JP5018653B2 - Image identification device - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To remove influence of disturbance light with high accuracy to identify a portion of an identification target of a face or an identification object with high accuracy. <P>SOLUTION: A face image in lighting time of a first light source 14 and a face image in lighting time of a second light source 16 are imaged by an imaging device 12. A characteristic amount related to a difference of each of a hair area and a skin area is calculated by a characteristic amount calculation part 27, and a correction amount is calculated by a correction amount calculation part 28 such that a distribution of the previously obtained characteristic amount related to each of the hair area and the skin area is acquired. The face image in lighting time of the second light source 16 is corrected by a correction part 30, and the characteristic amount related to the difference is calculated about each pixel by use of the corrected face image by the characteristic amount calculation part 27. Threshold value processing is performed to the calculated characteristic amount of each pixel by a candidate area extraction part 34 to extract a candidate area, and an area representing a black eye of the face is identified by a portion identification part 36. <P>COPYRIGHT: (C)2010,JPO&amp;INPIT

Description

  The present invention relates to an image identification device, and more particularly, to an image identification device that identifies a part of an identification target person's face or an identification target object from a captured image.

Conventionally, an object photographed in a state in which irradiation of light of the first wavelength band and light of the second wavelength band is turned on and off, the object and the standard white plate are imaged, and illuminated with the light of the first wavelength band Pixel value of an image of an object, pixel value of an image of a white plate photographed with light in the first wavelength range, pixel value of an image of an object photographed with light in the second wavelength range And the pixel value of the image of the white plate photographed in the state illuminated with the light of the second wavelength range, the influence of disturbance light is removed using the calculation result, and each part of the object, for example, the skin of the driver A region discriminating apparatus that discriminates hair and eyeball is known (Patent Document 1).
JP 2006-242909 A

  However, the technique of Patent Document 1 has a problem in that the influence of disturbance light cannot be removed with high accuracy because the standard white plate and the object are not always irradiated with disturbance light. In addition, when the installed standard white plate is hidden by an object other than the target, there is a problem that the influence of disturbance light is erroneously removed without being detected.

  The present invention has been made to solve the above-described problems, and is an image identification device capable of accurately removing the influence of disturbance light and accurately identifying a part to be identified as a face and an identification object. The purpose is to provide.

  In order to achieve the above object, an image identification device according to a first invention has sensitivity in a wavelength region including a first wavelength and a second wavelength different from the first wavelength, and the first wavelength And irradiating the face of the person to be identified with light in the first wavelength range that does not contain the second wavelength, or the first light from the reflected light reflected from the face of the person to be identified The first wavelength is output by imaging the face of the person to be identified by transmitting light in the wavelength range of the second wavelength, and the second wavelength includes the second wavelength and does not include the first wavelength. Irradiating the face of the person to be identified with the light of the region or imaging the face of the person to be identified by transmitting the light of the second wavelength region from the reflected light Imaging means for outputting image data; and the identification target person in the first image data And calculating a feature amount relating to a difference between the image data of the region representing the predetermined first part and the image data of the region representing the first part in the second image data, and the first image data First feature amount calculation for calculating a feature amount related to a difference between image data of a region representing a second portion different from the first portion of the face and image data of a region representing the second portion in the second image data. And a feature amount related to the difference of the region representing the first part and a feature amount related to the difference of the region representing the second part calculated by the first feature amount calculating means, The image data is corrected so that the feature quantity related to the difference of the area representing the first part and the feature quantity related to the difference of the area representing the second part are obtained. Correction amount calculating means for calculating a correction amount for correcting the image data, and correcting at least one of the first image data and the second image data based on the correction amount calculated by the correction amount calculating means A second feature amount calculating unit that calculates a feature amount relating to a difference between the first image data and the second image data for each pixel, and each pixel calculated by the second feature amount calculating unit. Of the first image data and the second image data are compared with the feature amount related to the difference obtained in advance for the image data of the region representing the part to be identified of the face. And detecting means for detecting the part to be identified from at least one side.

  According to the image identification device of the first invention, the imaging device having sensitivity in a wavelength range including the first wavelength and the second wavelength different from the first wavelength includes the first wavelength and the second wavelength. Irradiating the face of the identification target person with light in the first wavelength range that does not include the wavelength of the identification target, or transmitting the light in the first wavelength range from the reflected light reflected from the face of the identification target person First image data is output by imaging the face of the person to be identified. Further, the imaging means irradiates the face of the person to be identified with light in the second wavelength range including the second wavelength and not including the first wavelength, or images the reflected light from the second wavelength range. The second image data is output by imaging the face of the person to be identified by transmitting light.

  Image data of a region representing a predetermined first part of the face of the person to be identified in the first image data and image data of a region representing the first part in the second image data by the first feature amount calculating means Between the image data of the area representing the second part different from the first part of the face in the first image data and the image data of the area representing the second part in the second image data. A feature amount related to the difference is calculated.

  Then, the correction amount calculation means is obtained in advance based on the feature amount relating to the difference between the regions representing the first part and the feature amount relating to the difference between the areas representing the second part calculated by the first feature amount calculating means. A correction amount for correcting the image data is calculated so that a feature amount relating to the difference between the regions representing the first part and a feature amount relating to the difference between the regions representing the second part are obtained.

  The first image when the second feature amount calculation unit corrects at least one of the first image data and the second image data based on the correction amount calculated by the correction amount calculation unit. A feature amount relating to the difference between the data and the second image data is calculated for each pixel. The detection unit compares the feature amount related to the difference between the pixels calculated by the second feature amount calculation unit with the feature amount related to the difference obtained in advance for the image data of the region representing the part to be identified of the face, A part to be identified is detected from at least one of the first image data and the second image data.

  As described above, the correction amount is calculated based on the feature amount related to the difference between the regions representing the first part of the face and the feature amount related to the difference between the regions representing the second part, and the corrected image data is used to calculate the face. By detecting the part to be identified, it is possible to accurately remove the influence of disturbance light and accurately identify the part to be identified.

  The image identification device according to the first invention updates a feature amount relating to a predetermined difference based on a feature amount relating to a difference in image data of an area representing a part of the face of the person to be identified detected by the detecting means. Update means may further be included. Thereby, the part of the face of the person to be identified can be accurately detected in accordance with the characteristics of the individual of the person to be detected and the characteristics of the light source.

  The 1st site | part which concerns on 1st invention can be made into a hair part, and a 2nd site | part can be made into a skin part.

  Each of said 1st wavelength range and said 2nd wavelength range can be made into the range of 800-1100 nm. The first wavelength range can be 800 to 950 nm, and the second wavelength range can be 950 to 1100 nm.

  As described above, according to the image identification device of the present invention, it is possible to accurately remove the influence of disturbance light from the image data and accurately identify the part or identification object of the face identification target. An effect is obtained.

  Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings. In the present embodiment, a case where the present invention is applied to an image identification device that is mounted on a vehicle and identifies a black eye region of a driver's face will be described as an example.

