JP2005108195A - Object identification unit, method, and program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To identify whether or not a predetermined object such as a face is included in an image in a comparatively short processing time. <P>SOLUTION: A feature calculating part 4 calculates a first feature C1 not requiring normalization of the identification target image S0, and a normalized second feature C2. A first identification part 8 identifies whether or not a face candidate is included in the identification target image S0 on the basis of the first feature C1 calculated from the identification target image by referring to first reference data R1 having carried out learning in regard to a multiplicity of face images, and first features C1 of images that are not faces. If a face candidate is included, a second identification part 10 identifies whether or not the face candidate is a face by referring to second reference data R2 having carried out learning in regard to a multiplicity of face images, and second features C2 of images that are not faces. <P>COPYRIGHT: (C)2005,JPO&NCIPI

Description

  The present invention relates to an object identification apparatus and method for identifying whether a predetermined object such as a face is included in an image, and a program for causing a computer to execute the object identification method.

  Image data obtained by a digital camera or image data obtained by reading an image recorded on a film is reproduced as a hard copy such as a print or as a soft copy on a display. An image represented by such image data often includes a human face, and the brightness, gradation, color, sharpness, etc. of the image data are set so that the face has appropriate brightness and color. Image processing for correcting the image is performed. When image processing is performed on image data in this way, it is necessary to detect a face area corresponding to a human face from an image represented by the image data. For this reason, various methods for identifying whether or not a predetermined object such as a face is included in the image have been proposed.

  For example, Non-Patent Document 1 normalizes a luminance value, which is a feature amount used when detecting a face, refers to a learning result of a neural network that has learned the face, and determines whether or not the face is included in the image. It is a technique to identify. Non-Patent Document 2 obtains a high-frequency component such as an edge included in an image as a feature value used for detection of an object, normalizes this feature value, and is called machine learning called boosting. This is a method for identifying whether or not an object is included in an image by referring to a learning result on a feature amount using the above method. These methods of Non-Patent Documents 1 and 2 normalize feature quantities used for detecting an object such as a face, and therefore can accurately identify whether or not an object is included in an image.

In addition, Non-Patent Document 3 detects a tumor shadow that is one of the characteristic forms particularly in breast cancer. For example, on an X-ray negative film, the tumor shadow has a slightly lower concentration value than the surroundings. Using the fact that the gradient vector at any pixel in the tumor shadow is near the center of the tumor shadow, the distribution of the orientation of the gradient vector in the image is evaluated, and the area concentrated at a specific point is determined. This is a method of extracting as a candidate for a shadow of a tumor. Further, Patent Document 1 learns a feature pattern of an object such as a face using Kohonen's self-organization, which is one method of a neural network, and refers to the learning result to identify a candidate object and an object. By determining whether or not the feature portion is included in the learned feature pattern, and further determining whether or not the positional relationship of the feature portion of the candidate object matches the positional relationship of the feature portion of the target object This is a method for determining whether or not a candidate for an object is an object.
Henry A. Rowley, Shumeet Baluja, and Takeo Kanada, "Neural Network-Based Face Detection", volume 20, number 1, pages 23-38, January 1998. Rainer Lienhart, Jochen Maydt, "An Extended Set of Haar-like Features for Rapid Object Detection", International Conference on Image Processing. Obata et al., “Detection of mass shadow in DR image (iris filter)”, IEICE Transactions, D-II Vol.J75-D-II No.3, P663-670, March 1992 Japanese Patent Laid-Open No. 5-282457

  However, the above-described methods of Non-Patent Documents 1 and 2 have a problem that the amount of calculation increases because the feature amount used for detection of the object is normalized, and the processing time required for identification becomes long. There is. In addition, since the method of Non-Patent Document 3 only evaluates the distribution of the gradient vector direction, it can detect an object having a simple shape such as a mass shadow, but can detect a complicated object such as a human face. The object cannot be detected. Moreover, since the method of patent document 1 has many objects to determine, processing requires a long time.

  The present invention has been made in view of the above circumstances, and an object thereof is to identify whether or not a predetermined object such as a face is included in an image in a relatively short processing time.

An object identification apparatus according to the present invention includes an image input unit that receives an input of an image to be identified
First feature amount calculating means for calculating a first feature amount that is not required for normalization used for identifying a predetermined object from the image of the identification target;
The first reference data that predefines the first feature quantity and the identification condition corresponding to the first feature quantity are referred to based on the first feature quantity calculated from the image to be identified. First identifying means for identifying whether or not a predetermined object candidate is included in the image to be identified;
When the first identifying means identifies that the predetermined object candidate is included, a second normalized feature value used for identifying the predetermined object is calculated from the predetermined object candidate. A feature amount calculating means;
Based on the normalized second feature amount calculated from the predetermined object candidate, second reference data that predefines the second feature amount and an identification condition corresponding to the second feature amount. With reference to the above, a second identification unit for identifying whether or not the predetermined object candidate is the predetermined object is provided.

