corner of the eye of the image IM.

Further, contour curve L1 is, when in the position P ₃ away from the remarkable face F from the outline OP of the face F, as indicated by white circles in FIG. 10, the position and the reference curve S1 is a point of substantially all Distributed independently.

  For example, as illustrated in FIGS. 7 and 10, the calculation unit 33 sets the area A1 including the reference curve S1 based on statistics and the like, and plots each pixel group selected from the image IM by the selection unit 31. From these points, the points included in the area A1 are detected. Then, the number of detected points is divided by the number of all points (40), and the result is sequentially calculated and output as likelihood.

Next, the calculation unit 33 determines a point defined by the y coordinate in the template coordinate system of the corresponding pixel for the pixels a _{1 to} a ₄₀ of the template T2 and the angle θ _N as the edge direction of the corresponding pixel, as y Plot in the -θ coordinate system.

As an example, FIG. 11 shows points plotted in the y-θ coordinate system together with the reference curve S2. FIG. 12 shows a normal vector N _N of the curve L2 starting from the pixels a _{1 to} a ₄₀ that overlap the curve L2 among the pixels a (m, n) constituting the template T2. The reference curve S2 in FIG. 11 is on the y-θ coordinate system defined by the y coordinate in the template coordinate system of the pixels a _{1 to} a ₄₀ of the template T2 and the angle formed between each normal vector N _N and the x axis. It is a curve that passes through the points. Based on the positional relationship between the calculation unit 33 and the points plotted in the y-θ coordinate system with respect to the reference curve S2, the pixel group selected from the image IM by the selection unit 31 is the driver 100 shown in FIG. The likelihood that is the contour OP of the face F is calculated. Specifically, for example, as illustrated in FIG. 11, the calculation unit 33 sets an area A2 including the reference curve S2 based on statistics, and plots each pixel group selected from the image IM by the selection unit 31. The points included in the area A2 are detected from the 40 points. Then, the number of detected points is divided by the number of all points (40), and the result is sequentially calculated and output as likelihood.

  The face position detection unit 34 specifies a pixel group when the likelihood having the largest value is calculated from the pixel group selected by the template T1. Next, the pixel group when the likelihood having the largest value is calculated from the pixel group selected by the template T2 is specified. Then, the position information of the face F of the driver 100 is detected and output on the assumption that the contour OP of the face F of the driver 100 is constituted by the specified set of pixels.

  As described above, in the first embodiment, the contour defined in advance along the contour OP of the face F from the pixels constituting the image IM obtained by imaging the face F of the driver 100. A pixel group composed of pixels arranged so as to coincide with the curves L1 and L2 is sequentially selected. Next, the likelihood that the selected pixel group constitutes the contour OP of the face F of the driver 100 is calculated. And the position in the image IM of the face F is detected as the pixel determined based on this likelihood comprises the outline OP of the face F. FIG.

  According to this, since straight edges that do not constitute the contour of the face F are not extracted as the contour of the face F, the right and left contours OP of the face F can be accurately detected. Specifically, when a shadow is formed on the face F, even if a straight edge appears in the image of the face F, the pixels constituting the image IM curve smoothly as determined in advance by statistics or the like. It is selected based on the contour curves L1 and L2, and it is determined whether or not the selected pixel constitutes the contour OP of the face F. Therefore, it is possible to accurately detect the face F of the driver 100 without being affected by a shadow or the like appearing on the face F.

<< Second Embodiment >>
Next, a second embodiment of the present invention will be described with reference to FIGS. In addition, about the structure same or equivalent to 1st Embodiment, while using an equivalent code | symbol, the description is abbreviate | omitted or simplified.

  The face position detection device 10 according to the present embodiment is the face position according to the first embodiment in that the image processing device 30 is realized by the same configuration as a general computer or a device such as a microcomputer. This is different from the detection device 10.

  FIG. 13 is a block diagram illustrating a physical configuration of the face position detection apparatus 10. As illustrated in FIG. 13, the face position detection device 10 includes an imaging device 20 and an image processing device 30 including a computer.