  As shown in FIG. 1, an image identification device 10 according to the first exemplary embodiment includes an imaging device 12 provided to image a range including a driver's face as a person to be identified, and a first wavelength range. A first light source 14 provided to irradiate a range including a driver's face with near infrared light (for example, a wavelength range of 800 nm to 950 nm including 870 nm) and a second wavelength range (for example, including 970 nm) A second light source 16 provided to irradiate a range including a driver's face with near infrared light in a wavelength range of 950 nm to 1100 nm, on / off of irradiation by the first light source 14 and the second light source 16, and an imaging device 12 and a computer 18 that performs image processing on the captured face image and identifies the black eye region of the driver's face.

  The imaging device 12 images a range including the face of the driver who is the identification target, generates an image signal of the image, and A / D converts the image signal generated by the imaging unit. An A / D converter (not shown) and an image memory (not shown) for temporarily storing the A / D converted image signal are provided. In addition, the imaging element of the imaging unit of the imaging device 12 has sensitivity in a wavelength range including the first wavelength range and the second wavelength range.

  The first wavelength range of the light irradiated by the first light source 14 and the second wavelength range of the light irradiated by the second light source 16 have different main wavelength ranges, and the second wavelength range includes the second wavelength range. If the main wavelength 970 nm is not included and the second wavelength region does not include the main wavelength 870 nm of the first wavelength region, a part of the wavelength region may overlap.

  The computer 18 includes a CPU, a ROM that stores an identification program for an identification processing routine, which will be described later, a RAM that stores data, and a bus that connects these. When the computer 18 is described in terms of functional blocks divided for each function realizing means determined based on hardware and software, as shown in FIG. 1, the imaging device 12 continuously captures two face images. The imaging light source control unit 20 that switches each of the first light source 14 and the second light source 16 on and off in accordance with the imaging timing (shutter timing) of the imaging device 12 and the lighting of the first light source 14 output from the imaging device 12 Based on the difference value of each pixel of the two face images input by the image input unit 22 and the image input unit 22 that inputs the face image that is the gray image at the time and the face image when the second light source 16 is turned on. A hair skin region extraction unit 26 for extracting each of a region representing hair and a region representing skin from the face image, and a region representing the extracted hair in the two face images, and Based on the image data of the area representing the area, the feature amount related to the difference between the area representing the hair and the area representing the skin is calculated, and two faces are used using the face image corrected by the correction unit 30 described later. A feature amount calculation unit 27 that calculates a feature amount related to a difference from the image for each pixel, and a correction amount for correcting the face image based on the extracted feature amounts of the region representing hair and the region representing skin. A correction amount calculation unit 28 to be calculated and a correction unit 30 to correct the face image input by the image input unit 22 according to the calculated correction amount are provided.

  The imaging light source control unit 20 controls the imaging device 12 to continuously capture two face images, and in accordance with the imaging timing of the imaging device 12, the second light source 16 is captured at the time of imaging one face image. The first light source 14 and the second light source 16 are turned on so that the first light source 14 is turned on and the second light source 16 is turned on when the other face image is captured. Control. Thereby, the face image photographed when the first light source 14 is turned on and the face image photographed when the second light source 16 is turned on are sequentially output from the imaging device 12.

  The image input unit 22 includes, for example, an A / D converter and an image memory that stores image data of one screen.

  With skin and hair, spectral reflection characteristics as shown in FIG. 2 are obtained. In FIG. 2, the horizontal axis indicates the wavelength of light, and the vertical axis indicates the reflectance. In the near-infrared wavelength range of 800 nm to 1000 nm, the reflectance of hair has a characteristic of rising right (the longer the wavelength, the higher the reflectance), and the reflectance of skin is decreasing to the right (the longer the wavelength, the higher the reflectance). Is low). Further, when comparing the reflectance of near infrared light at two wavelengths in the near infrared region, for example, 870 nm and 970 nm, the skin strongly reflects near infrared light of 870 nm, and hair and black eyes are 970 nm. It has the characteristic of strongly reflecting near infrared light. Therefore, when an image captured by irradiating 970 nm near-infrared light is subtracted from an image captured by irradiating 870 nm near-infrared light (calculating the difference between the corresponding pixels), pixels in the region representing skin Is a positive value, and the difference value of the pixels in the region representing the hair and black eyes is a negative value.

  Therefore, as shown in the following equation (1), the hair skin region extraction unit 26 determines the target pixel based on the difference value of each pixel of the two face images, and the positive threshold value and the negative threshold value regarding the difference value. Identify hair and at the same time identify whether it represents skin and extract a region representing hair and a region representing skin.

Here, f ₈₇₀ (i, j) is the pixel value of the pixel (i, j) in the face image captured when the first light source 14 is irradiated with light in the first wavelength range including the wavelength of 870 nm. F ₉₇₀ (i, j) represents the pixel value of the pixel (i, j) in the face image captured when the second light source 16 is irradiated with light in the second wavelength range including a wavelength of 970 nm. ing. Th ₁ is a positive threshold value, and Th ₂ is a negative threshold value. As described above, since the difference value of the pixel value differs greatly between the skin region and the hair region, the brightness balance of the two wavelength regions of the first light source 14 and the second light source 16 in the face image due to the influence of disturbance light. Even in a state in which the state has collapsed, it is possible to search for a threshold value for separating the two relatively easily. Therefore, it is sufficient to examine the positive threshold value and the negative threshold value for separating the skin region and the hair region experimentally or statistically and set the threshold values in advance.

The feature amount calculation unit 27 includes a pixel value of each pixel in the area representing the hair of the two face images, a pixel value of each pixel in the area representing the skin of the two face images, and each pixel of the two face images. On the basis of the pixel value, the following equation (2) is calculated for each pixel to calculate a feature value f _n (i, j) relating to the difference between the pixels (i, j).

  Depending on the distance from the light source to the imaging target, the brightness of the target in the captured image changes. Therefore, in the above equation (2), the difference between the pixel values of the target pixel is normalized by the sum of the pixel values of the target pixel in the two captured images having different wavelength ranges of the light irradiated on the object, and depends on the distance. The feature quantity that is not affected by the brightness change is calculated. Moreover, since it has normalized, the feature-value represented by said Formula (2) becomes a value within the range of -1 to 1.

  Further, as disclosed in Patent Document 1, skin, hair, and black eyes (pupil, iris) have different spectral reflection characteristics in the near-infrared region. Therefore, as shown in FIG. Different distributions of feature quantities are obtained for regions representing. Note that the distribution of feature amounts in the black eye region is close to the distribution of feature amounts in the hair region, unlike the distribution of feature amounts in the skin region and the white eye region.