  Examples of the “predetermined object” include an object that has a substantially constant shape and can be arranged to have a substantially constant size. Specifically, a human face, a vehicle, a road sign, and the like can be set as predetermined objects.

  The “feature amount” refers to a parameter representing the feature of the image, a gradient vector representing the density gradient of each pixel in the image, color information (hue, saturation), density, texture feature, depth information of each pixel, Any feature such as a feature of an edge included in the image may be expressed.

  The “first feature amount that does not require normalization” is a feature amount that does not depend on changes in image brightness or contrast. For example, the direction in which the density changes in each pixel of the image and the magnitude of the change, that is, the gradient vector representing the density gradient, is large depending on the density of the pixel and the amount of change in contrast in a specific direction viewed from the pixel. However, the direction of the gradient vector does not change even if the size changes. Further, the color information such as hue does not change even if the image density changes. Therefore, the direction of the gradient vector, color information, and the like can be used as the first feature amount.

  The “second feature amount” means that if the feature amount is used as it is, the feature amount of an image is compared with the same type of feature amount included in the image or the same type of feature amount in another image. It is a feature quantity that cannot be distinguished whether it is large or small and depends on changes in brightness and contrast of the image. As a method for normalizing the second feature amount, for example, when the second feature amount is calculated for each pixel, the predetermined feature is used by using the second feature amounts of all the pixels constituting the predetermined target candidate. A method of normalizing the second feature amount of each pixel included in the candidate, or a second of a plurality of pixels within a predetermined range including a pixel to be normalized among all pixels constituting the predetermined object candidate A method of normalizing the second feature value of the target pixel using the feature value can be used.

  The “identification condition” refers to a condition for discriminating between a predetermined object and an object that is not, using the feature amount as an index.

  In the object identification device according to the present invention, the first reference data includes a plurality of sample images known to be the predetermined object and a plurality of samples known to be not the predetermined object. The first feature amount included in a large number of sample image groups including images may be obtained by learning in advance by a machine learning technique such as neural network or boosting.

Here, when the predetermined object is a face, the first reference data includes a first region in a predetermined range including a left eye and a left cheek in a sample image that is known to be the predetermined object; Corresponding to the first feature amount included in the second region of the predetermined range including the right eye and the right cheek, and the first and second regions in the sample image that is known not to be the predetermined object. It is obtained by learning the first feature amount included in each region,
The first feature amount calculating means may be means for calculating the first feature amount from each region corresponding to the first and second regions in the image to be identified.

Further, the first reference data includes the first feature amount included in a third region of a predetermined range including both eyes in the sample image known to be the predetermined object, and is not the predetermined object. It is obtained by further learning the first feature amount included in an area corresponding to the third area in the sample image that is known.
The first feature amount calculating unit may be a unit that calculates the first feature amount from each region corresponding to the first to third regions in the image to be identified.

  Further, in the object identification device according to the present invention, the second reference data includes a plurality of sample images known to be the predetermined object and a plurality of samples known to be not the predetermined object. The second feature amount included in a large number of sample image groups including images may be obtained by learning in advance by a machine learning method.

Here, when the predetermined object is a face, the second reference data includes a first region in a predetermined range including a left eye and a left cheek in a sample image that is known to be the predetermined object; Corresponding to the second feature amount included in the second region of the predetermined range including the right eye and the right cheek, and the first and second regions in the sample image that is known not to be the predetermined object. It is obtained by learning the second feature amount included in each region,
The second feature amount calculating unit may be a unit that calculates the second feature amount from each region corresponding to the first and second regions in the image to be identified.

In addition, the second reference data is not the predetermined feature and the second feature amount included in the third region of the predetermined range including both eyes in the sample image that is known to be the predetermined target. It is obtained by further learning the second feature amount included in a region corresponding to the third region in the sample image that is known.
The second feature quantity calculating means may be means for calculating the second feature quantity from each area corresponding to the first to third areas in the image to be identified.

  In the object identification device according to the present invention, the first feature amount may be a gradient vector direction or color information in each pixel on the image.

  The “gradient vector” represents the direction in which the density at each pixel of the image changes and the magnitude of the change.

  In the object identification device according to the present invention, the second feature amount may be a direction and a magnitude of a gradient vector in each pixel on the image.

  Further, in the object identification device according to the present invention, it is determined whether or not the identification result by the first identification unit satisfies a predetermined request, and when the determination is affirmative, the image is identified from the identification target image. Only the first feature quantity is calculated, and the first and second feature quantity calculation means, and the first feature quantity calculation means to identify the predetermined object candidate identified by the first identification means from the predetermined object, and Control means for controlling the first and second identification means may be further provided.