  The image processing apparatus 30 includes a central processing unit (CPU) 30a, a main storage unit 30b, an auxiliary storage unit 30c, a display unit 30d, an input unit 30e, an interface unit 30f, and a system bus 30g that interconnects the above units. It is configured.

  CPU30a performs the process mentioned later to the image acquired by the imaging device 20 according to the program memorize | stored in the auxiliary storage part 30c.

  The main storage unit 30b includes a RAM (Random Access Memory) and the like, and is used as a work area for the CPU 30a.

  The auxiliary storage unit 30c includes a nonvolatile memory such as a ROM (Read Only Memory), a magnetic disk, and a semiconductor memory. The auxiliary storage unit 30c stores programs executed by the CPU 30a, various parameters, and the like. In addition, information about an image output from the imaging device 20 and information including a processing result by the CPU 30a are sequentially stored.

  The display unit 30d includes a CRT (Cathode Ray Tube), an LCD (Liquid Crystal Display), or the like, and displays a processing result of the CPU 30a.

  The input unit 30e includes a pointing device such as a keyboard and a mouse. The operator's instruction is input via the input unit 30e and notified to the CPU 30a via the system bus 30g.

  The interface unit 30f includes a serial interface or a LAN (Local Area Network) interface. The imaging device 20 is connected to the system bus 30g via the interface unit 30f.

  The flowchart in FIG. 14 corresponds to a series of processing algorithms of a program executed by the CPU 30a. Hereinafter, the process executed by the image processing apparatus 30 will be described with reference to FIG. This process is executed after the face position detection device 10 is activated and information related to the image IM captured by the imaging device 20 is output.

  In the first step S101, as shown in FIG. 6 as an example, the CPU 30a moves the template T1 on the image IM in the X-axis direction or the Y-axis direction, and selects the template T1 from the pixels constituting the image IM. A pixel group composed of 40 corresponding pixels overlapping with the colored pixels is sequentially selected.

  In the next step S102, the CPU 30a moves the template T2 on the image IM in the X-axis direction or the Y-axis direction, and from among the pixels constituting the image IM, the 40 pixels that overlap the colored pixels of the template T2. A pixel group consisting of corresponding pixels is sequentially selected.

  In the next step S103, the CPU 30a detects the direction of the edge of the image IM in each of the pixels constituting the pixel group selected in steps S101 and S102.

Specifically, the CPU 30a uses the operator represented by the above formula (1), and the luminance gradient magnitudes AX ₁ to AX _{40 in} the X-axis direction of the image IM in each of the 40 pixels constituting the pixel group. Is detected. Next, using the operator represented by the above equation (2), the magnitudes of the luminance gradients AY ₁ to AY ₄₀ in the Y-axis direction of the image IM in each of the 40 pixels constituting the pixel group are detected. Then, the CPU 30a detects the edge direction in each of the 40 pixels constituting the pixel group as an angle θ _N (N = 1, 2,..., 40) expressed by the above equation (3).

In the next step S104, the CPU 30a determines a point defined by the y coordinate in the template coordinate system of the corresponding pixel for the pixels a _{1 to} a ₄₀ of the template T1 and the angle θ _N as the edge direction in the corresponding pixel, Plot in the y-θ coordinate system. Thereby, in the y-θ coordinate system, for example, as shown in FIG. 7, points indicated by white squares are plotted. Next, the CPU 30a determines a point defined by the y coordinate in the template coordinate system of the corresponding pixel for the pixels a _{1 to} a ₄₀ of the template T2 and the angle θ _N as the edge direction of the corresponding pixel as y−θ. Plot in the coordinate system.

  In the next step S105, the CPU 30a selects in steps S101 and S102 based on the positional relationship between the reference curves S1 and S2 shown in the y-θ coordinate system and the points plotted in the y-θ coordinate system. The likelihood that the obtained pixel group is, for example, the contour OP of the face F of the driver 100 shown in FIG. 2 is calculated.