  In addition, the intensity balance of light in two wavelength regions (a wavelength region including 870 nm and a wavelength region including 970 nm) is affected by the amount of light in the two wavelength regions included in the disturbance light. Therefore, when strong disturbance light such as sunlight is present, the value of the feature amount calculated by the above equation (2) changes, and as shown in FIGS. 4 (A) and 4 (B) Since the distribution of the feature quantity of the area representing the hair and the hair is biased, it becomes difficult to separate the area representing the skin and the hair based on the distribution of the feature quantity. In addition, as shown in FIG. 3 above, the distribution of the feature amount of the region representing the pupil and the iris is positioned between the distribution of the feature amount of the region representing the skin and the distribution of the feature amount of the region representing the hair. Separation of areas representing skin and hair based on the distribution of feature amounts becomes even more difficult, making it impossible to identify with high accuracy.

  Therefore, the correction amount calculation unit 28 has no influence of disturbance light as shown in FIG. 4C, in which the distribution of the feature amounts of the extracted region representing hair and the region representing skin is determined in advance. As described below, the correction amount for correcting the image is calculated so that the distribution of the feature amount having a large difference in the feature amount axis in the case is obtained.

  First, in order to correct the intensity balance between the first wavelength region including 870 nm and the second wavelength region including 970 nm in the face image, the correction amount for correcting the face image when the second light source 16 is turned on is α. The feature amount related to the difference when corrected by the amount α is expressed by the following equation (3).

  By varying the correction amount α in various ways, the distribution of the feature amounts of the skin region and the hair region calculated by the above equation (3) is the characteristic amount of each of the skin region and the hair region stored in advance. A correction amount α when the distribution is close is obtained, and a correction amount for correcting the face image when the second light source 16 is turned on is determined.

  Note that the difference between the distribution of feature amounts calculated from the skin region and the hair region is calculated so that the relationship between the distribution of the feature amount of the skin region and the distribution of the feature amount of the hair region is large. The amount of correction to be increased may be determined.

  The correction unit 30 corrects the pixel value of each pixel of the face image when the second light source 16 is turned on according to the correction amount calculated by the correction amount calculation unit 28. As a result, the distribution of the feature amount related to the difference calculated from the face image when the first light source 16 is turned on and the corrected face image when the second light source 16 is turned on is as shown in FIG. Therefore, it becomes easy to separate the distribution of the feature amount of the region representing the pupil or the iris located in the middle of the distribution of the feature amount of the region representing each of the skin and the hair, and the black eye region can be identified.

  Further, as shown in FIG. 1, the computer 18 is calculated using a threshold value storage unit 32 that stores a threshold value that defines a feature amount range based on the feature amount distribution of the black eye region, and a corrected face image. Based on the feature amount of each pixel related to the difference and the threshold value related to the feature value stored in the threshold value storage unit 32, a candidate region extraction unit 34 that extracts a candidate region of a black eye region from the face image, and the extracted candidate region A part identifying unit 36 for identifying a region representing a black eye as a part to be identified, and an updating unit 38 for updating the threshold value stored in the threshold value storage unit 32 based on the distribution of the feature amount of the identified black eye region. I have.

  The threshold value stored in the threshold value storage unit 32 is determined as follows. First, from the face image from which the influence of disturbance light has been removed or from the face image without the influence of disturbance light, the feature amount distribution of the black eye region is obtained, and based on the obtained feature amount distribution of the black eye region, The range of the feature amount of the region is determined, a threshold value that defines the determined feature amount range is determined, and stored in the threshold value storage unit 32.

  Based on the feature amount of each pixel relating to the difference calculated using the corrected face image, the candidate region extraction unit 34 has the feature amount within the range determined by the threshold from the inside of the extracted skin region. The region is extracted as a candidate region for the black eye region. At this time, since the distribution of the feature amount of the black eye region is close to the distribution of the feature amount of the hair region, the region representing the eyebrow is also extracted as a candidate region for the black eye region.

  The part identifying unit 36 excludes the upper candidate area from the positional relationship of the extracted candidate area of the black eye area in the face image, and expresses the black eye from the candidate area of the black eye area. A region is detected to identify a region representing a black eye.

  The reflection characteristics of near-infrared light in the black eye region have individual differences, and the reflectivity varies depending on the region (pupil and iris) in the black eye. Therefore, as shown in FIGS. 5A and 5B, the size of the pupil Depending on the brightness of the surrounding environment and the brightness of the light source, the value and distribution of the feature amount regarding the difference change. Further, as shown in FIGS. 6A and 6B, the feature quantity regarding the difference is also affected by the irradiation intensity of the light source (LED, etc.) and the imaging conditions (shutter speed and gain control). The value and distribution of the feature value change according to individual differences and surrounding brightness.

  Therefore, in the update unit 38, the threshold value stored in the threshold value storage unit 32 is obtained from the distribution of the feature amount considering the size of the driver's pupil and the intensity of light emitted from the first light source 14 and the second light source 16. The threshold value stored in the threshold value storage unit 32 is updated in accordance with the distribution of the feature amount of the identified black eye region so that the threshold value is set to the threshold value.

  Next, the operation of the image identification device 10 according to the first embodiment will be described. Below, the case where the face of the driver who is a candidate for identification is located in the imaging range of imaging device 12, the irradiation range of the 1st light source 14, and the irradiation range of the 2nd light source 16 is explained as an example.

  In the computer 18, an identification processing routine shown in FIG. 7 is executed. First, in step 100, the first light source 14 and the second light source 16 are controlled to be turned on and off, and when the first light source 14 is turned on and only the second light source 16 is turned on, the imaging device 12 causes the driver to Control is performed so that the region including the face is continuously imaged.

  In step 102, a face image when the first light source 14 is turned on and a face image when the second light source 16 is turned on are acquired from the imaging device 12, and in step 104, 2 acquired in step 102 above. Based on the pixel values of one face image, a difference value of each pixel is calculated.

  Then, in step 106, based on the difference value of each pixel calculated in step 104, a region representing hair is extracted according to the above equation (1). In step 108, each pixel calculated in step 104 is extracted. Based on the difference value, an area representing the skin is extracted according to the above equation (1).

  In the next step 109, for the region representing the hair extracted in step 106, the feature amount of each pixel is calculated according to the above equation (2) based on the pixel values of the two face images acquired in step 102. In the same manner, the feature amount of each pixel is calculated for the region representing the skin extracted in step 108.

  In step 110, the distribution of the feature amount of the region representing the hair calculated in step 109, the distribution of the feature amount of the region representing the skin, and each of the predetermined region representing the hair and the region representing the skin. Based on the distribution of the feature amount, a correction amount for correcting the face image when the second light source 16 is turned on is calculated.

  In step 112, the pixel value of each pixel of the face image obtained when the second light source 16 is turned on acquired in step 102 is corrected according to the correction amount calculated in step 110.

  In the next step 114, the pixel value of the face image when the first light source 14 is turned on acquired in step 102 and the pixel value of the face image when the second light source 16 is corrected corrected in step 112 are used. Based on the above equation (2), the feature quantity related to the difference is calculated for each pixel.