  In the object identification device according to the present invention, the predetermined object included in the image identified by the second identification unit based on still another feature amount calculated from the predetermined object candidate, You may make it further provide at least 1 other identification means which identifies whether it is truly a predetermined target object.

  The object identification device according to the present invention may further include an extraction unit that extracts the predetermined object from the image to be identified.

  The object identification device according to the present invention may further include an output unit that outputs the information representing the position of the predetermined object in the identification target image to the identification target image.

  An imaging apparatus such as a digital camera or a camera-equipped mobile phone according to the present invention is characterized by including the object identification apparatus according to the present invention.

The object identification method according to the present invention accepts input of an image to be identified,
Calculating a first feature amount that is not required for normalization used for identification of a predetermined object from the image of the identification object;
The first reference data that predefines the first feature quantity and the identification condition corresponding to the first feature quantity are referred to based on the first feature quantity calculated from the image to be identified. , Identifying whether a predetermined target candidate is included in the image to be identified,
When the first identification means identifies that the predetermined object candidate is included, a normalized second feature value used for identification of the predetermined object is calculated from the predetermined object candidate,
Based on the normalized second feature amount calculated from the predetermined object candidate, second reference data that predefines the second feature amount and an identification condition corresponding to the second feature amount. With reference to the above, it is characterized by identifying whether or not the predetermined object candidate is the predetermined object.

  In addition, you may provide as a program for making a computer perform the target object identification method by this invention.

  According to the present invention, the first feature amount that does not require normalization is calculated from the image to be identified. Then, the first reference data is referred to based on the first feature amount, and it is identified whether or not the predetermined target candidate is included in the image to be identified (first identification). Then, when it is identified that the predetermined object candidate is included, a normalized second feature amount is calculated from the predetermined object candidate, and then the second reference data is based on the second feature amount. Reference is made to identify (second identification) whether or not the predetermined object candidate is the predetermined object. Here, in the first identification, since the first feature amount that does not need to be normalized is used, even if it is determined whether or not the predetermined target candidate is included in the entire identification target image, it is not much. The amount of calculation is not large, and as a result, it is possible to identify whether or not the predetermined target candidate is included in the image to be identified at a relatively high speed. On the other hand, in the second identification, since the normalized second feature amount is used, it can be accurately identified whether or not the predetermined object is included, but the calculation amount increases. However, in the present invention, it is only the portion of the predetermined object candidate in the image to be identified that calculates the second normalized feature amount and performs the second identification. The amount of computation is reduced, and as a result, the time required for the identification process is short. Therefore, according to the present invention, it is possible to identify at high speed and with high accuracy whether or not a predetermined object is included in the image to be identified.

  Moreover, the identification performance of a predetermined object can be further improved by obtaining the first and second reference data by learning in advance by a machine learning method.

  Further, when the predetermined object is a face, during the learning, the first region including the left eye and the left cheek, the second region including the right eye and the right cheek in the sample image, and the second region including both eyes. By using the first and second feature quantities included in the region 3, the learning time can be significantly shortened. When identifying whether or not a predetermined object is included in the image to be identified, the first and second feature values included in the first and second regions, and further the third region are identified. It has been confirmed by the applicant's experiment that it greatly contributes to the improvement of performance. Therefore, learning of the first and second reference data using the first and second feature amounts included in the first and second regions, and further in the third region during learning. Accordingly, it is possible to further improve the identification performance as to whether or not the predetermined object is included in the image to be identified.

  In addition, by calculating the first and second feature amounts from the first and second regions and further from the regions corresponding to the third region from the identification target image, the first and second feature amounts are calculated from the entire identification target image. Since the range for calculating the first and second feature amounts is smaller than when calculating the first and second feature amounts, the calculation time can be shortened.

  Also, the first feature amount is included in the image by setting the direction or color information of the gradient vector at each pixel on the image or the second feature amount as the direction and size of the gradient vector at each pixel on the image. Whether or not the predetermined object is included in the image to be identified can be identified with high accuracy using the feature amount that is relatively easy to calculate.

  Further, it is determined whether or not the identification result satisfies a predetermined request by the first identification unit, and when this determination is affirmed, only the first feature amount is calculated and the first identification is performed, By identifying the identified predetermined object candidate as the predetermined object, the calculation of the normalized second feature amount and the second identification are omitted when the first identification is accurately performed. Therefore, it is possible to identify whether or not the predetermined object is included in the image to be identified at a higher speed.

  Further, by extracting the identified predetermined object, the predetermined object can be extracted with high accuracy from the image of the identification target.