As described above, the points plotted in the y-θ coordinate system are such that the contour line L1 defined by the pixels a _{1 to} a ₄₀ of the template T1 is almost the contour OP of the face F as shown in FIG. when in matching position P ₂ to, as indicated by the white squares in FIG. 7, almost all the points are distributed along a reference curve S1.

Meanwhile, the point to be plotted on y-theta coordinate system, for example as shown in FIG. 9, the contour line L1, when in the position P ₁ of the inner bit face F from the outline OP of the face F is in FIG. 10 As indicated by the colored squares, most of the points are distributed along the reference curve S1, but some points are distributed away from the reference curve S1. A point appearing in a region C1 spaced upward from the reference curve S1 is a point corresponding to a pixel near the eyebrows of the image IM. A point appearing in a region C2 spaced downward from the reference curve S1 is a point corresponding to the vicinity of the corner of the eye of the image IM.

Also, the contour line L1, when in the position P ₃ away from the remarkable face F from the outline OP of the face F, as indicated by white circles in FIG. 10, the position and the reference curve S1 is a point of substantially all Distributed independently.

  The CPU 30a sets a region A1 including the reference curve S1 as shown in FIGS. 7 and 10, and sets a region A2 including the reference curve S2 as shown in FIG. And the point contained in area | region A1 or area | region A2 is detected from 40 points plotted for every pixel group selected from the image IM. Then, the number of detected points is divided by the number of all points (40), and the result is sequentially calculated and output as likelihood.

  In the next step S106, the CPU 30a specifies the pixel group when the likelihood having the largest value is calculated from the pixel group selected by the template T1. Next, the CPU 30a specifies the pixel group when the likelihood with the largest value is calculated from the pixel group selected by the template T2.

  In the next step S107, the CPU 30a detects and outputs the position information of the face F, assuming that the contour OP of the face F of the driver 100 is constituted by the set of pixels specified in step S106.

  As described above, in the second embodiment, the contour defined in advance along the contour OP of the face F from the pixels constituting the image IM obtained by imaging the face F of the driver 100. A pixel group composed of pixels arranged so as to coincide with the curves L1 and L2 is sequentially selected. Next, the likelihood that the selected pixel group constitutes the contour OP of the face F of the driver 100 is calculated. And the position in the image IM of the face F is detected as the pixel determined based on this likelihood comprises the outline OP of the face F. FIG.

  According to this, since straight edges that do not constitute the contour of the face F are not extracted as the contour of the face F, the right and left contours OP of the face F can be accurately detected. Specifically, when a shadow is formed on the face F, even if a straight edge appears in the image of the face F, the pixels constituting the image IM curve smoothly as determined in advance by statistics or the like. It is selected based on the contour curves L1 and L2, and it is determined whether or not the selected pixel constitutes the contour OP of the face F. Therefore, it is possible to accurately detect the face F of the driver 100 without being affected by a shadow or the like appearing on the face F.

  As mentioned above, although embodiment of this invention was described, this invention is not limited by the said embodiment.

  For example, in each of the above embodiments, a template composed of 400 pixels arranged in a matrix of 40 rows and 10 columns is used. However, the present invention is not limited to this, and an appropriate size corresponding to the screen resolution and the screen size is used. A template may be used.

  In each of the above embodiments, the likelihood is calculated by dividing the number of points included in the region A1 or the region A2 provided in the y-θ coordinate system by the total number of points. However, the present invention is not limited to this. For example, the likelihood is calculated based on the ratio between the number of points included in the region A1 or the region A2 and the number of points not included in the region A1 or the region A2. It may be calculated. Further, a value obtained by subtracting the number of points not included in the region A1 or the region A2 from the number of points included in the region A1 or the region A2 may be calculated as the likelihood. Further, the likelihood described above may be determined based on the sum of errors at each point by calculating the distance between the reference curve S1 or the reference curve S2 and each point as an error.