  Then, in step 116, the range in which the feature amount is defined by the threshold stored in the threshold storage unit 32 from the inside of the skin region extracted in step 108 based on the feature amount calculated in step 114. The inner area is extracted as a candidate area for the black eye area. In the next step 118, a region representing a black eye is identified from the candidate regions of the black eye region extracted in step 116 based on the positional relationship of the candidate regions of the black eye region in the face image.

  In step 120, the threshold value stored in the threshold value storage unit 32 is updated based on the feature amount distribution of the region representing the black eye identified in step 118, and the identification processing routine is terminated.

  As described above, according to the image identification device according to the first embodiment, the distribution of the feature amount related to the difference calculated for the region representing the facial hair and the feature amount related to the difference calculated for the region representing the skin. By calculating the correction amount for correcting the face image so that the distribution of the predetermined feature amount is the distribution of the predetermined feature amount, and detecting the black eye region of the face using the corrected face image, disturbance light Thus, the black eye region of the face can be accurately identified.

  In addition, it is possible to accurately detect a region representing the black eye of the driver's face corresponding to the individual difference in the reflection characteristics of the driver's black eyes and the individual difference in the irradiation intensity of the light source.

  In addition, since the reflection characteristics of the skin and hair included in the face area are known, an accurate correction amount can be calculated from the brightness of the skin and hair area in the face image, and the sunlight irradiated to the driver It can correct | amend so that the influence of disturbance light, such as.

  In the above embodiment, the case where a region representing a black eye is detected from the candidate region of the black eye region based on the positional relationship between the candidate regions has been described as an example. However, the present invention is not limited to this. A region representing a black eye may be detected by a pattern recognition method based on the distribution of the feature amount in the candidate region. In a face image as shown in FIGS. 8A and 8B, a feature amount distribution as shown in FIG. 8C is obtained in a region representing a black eye. Therefore, a pattern using this feature amount distribution is used. An area representing a black eye may be detected from the candidate area by the recognition method.

  Moreover, although the case where only the face image when the second light source is turned on according to the correction amount has been described as an example, the present invention is not limited to this, and the face image when the first light source is turned on and the second light source is turned on. You may make it correct | amend both with the face image of time. In this case, the face when the first light source is turned on when the distribution of the feature amounts of the extracted hair region and the skin region is close to the predetermined feature amount distribution. A correction amount for correcting the image and a correction amount for correcting the face image when the second light source is turned on may be obtained. Alternatively, only the face image when the first light source is turned on may be corrected.

  Further, the case where the skin region and the hair region are extracted based on the difference value between the two face images has been described as an example, but the present invention is not limited to this, and each of the skin region and the hair region in the face image is described. These positions may be set in advance. Further, a human head region may be searched for face images from edge information or the like, and a skin region and a hair region may be extracted using the search result.

  Further, in order to calculate the correction amount, the same calculation formula expressed by the above equation (2) is used when calculating the feature amount regarding the difference and when calculating the feature amount regarding the difference from the corrected face image. However, the present invention is not limited to this. However, the present invention is not limited to this, and the case where the feature amount related to the difference is calculated to calculate the correction amount and the corrected face image are used. The feature amount related to the difference may be calculated by using different calculation formulas in the case of calculating the feature amount related to the difference.

  Next, a second embodiment will be described. In addition, about the part which becomes the structure similar to 1st Embodiment, the same code | symbol is attached | subjected and description is abbreviate | omitted.

  The second embodiment is different from the first embodiment in that the correction amount is calculated based on the face image when the light source is turned off.

  As illustrated in FIG. 9, the computer 218 of the image identification device 210 according to the second embodiment causes the imaging device 12 to continuously capture three face images, and the imaging timing (shutter timing) of the imaging device 12. ), The imaging light source control unit 220 that switches each of the first light source 14 and the second light source 16 on and off, the face image when the first light source 14 that is output from the imaging device 12 is turned on, and the second light source 16 The correction amount of the face image is calculated based on the image input unit 222 that inputs the face image when the light source is turned on and the face image when both the first light source 14 and the second light source 16 are turned off, and the face image when the light source is turned off. And a correction amount calculating unit 228 that performs correction for correcting the face image when the first light source 14 is turned on and the face image when the second light source 16 is turned on, which are input by the image input unit 222, according to the calculated correction amount. 230, a feature amount calculation unit 227 that calculates a feature amount related to a difference for each pixel using the face image corrected by the correction unit 230, a threshold storage unit 32, a candidate region extraction unit 34, and a part identification unit 36 and an update unit 38.

  The imaging light source control unit 220 controls the imaging device 12 to continuously capture three face images, and at the time of capturing the first face image in accordance with the imaging timing of the imaging device 12, With the light source 16 turned off, the first light source 14 is turned on, and when the second face image is picked up, the first light source 14 is turned off and the second light source 16 is turned on to pick up the third face image. Sometimes, the first light source 14 and the second light source 16 are controlled so that the first light source 14 and the second light source 16 are turned off. As a result, a face image taken when the first light source 14 is turned on, a face image taken when the second light source 16 is turned on, and a face image taken when the light source is turned off are sequentially output from the imaging device 12. Is done.

  Since the value of each pixel of the face image when the light source is turned off is estimated to be a disturbance light component, the correction amount calculation unit 228 calculates the pixel value of each pixel of the face image captured when the light source is turned off as the correction amount. To do.

  In accordance with the calculated correction amount of each pixel, the correction unit 230 responds from the pixel value of each pixel of the face image taken when the first light source 14 is turned on and the face image taken when the second light source 16 is turned on. Correction is performed by subtracting the correction amount to be applied.

The feature amount calculation unit 227 applies the pixel value to each pixel (i, j) based on the corrected face image when the first light source 14 is turned on and the corrected face image when the second light source 16 is turned on. On the other hand, the feature quantity f _n (i, j) regarding the difference is calculated according to the above equation (2).

  Next, the contents of the identification processing routine according to the second embodiment will be described with reference to FIG. Note that the same processes as those of the first embodiment are denoted by the same reference numerals, and description thereof is omitted.

  First, in step 250, the first light source 14 and the second light source 16 are controlled to be turned on / off, and only the first light source 14 is turned on, only the second light source 16 is turned on, and the first light source 14 and the second light source 14 are turned on. When both the light source 16 and the light source 16 are turned off, the imaging device 12 performs control so that an area including the driver's face is continuously imaged. In the next step 252, a face image when the first light source 14 is turned on, a face image when the second light source 16 is turned on, and a face image when the light source is turned off are acquired from the imaging device 12.

  In step 254, a correction amount for correcting the face image is calculated for each pixel based on the face image when the light source is extinguished acquired in step 252. In step 256, the correction amount acquired in step 252 is calculated. For each of the face image when the first light source 14 is turned on and the face image when the second light source 16 is turned on, the pixel value of each pixel is corrected according to the correction amount calculated in step 254.

  In the next step 114, based on the pixel value of the face image when the first light source 14 is lit corrected in step 256 and the pixel value of the face image when the second light source 16 is lit, the above equation (2) In accordance with the above, a feature amount regarding the difference is calculated for each pixel.

  In step 116, a candidate area for the black eye region is extracted based on the feature amount calculated in step 114. In the next step 118, the candidate area for the black eye region extracted in step 116 is extracted. An area representing a black eye is identified.

  In step 120, the threshold value stored in the threshold value storage unit 32 is updated based on the feature amount distribution of the region representing the black eye identified in step 118, and the identification processing routine is terminated.

  As described above, according to the image identification device according to the second embodiment, the correction amount is calculated based on the face image captured when the light source is turned off, and the face is corrected using the corrected face image. By detecting the area representing the black eye, it is possible to accurately remove the influence of disturbance light and accurately identify the area representing the black eye.

  In addition, since it is possible to estimate how much disturbance light is included from the brightness of the face image when the light source is turned off, it is possible to calculate an appropriate image correction amount for black eye identification, and Correction can be made so as to remove the influence of the disturbance light being irradiated.

  In addition, although the case where both the face image when the first light source is lit and the face image when the second light source is lit is corrected according to the correction amount is described as an example, the present invention is not limited to this. Only the face image when lit or only the face image when the second light source is lit may be corrected. In this case, a correction amount for correcting only the face image when the first light source is turned on or a correction amount for correcting only the face image when the second light source is turned on based on the face image when the light source is turned off. Should be requested.

  Next, a third embodiment will be described. In addition, about the part which becomes the structure similar to 1st Embodiment, the same code | symbol is attached | subjected and description is abbreviate | omitted.

  The third embodiment is different from the first embodiment in that the correction amount for correcting the face image is calculated based on the illuminance measured by the illuminometer.

  As illustrated in FIG. 11, an image identification device 310 according to the third embodiment includes an imaging device 12, a first light source 14, a second light source 16, and an illuminance of light incident on an imaging device of the imaging device 12. Illuminance meter 312 provided to measure illuminance, and on / off of irradiation by first light source 14 and second light source 16 and imaging by imaging device 12, and using the illuminance measured by illuminance meter 312, And a computer 318 that performs image processing on the captured face image and identifies a region representing the black eye of the driver's face.

  The computer 318 also includes a correction amount calculation unit 328 that calculates the correction amount of the face image based on the illuminance measured by the imaging light source control unit 20, the image input unit 22, and the illuminance meter 312, and the calculated correction. A correction unit 330 that corrects each of the two face images input by the image input unit 222 according to the amount, a feature amount calculation unit 227, a threshold storage unit 32, a candidate area extraction unit 34, and a part identification unit 36 And an update unit 38.

  Here, the principle of the present embodiment will be described. The spectral intensity of disturbance light such as sunlight and road ambient lighting can be known in advance. Further, the spectral sensitivity of the image sensor of the imaging device and the spectral transmittance of the lens can be known in advance. Therefore, if the spectral intensity, spectral sensitivity, and spectral transmittance described above are examined in advance and the illuminance of light incident on the imaging device is measured, the ratio of disturbance light included in the pixel value of the captured image can be estimated. Is possible.

  Therefore, in this embodiment, the correction amount calculation unit 328 measures the spectral intensity of disturbance light, the spectral sensitivity of the imaging device of the imaging device 12, and the spectral transmittance of the lens, and the illuminometer 312. Based on the illuminance of the measured light, the ratio of disturbance light included in the pixel value of the captured image is calculated as the amount of the measured light received by the imaging device 12, and the calculated ratio is used to correct the face image. Correction amount.

  The correction unit 330 corrects the pixel value of each pixel of the face image shot when the first light source 14 is turned on and the face image shot when the second light source 16 is turned on according to the calculated correction amount.

  Next, an identification processing routine according to the third embodiment will be described with reference to FIG. Note that the same processes as those of the first embodiment are denoted by the same reference numerals, and description thereof is omitted.

  First, in step 100, the first light source 14 and the second light source 16 are controlled to be turned on and off, and when the first light source 14 is turned on and only the second light source 16 is turned on, the imaging device 12 causes the driver to Control is performed so that the region including the face is continuously imaged.

  In step 102, a face image when the first light source 14 is turned on and a face image when the second light source 16 is turned on are acquired from the imaging device 12, and measured by the illuminometer 312 in the next step 350. Get the illuminance of light.

  In step 352, based on the illuminance of the light acquired in step 350, the spectral intensity of disturbance light, the spectral sensitivity of the imaging device of the imaging device 12, and the spectral transmittance of the lens obtained in advance. A correction amount for correcting the face image is calculated, and in step 354, for each of the face image when the first light source 14 is turned on and the face image when the second light source 16 is turned on, acquired in step 102, The pixel value of each pixel is corrected according to the correction amount calculated in step 352.

  In the next step 114, based on the pixel value of the face image when the first light source 14 is lit corrected in step 354 and the pixel value of the face image when the second light source 16 is lit, the above equation (2) In accordance with the above, a feature amount regarding the difference is calculated for each pixel.

  In step 116, a candidate area for the black eye region is extracted based on the feature amount calculated in step 114. In the next step 118, the candidate area for the black eye region extracted in step 116 is extracted. An area representing a black eye is identified.

  In step 120, the threshold value stored in the threshold value storage unit 32 is updated based on the feature amount distribution of the region representing the black eye identified in step 118, and the identification processing routine is terminated.

  As described above, according to the image identification device according to the third embodiment, the correction amount for correcting the face image is calculated and corrected based on the illuminance of the light measured by the illuminometer. By detecting the area representing the black eyes of the face using the face image, the area representing the black eyes of the face can be accurately identified by accurately removing the influence of ambient light.

  In addition, since the spectrum distribution of disturbance light is examined in advance and the illuminance of the light is measured with an illuminometer, it can be estimated how much disturbance light is included. The correction amount can be calculated, and correction can be made so as to remove the influence of disturbance light applied to the driver.

  In addition, although the case where both the face image when the first light source is lit and the face image when the second light source is lit is corrected according to the correction amount is described as an example, the present invention is not limited to this. Only the face image when lit or only the face image when the second light source is lit may be corrected. In this case, the ratio of disturbance light included in the pixel value of the captured image is calculated based on the spectral intensity of disturbance light obtained in advance and the illuminance of the measured light, and based on the ratio of disturbance light The correction amount for correcting only the face image when the first light source is turned on or the correction amount for correcting only the face image when the second light source is turned on may be obtained.

  Next, a fourth embodiment will be described. In addition, the same code | symbol is attached | subjected about the part which becomes the same structure as 1st Embodiment, and description is abbreviate | omitted.

  In the fourth embodiment, a near-infrared light in each of the first wavelength range and the second wavelength range is transmitted from the reflected light from the driver's face using a bandpass filter to capture a face image. This is different from the first embodiment.