  In the above embodiments, the pixel group constituting the right outline OP of the face F and the pixel group constituting the left outline OP are specified based on the likelihood, but the right and left outlines of the face F are specified. Each pixel group constituting the OP may be specified in consideration of the objectivity of the contour OP. For example, the symmetry of the angle of the edge, that is, the distribution of the points plotted in the y-θ coordinate system, between the pixel group constituting the right outline OP of the face F and the pixel group constituting the left outline OP. You may specify in consideration of the symmetry about ay axis. According to this, for example, a set of pixel groups that are significantly deviated in the vertical direction (Y-axis direction) is not specified as a pixel group that constitutes the left and right outlines OP, so that the detection accuracy of the face position is improved. It becomes possible.

  In addition, the first pixel group that forms the right contour of the face F and the second pixel group that forms the left contour OP are in a relative positional relationship in the y-axis direction between the pixels that form each pixel group. You may decide based on. Specifically, the first axis is set such that the distance in the y-axis direction between the pixels constituting the first pixel group and the pixels constituting the second pixel group corresponding to the pixel is equal to or less than a predetermined threshold. The pixel group and the second pixel group may be respectively determined. Also in this case, for example, a set of pixel groups that are significantly deviated in the vertical direction (Y-axis direction) is not specified as a pixel group that constitutes the left and right outlines OP, so that the detection accuracy of the face position is improved. It becomes possible.

  In each of the above embodiments, the pixel group is selected using the templates T1 and T2 based on the set of contour curves L1 and L2. However, a plurality of contour curves other than the contour curves L1 and L2 are set in advance. For example, when the calculated likelihood is equal to or less than a predetermined threshold, a pixel group may be selected using a template based on another contour curve. According to this, since the detection error based on the individual difference in the contour is reduced, it is possible to improve the detection accuracy of the face position.

  Further, in each of the above embodiments, as can be seen with reference to FIG. 6, the pixel groups are selected by superimposing the templates T <b> 1 and T <b> 2 on all the regions of the image IM. However, the present invention is not limited to this. For example, once the position of the driver's face is detected, the templates IM and T2 are matched with the image IM based on the detection result of the past face position. An area to be made may be defined. Thereby, since the amount of data handled when detecting the face position is reduced, the face position can be detected quickly.

  The functions of the image processing apparatus 30 according to each of the above embodiments can be realized by dedicated hardware or by a normal computer system.

  In the second embodiment, programs stored in the auxiliary storage unit 30c of the image processing apparatus 30 are a flexible disk, a CD-ROM (Compact Disk Read-Only Memory), a DVD (Digital Versatile Disk), an MO (Magneto). A device that executes the above-described processing may be configured by storing and distributing in a computer-readable recording medium such as -Optical disk) and installing the program in the computer.

  Further, the program may be stored in a disk device or the like included in a predetermined server device on a communication network such as the Internet, and may be downloaded onto a computer by being superimposed on a carrier wave, for example.

  Further, the program may be activated and executed while being transferred via a communication network.

  The program may be executed entirely or partially on the server device, and the above-described image processing may be executed while transmitting / receiving information regarding the processing via the communication network.

  Note that when the above functions are realized by sharing an OS (Operating System) or when the functions are realized by cooperation between the OS and an application, only the part other than the OS may be stored in a medium and distributed. It may also be downloaded to a computer.

  It should be noted that the present invention can be variously modified and modified without departing from the broad spirit and scope of the present invention. Further, the above-described embodiment is for explaining the present invention, and does not limit the scope of the present invention.

  The face position detection device, the face position detection method, and the program according to the present invention are suitable for detecting the position of the driver's face.