  As shown in FIG. 13, the image identification device 410 according to the fourth embodiment includes a first wavelength region (e.g., 800 nm to 870 nm including 870 nm) among the light reflected from the imaging device 12 and the driver's face. A band-pass filter 414 that switches near-infrared light in a wavelength range of 950 nm and near-infrared light in a second wavelength range (for example, a wavelength range of 950 nm to 1100 nm including 970 nm) and transmits it to the imaging device 12; A computer 418 that controls switching of the wavelength range transmitted by the band-pass filter 414 and imaging by the imaging device 12, and performs image processing on the captured face image to identify a region representing the black eyes of the driver's face; It has.

  The band-pass filter 414 includes a filter that transmits near-infrared light in the first wavelength range, a filter that transmits near-infrared light in the second wavelength range, and a mechanism that switches between the two filters. According to the control, the wavelength range of the light that is transmitted through the imaging device 12 among the light reflected from the driver's face is switched to either the first wavelength range or the second wavelength range.

  The computer 418 continuously captures two face images by the imaging device 12 and sets the wavelength range of light transmitted by the band-pass filter 414 in accordance with the imaging timing (shutter timing) of the imaging device 12 for the first time. An imaging filter control unit 420 that switches to either the wavelength range or the second wavelength range, and a face image and near-infrared light in the second wavelength range that are output from the imaging device 12 when transmitting near-infrared light in the first wavelength range An image input unit 422 for inputting a face image at the time of transmission, a hair skin region extraction unit 26, a feature amount calculation unit 27, a correction amount calculation unit 28, a correction unit 30, a threshold storage unit 32, and candidate region extraction Unit 34, part identifying unit 36, and updating unit 38.

  In the identification processing routine according to the fourth embodiment, the wavelength range of the light transmitted by the bandpass filter 414 is controlled to be switched from one of the first wavelength range and the second wavelength range to the other, and the first wavelength When the near-infrared light in the region is transmitted and the near-infrared light in the second wavelength region is transmitted, the imaging device 12 performs control so that the region including the driver's face is continuously imaged.

  And the face image at the time of near-infrared light transmission of the 1st wavelength range and the face image at the time of near-infrared light transmission of the 2nd wavelength range are acquired from the imaging device 12, and the pixel of two acquired face images Based on the value, the difference value of each pixel is calculated. Next, based on the calculated difference value of each pixel, similarly to the first embodiment, a region representing hair is extracted and a region representing skin is extracted according to the above equation (1).

  Then, as in the first embodiment, the feature amount of each pixel relating to the difference is calculated according to the above equation (2) for the region representing the extracted hair, and the same for the region representing the extracted skin. Then, the feature amount of each pixel relating to the difference is calculated. Next, based on the distribution of the feature amount of the area representing the calculated hair, the distribution of the feature amount of the area representing the skin, and the distribution of the feature amount of each of the area representing the hair and the area representing the skin. Thus, the correction amount for correcting the face image when transmitting near-infrared light in the second wavelength region is calculated by the same method as in the first embodiment. And the pixel value of each pixel of the face image at the time of near-infrared light transmission of the acquired 2nd wavelength range is correct | amended according to the calculated correction amount.

  Next, based on the acquired pixel value of the face image when transmitting near-infrared light in the first wavelength range and the corrected pixel value of the face image when transmitting near-infrared light in the second wavelength range, In accordance with the above equation (2), a feature amount related to the difference is calculated for each pixel. Then, based on the calculated feature value, a region where the feature value is within the range defined by the threshold value stored in the threshold value storage unit 32 from the inside of the skin region extracted above is selected as a candidate for the black eye region. Extract as a region. Next, an area representing a black eye is identified from the extracted candidate areas for the black eye area based on the positional relationship of the candidate areas for the black eye area in the face image.

  Then, the threshold value stored in the threshold value storage unit 32 is updated based on the feature amount distribution of the area representing the identified black eye, and the identification processing routine is terminated.

  In this way, the correction amount for correcting the face image is calculated using the face image when transmitting near-infrared light in the first wavelength region and the face image when transmitting near-infrared light in the second wavelength region. By detecting the black eye area of the face using the corrected face image, it is possible to accurately remove the influence of disturbance light and accurately identify the black eye area of the face.

  Next, a fifth embodiment will be described. In addition, about the part which becomes the structure similar to 1st Embodiment-4th Embodiment, the same code | symbol is attached | subjected and description is abbreviate | omitted.

  In the fifth embodiment, a near-infrared light in each of the first wavelength region and the second wavelength region is transmitted from the reflected light from the driver's face using a bandpass filter to capture a face image. This is different from the third embodiment.

  As illustrated in FIG. 14, the image identification device 510 according to the fifth embodiment is configured to switch and pick up an image with the imaging device 12, the bandpass filter 414, the illuminance meter 312, and the wavelength band transmitted by the bandpass filter 414. A computer 518 that controls imaging by the device 12 and performs image processing on the captured face image using the illuminance measured by the illuminometer 312 to identify a region representing the black eyes of the driver's face; It has.

  The computer 518 includes an imaging filter control unit 420, an image input unit 422, a correction amount calculation unit 328, a correction unit 330, a feature amount calculation unit 227, a threshold value storage unit 32, a candidate area extraction unit 34, a part An identification unit 36 and an update unit 38 are provided.

  In the correction amount calculation unit 328, the spectral intensity of disturbance light, the spectral sensitivity of the imaging device of the imaging device 12, the spectral transmittance of the lens, the spectral transmittance of each filter of the bandpass filter 414, and the illuminometer Based on the illuminance of the light measured by 312, the ratio of disturbance light included in the pixel value of the captured image is used as a correction amount for correcting the face image for each of the first wavelength region and the second wavelength region. ,calculate.

  In the identification processing routine according to the fifth embodiment, the wavelength range of the light transmitted by the bandpass filter 414 is controlled to be switched from one of the first wavelength range and the second wavelength range to the other, and the first wavelength When the near-infrared light in the region is transmitted and the near-infrared light in the second wavelength region is transmitted, the imaging device 12 performs control so that the region including the driver's face is continuously imaged.

  Then, a face image when transmitting near-infrared light in the first wavelength region and a face image when transmitting near-infrared light in the second wavelength region are acquired from the imaging device 12, and then measured by the illuminometer 350. Get the illuminance of the light.

  Then, the illuminance of the acquired light, the spectral intensity of disturbance light, the spectral sensitivity of the imaging device of the imaging device 12, the spectral transmittance of the lens, and the spectral transmittance of each filter of the bandpass filter 414, which are obtained in advance, Based on the above, a correction amount for correcting the face image is calculated for each of the face image when transmitting near infrared light in the first wavelength range and the face image when transmitting near infrared light in the second wavelength range, For each of the acquired face image when transmitting near-infrared light in the first wavelength region and face image when transmitting near-infrared light in the second wavelength region, the pixel value of each pixel is corrected according to the corresponding correction amount.

  Next, based on the corrected pixel value of the face image when transmitting near-infrared light in the first wavelength range and the pixel value of the face image when transmitting near-infrared light in the second wavelength range, the above (2) In accordance with the equation, a feature amount relating to the difference is calculated for each pixel. Then, based on the calculated feature amount, a candidate area for the black eye area is extracted, and an area representing the black eye is identified from the extracted candidate areas for the black eye area.

  Then, the threshold value stored in the threshold value storage unit 32 is updated based on the feature amount distribution of the area representing the identified black eye, and the identification processing routine is terminated.

  In this way, the correction amount for correcting the face image is calculated using the face image when transmitting near-infrared light in the first wavelength region and the face image when transmitting near-infrared light in the second wavelength region. By detecting the black eye area of the face using the corrected face image, it is possible to accurately remove the influence of disturbance light and accurately identify the black eye area of the face.

  Next, a sixth embodiment will be described. In addition, about the part which becomes the structure similar to 1st Embodiment-4th Embodiment, the same code | symbol is attached | subjected and description is abbreviate | omitted.

  In the sixth embodiment, a near-infrared light in each of the first wavelength region and the second wavelength region is transmitted from reflected light from the driver's face using a bandpass filter, and a face image is captured. This is different from the second embodiment.

  As shown in FIG. 15, the image identification device 610 according to the sixth embodiment operates an imaging device 12, a bandpass filter 414, and light in a wide wavelength range including the first wavelength range and the second wavelength range. The light source 614 that irradiates the person's face, the switching of the wavelength range transmitted by the band-pass filter 414, the imaging by the imaging device 12, and the on / off of the light source 614 are controlled, and image processing is performed on the captured face image And a computer 618 for identifying a region representing the black eyes of the driver's face.

  The computer 618 continuously captures four face images by the imaging device 12 and sets the wavelength range of light transmitted by the band-pass filter 414 in accordance with the imaging timing (shutter timing) of the imaging device 12 for the first time. The imaging light source filter control unit 620 that switches between the wavelength range and the second wavelength range and switches the light source 614 on and off, and the light source 614 that is output from the imaging device 12 and is in the first wavelength range The face image when transmitting near-infrared light, when the light source 614 is turned on, and when the face image when transmitting near-infrared light in the second wavelength range, when the light source 614 is turned off, and the first An image input unit 622 for inputting a face image when transmitting near-infrared light in the wavelength range and a face image when light source 614 is turned off and transmitting near-infrared light in the second wavelength range, and turning off the light source Time and first A correction amount calculation unit that calculates a correction amount of the face image based on each of the face image when transmitting near-infrared light in the wavelength region and the face image when the light source is turned off and transmitting near-infrared light in the second wavelength region. 628 and the face image when the light source is turned on and near-infrared light is transmitted in the first wavelength range and the near red color in the second wavelength range when the light source is turned on and input by the image input unit 622 according to the calculated correction amount A correction unit 630 that corrects each face image when transmitting external light, a feature amount calculation unit 227, a threshold storage unit 32, a candidate area extraction unit 34, a part identification unit 36, and an update unit 38 are provided. Yes.

  The imaging light source filter control unit 620 controls the imaging device 12 to continuously capture four face images. In addition, the imaging light source filter control unit 620 turns on the light source 614 and adjusts the wavelength range of the light transmitted by the bandpass filter 414 at the time of imaging the first face image in accordance with the imaging timing of the imaging device 12. Switching to the first wavelength range, the second wavelength range is switched to the wavelength range of the light transmitted by the bandpass filter 414 while the light source 614 is turned on when the second face image is captured. Further, the imaging light source filter control unit 620 turns off the light source 614 and switches the wavelength range of light transmitted by the band pass filter 414 to the first wavelength range when the third face image is captured. When the face image is captured, the wavelength range of light transmitted by the bandpass filter 414 is switched to the second wavelength range while the light source 614 is turned off.

  The correction amount calculation unit 628 calculates the pixel value of each pixel of the face image captured when the light source 614 is turned off and transmitted near-infrared light in the first wavelength region as the correction amount for the first wavelength region. In addition, the pixel value of each pixel of the face image captured when the light source 614 is turned off and the near-infrared light in the second wavelength region is transmitted is calculated as a correction amount for the second wavelength region.

  In accordance with the calculated correction amount of each pixel with respect to the first wavelength range, the correction unit 630 is configured for each pixel of the face image that is captured when the light source 614 is turned on and near-infrared light is transmitted in the first wavelength range. Correction is performed by subtracting the corresponding correction amount from the pixel value. Further, the correction unit 630 determines each face image captured when the light source 614 is turned on and near-infrared light is transmitted in the second wavelength range according to the calculated correction amount of each pixel for the second wavelength range. Correction is performed by subtracting the corresponding correction amount from the pixel value of the pixel.

  In the identification processing routine according to the sixth embodiment, the wavelength range of the light transmitted by the band pass filter 414 is controlled to be switched from one of the first wavelength range and the second wavelength range to the other, and the light source 614 is turned on. Turn on and control the imaging device 12 to continuously capture an area including the driver's face when transmitting near-infrared light in the first wavelength range and transmitting near-infrared light in the second wavelength range. . In addition, the wavelength range of the light transmitted by the bandpass filter 414 is controlled so as to be switched from one of the first wavelength range and the second wavelength range to the other, and the light source 614 is turned off so that the near red in the first wavelength range is displayed. When the outside light is transmitted and the near-infrared light in the second wavelength range is transmitted, the imaging device 12 performs control so that the region including the driver's face is continuously imaged.

  Then, from the image pickup device 12, a face image when the light source is turned on and near-infrared light transmission in the first wavelength range, and a near-infrared light in the second wavelength range when the light source is turned on. A face image when transmitting, a face image when the light source is turned off and near-infrared light in the first wavelength range is transmitted, and a face image when the light source is turned off and near-infrared light in the second wavelength range The face image at the time of transmission is acquired.

  Then, based on the obtained two face images when the light source is turned off, a correction amount for correcting the face image when transmitting near-infrared light in the first wavelength region is calculated for each pixel, and the second wavelength is calculated. A correction amount for correcting the face image when transmitting near-infrared light in the region is calculated for each pixel.

  Then, when the acquired light source is turned on and the face image when the near-infrared light is transmitted in the first wavelength range, and when the light source is turned on and the near-wavelength light is transmitted in the second wavelength range For each of the face images, the pixel value of each pixel is corrected in accordance with the calculated corresponding correction amount. Next, the pixel value of the face image when the corrected light source is turned on and near-infrared light in the first wavelength range is transmitted, and the near-red color in the second wavelength range when the light source is turned on. Based on the pixel value of the face image at the time of transmission of external light, a feature amount related to the difference is calculated for each pixel.

  Then, based on the calculated feature amount, a candidate area for the black eye area is extracted, and an area representing the black eye is identified from the extracted candidate areas for the black eye area. Next, the threshold value stored in the threshold value storage unit 32 is updated based on the feature amount distribution of the area representing the identified black eye, and the identification processing routine is terminated.