DESCRIPTION OF SYMBOLS 10 Face position detection apparatus 20 Imaging apparatus 30 Image processing apparatus 30a CPU
30b Main storage unit 30c Auxiliary storage unit 30d Display unit 30e Input unit 30f Interface unit 30g System bus 31 Selection unit 32 Edge angle detection unit 33 Calculation unit 34 Face position detection unit 100 Driver F Face OP Contour IM image L1, L2 Contour curve S1, S2 reference curve T1, T2 template

Claims (11)

Imaging means for imaging a human face;
A selection unit that sequentially selects a pixel group including the pixels arranged in a contour curve along the outline of the face among the pixels constituting the image captured by the imaging unit;
Edge direction detection means for detecting the direction of the edge included in the image at the pixel for each pixel group selected by the selection means;
Calculation means for calculating a likelihood that the pixels constituting the pixel group constitute the contour of the face based on the positional relationship between the pixels and the direction of the edge of each pixel for each pixel group When,
A face position detecting device.   The face position detection apparatus according to claim 1, wherein the calculation unit calculates the likelihood based on a distribution of points defined by coordinates of the pixels and an orientation of the edge.   The calculation means calculates the likelihood based on a distribution of the points with respect to a reference curve indicating the coordinates of each of the pixels constituting the pixel group and a direction of a normal line of the contour curve at the coordinates. Item 3. The face position detection device according to Item 2.   The face position detection apparatus according to claim 2, wherein the calculation unit calculates the likelihood based on the number of the points within a range set based on statistics.   The pixel group is identified based on the likelihood, and the position of the face included in the image is detected assuming that the pixel of the identified pixel group is the pixel constituting at least part of the contour of the face The face position detection device according to claim 1, further comprising a face position detection unit that performs the operation. The pixel group is:
The pixel group consisting of the pixels arranged in a first contour curve shape along the right contour of the face;
The pixel group consisting of the pixels arranged in a second contour curve along the left contour of the face;
Including
The face position detecting means includes
Extracting one of the pixel groups as the first pixel group based on the likelihood from the pixel group consisting of the pixels arranged in the first contour curve shape,
From the pixel group consisting of the pixels arranged in the second contour curve shape, any one of the pixel groups is extracted as a second pixel group based on the likelihood,
6. The position of the face included in the image is detected by assuming that the pixels constituting the first pixel group and the second pixel group are the pixels constituting at least a part of the outline of the face. 2. A face position detecting device.   The face position detection device according to claim 5, wherein the face position detection unit extracts the pixel group when the likelihood is the highest.   The face position detection means includes the first pixel group and the second pixel group, the edge direction in the pixels constituting the first pixel group, and the edge direction in the pixels constituting the second pixel group. The face position detection device according to claim 6, wherein the face position detection device extracts the images based on the symmetry with respect to.   The face position detection unit relates the first pixel group and the second pixel group to the vertical direction of the face between the pixels constituting the first pixel group and the pixels constituting the second pixel group. The face position detection apparatus according to claim 6, wherein extraction is performed based on a relative positional relationship. A step of sequentially selecting a pixel group composed of the pixels arranged in a contour curve along the outline of the face, among pixels constituting an image of a human face;
Detecting a direction of an edge included in the image in the pixel for each of the selected pixel groups;
For each pixel group, the likelihood that the pixels constituting the pixel group constitute the contour of the face is calculated based on the distribution of points defined by the coordinates of the pixels and the direction of the edge. Process,
The pixel group is identified based on the likelihood, and the position of the face included in the image is detected assuming that the pixel of the identified pixel group is the pixel constituting at least part of the contour of the face And a process of
A face position detection method including: Computer
Means for sequentially selecting a pixel group composed of the pixels arranged in a contour curve along the outline of the face, among pixels constituting an image showing a human face;
Means for detecting an orientation of an edge included in the image in the pixel for each of the selected pixel groups;
For each pixel group, the likelihood that the pixels constituting the pixel group constitute the contour of the face is calculated based on the distribution of points defined by the coordinates of the pixels and the direction of the edge. Means,
The pixel group is identified based on the likelihood, and the position of the face included in the image is detected assuming that the pixel of the identified pixel group is the pixel constituting at least part of the contour of the face Means to
Program to function as.

Images