  Thus, the face image when the light source is turned off and near-infrared light is transmitted in the first wavelength range, and the face image when the light source is turned off and near-infrared light is transmitted in the second wavelength range Is used to calculate the correction amount for correcting the face image, and the corrected face image is used to detect the black eye area of the face, thereby accurately removing the influence of ambient light and The black eye area can be accurately identified.

  In the fourth to sixth embodiments, the band image is used to transmit the face image when transmitting near-infrared light in the first wavelength region and the near-infrared light in the second wavelength region. The case of capturing a face image during transmission has been described as an example. However, the present invention is not limited to this, and a prism that can separate light in the first wavelength range and light in the second wavelength range is used. A face image when transmitting near-infrared light in the first wavelength region and a face image when transmitting near-infrared light in the second wavelength region may be captured. In addition, after the reflected light from the driver's face is spectrally separated by the prism, the near-infrared light in the first wavelength range is transmitted by the bandpass filter to capture the face image, and similarly the reflected light is spectrally separated by the prism. After that, the near-infrared light in the second wavelength region may be transmitted by the band pass filter to capture the face image.

  In addition, band pass filters that transmit different wavelength ranges may be provided for each pixel or pixel column. In this case, a band-pass filter for transmitting near-infrared light in the first wavelength region and a band-pass filter for transmitting near-infrared light in the second wavelength region are installed in units of pixels or pixel columns. Then, a face image when transmitting near-infrared light in the first wavelength region and a face image when transmitting near-infrared light in the second wavelength region may be acquired from the captured image data.

  Further, a light source that irradiates the driver's face with light in a wide wavelength range including the first wavelength range and the second wavelength range may be further provided so that the light source is turned on when the face image is captured.

  Moreover, in said 1st Embodiment-6th Embodiment, although the 1st wavelength range and the 2nd wavelength range demonstrated as an example the case where it was contained in the wavelength range of near-infrared light, However, the present invention is not limited to this, and the first wavelength region and the second wavelength region may include a visible light wavelength region.

  Moreover, although the case where the feature amount related to the difference is calculated from the two face images corrected based on the calculated correction amount has been described as an example, the present invention is not limited to this, and the face image is not corrected. Based on the two face images before correction and the calculated correction amount, a feature amount related to the difference obtained when the two face images are corrected based on the correction amount may be calculated.

  Further, the case where a grayscale image is output from the imaging device has been described as an example, but the present invention is not limited to this, and a color image may be output from the imaging device.

It is a block diagram which shows the image identification device which concerns on the 1st Embodiment of this invention. It is a graph which shows the relationship between the wavelength range and the reflectance of the light about hair and skin. It is an image figure which shows distribution of the feature-value of each site | part of a face. (A) A graph showing a distribution of feature values biased by the influence of disturbance light, (B) a graph showing a distribution of feature quantities biased by the influence of disturbance light, and (C) a case where there is no influence of disturbance light It is a graph which shows distribution of feature-value. (A) Image diagram showing distribution of feature quantity when pupil size is small, and (B) Image diagram showing distribution of feature quantity when pupil size is large. (A) Image diagram showing distribution of feature quantity when light source irradiation intensity is low, and (B) Image diagram showing distribution of feature quantity when light source irradiation intensity is high. It is a flowchart which shows the content of the identification process routine in the image identification device which concerns on the 1st Embodiment of this invention. (A) An image diagram showing a face image, (B) An image diagram showing an enlarged image of the face image, and (B) An image diagram showing a distribution of feature quantities in a region representing a black eye. It is a block diagram which shows the image identification device which concerns on the 2nd Embodiment of this invention. It is a flowchart which shows the content of the identification process routine in the image identification device which concerns on the 2nd Embodiment of this invention. It is a block diagram which shows the image identification device which concerns on the 3rd Embodiment of this invention. It is a flowchart which shows the content of the identification process routine in the image identification device which concerns on the 3rd Embodiment of this invention. It is a block diagram which shows the image identification device which concerns on the 4th Embodiment of this invention. It is a block diagram which shows the image identification device which concerns on the 5th Embodiment of this invention. It is a block diagram which shows the image identification device which concerns on the 6th Embodiment of this invention.

Explanation of symbols

10, 210, 310, 410, 510, 610 Image identification device 12 Imaging device 14 First light source 16 Second light source 18, 218, 318, 418, 518, 618 Computer 20, 220 Imaging light source control unit 26 Hair skin region extraction unit 27, 227 Feature amount calculation unit 28, 228, 328, 628 Correction amount calculation unit 30, 230, 330, 630 Correction unit 32 Threshold storage unit 34 Candidate region extraction unit 36 Region identification unit 38 Update unit 312 Illuminometer 414 Band pass filter 420 Imaging Filter Control Unit 620 Imaging Light Source Filter Control Unit

Claims (5)

Light in a first wavelength range having sensitivity in a wavelength range including a first wavelength and a second wavelength different from the first wavelength, including the first wavelength and not including the second wavelength. Or imaging the face of the identification target person by transmitting light in the first wavelength range from the reflected light reflected from the face of the identification target person. And outputting the first image data and irradiating the face of the identification subject with the light of the second wavelength range including the second wavelength and not including the first wavelength, or Imaging means for outputting second image data by imaging the face of the person to be identified by transmitting light in the second wavelength range from the reflected light;
Features relating to a difference between image data of a region representing a predetermined first part of the face of the person to be identified in the first image data and image data of a region representing the first part in the second image data Calculating the amount, and image data of a region representing a second part different from the first part of the face in the first image data, and image data of a region representing the second part in the second image data; First feature value calculating means for calculating a feature value related to the difference between the two,
Based on the feature amount related to the difference of the region representing the first part and the feature amount related to the difference of the region representing the second part calculated by the first feature amount calculating means, the first feature is obtained in advance. Correction amount calculation means for calculating a correction amount for correcting image data so that a feature amount related to the difference of the region representing the part and a feature amount related to the difference of the region representing the second part are obtained;
The first image data and the second image data when at least one of the first image data and the second image data is corrected based on the correction amount calculated by the correction amount calculation means. Second feature value calculating means for calculating a feature value related to a difference from the image data for each pixel;
Comparing the feature amount relating to the difference of each pixel calculated by the second feature amount calculating means with the feature amount relating to the difference obtained in advance for the image data of the region representing the part of the face to be identified; Detecting means for detecting the part to be identified from at least one of the first image data and the second image data;
An image identification device including:   An update means for updating the predetermined feature quantity related to the difference based on the feature quantity related to the difference in the image data of the area representing the part of the face of the person to be identified detected by the detection means. Item 2. The image identification device according to Item 1. The first part is a hair part,
Wherein the second portion, the image identification apparatus according to claim 1 Symbol placement and the skin portion. The first wavelength band and each of the second wavelength band, the image identification apparatus according to any one of claims 1 to 3 which is in the range of 800 to 1100 nm. The first wavelength range is 800 to 950 nm,
Wherein the second wavelength range, the image identification apparatus according to any one of claims 1 to 4 in which the 950～1100Nm.