  In addition, by adding information indicating the position of the predetermined object in the identification target image to the identification target image and outputting it, the information given to the identification target can be referred to from the identification target image with high accuracy. An object can be extracted.

  Hereinafter, embodiments of the present invention will be described with reference to the drawings. FIG. 1 is a schematic block diagram showing the configuration of the object identification device according to the first embodiment of the present invention. As shown in FIG. 1, the object identification device 1 according to the first embodiment includes an image input unit 2 that receives input of identification target image data S0 representing an image to be identified, and an identification represented by the identification target image data S0. A feature value calculation unit 4 that calculates first and second feature values C1 and C2 from a target image (hereinafter also referred to as reference image S0) S0, and first and second reference data R1 and R2 to be described later are stored. Based on the first feature amount C1 calculated by the memory 6 and the feature amount calculation unit 4 and the first reference data R1 in the memory 6, the identification target image S0 includes the face of the person who is the predetermined target. When the first identification unit 8 for identifying whether or not a candidate is included and the first identification unit 8 identify that the face candidate is included in the identification target image S0, the feature amount calculation unit 4 Calculated second feature amount C2 and memory A second identification unit 10 for identifying whether or not the face candidate is a face of a person who is a predetermined object, and the first and second identification units 8 based on the second reference data R2 , 10 and an output unit 12 for outputting the identification results.

  The feature quantity calculation unit 4 calculates the first feature quantity C1 that is not required for normalization used for face identification from the identification target image S0, and extracts the second feature quantity C2 from the extracted face candidates as described later. It calculates from the image. Specifically, the direction of the gradient vector of the identification target image S0 is calculated as the first feature amount C1, and the gradient vector (that is, the direction and size) of the image in the face candidate is calculated as the second feature amount C2. Hereinafter, calculation of the gradient vector will be described. First, the feature amount calculation unit 4 performs a filtering process on the identification target image S0 using a horizontal edge detection filter shown in FIG. 2A to detect a horizontal edge in the identification target image S0. The feature amount calculation unit 4 performs filtering processing on the identification target image S0 using a vertical edge detection filter illustrated in FIG. 2B to detect vertical edges in the identification target image S0. Then, as shown in FIG. 3, a gradient vector K for each pixel is calculated from the horizontal edge size H and the vertical edge size V of each pixel on the identification target image S0.

  The direction of the gradient vector K is defined as a first feature amount C1. Specifically, a value from 0 to 359 degrees based on a predetermined direction of the gradient vector K (for example, the x direction in FIG. 3) is set as the first feature amount C1.

  In the case of a human face as shown in FIG. 4 (a), the gradient vector K calculated in this way is the eye in a dark part such as the eyes and mouth as shown in FIG. 4 (b). It faces the center of the mouth and faces outward from the position of the nose in a bright part like the nose. Further, since the change in density is larger in the eyes than in the mouth, the magnitude of the gradient vector K is larger in the eyes than in the mouth.

  Here, the second feature amount C2 is calculated only within the face candidate. Further, the magnitude of the gradient vector K of the second feature quantity C2 is normalized. This normalization obtains a histogram of the magnitude of the gradient vector K at all the pixels in the face candidate, and the distribution of the magnitude is a value that each pixel in the face candidate can take (0 to 255 if 8 bits). The histogram is smoothed so as to be uniformly distributed, and the magnitude of the gradient vector K is corrected. For example, in the case where the gradient vector K is small and the histogram is distributed to the side where the gradient vector K is small as shown in FIG. The magnitude of the gradient vector K is normalized so that it extends over the region so that the histogram is distributed as shown in FIG. In order to reduce the amount of calculation, as shown in FIG. 5C, the distribution range in the histogram of the gradient vector K is divided into, for example, five, and the frequency distribution divided into five is shown in FIG. 5D. It is preferable to normalize so that the value of 0 to 255 is in a range divided into five.

  Here, when shooting, the brightness and direction of illumination vary depending on the conditions at the time of shooting, so the brightness and direction of illumination differ for each identification target image S0. As described above, if the gradient vector K is obtained as it is for each of the identification target images S0 having different brightness and illumination directions, the magnitude of the gradient vector at the eye position is different even though the face is the same. Whether the face candidate is a face cannot be identified. In this case, the magnitude of the gradient vector K may be normalized for the entire identification target image S0. However, the normalization takes a long time because of the large amount of calculation. For this reason, in the present embodiment, the amount of computation is reduced by normalizing the second feature amount only for the face candidate identified by the first identification unit 8 instead of the entire identification target image S0. Processing time is shortened.

  Note that the feature amount calculation unit 4 calculates first and second feature amounts C1 and C2 at each stage of the transformation of the identification target image S0 and the face candidate, as will be described later.

  The first reference data R1 stored in the memory 6 is the first reference data in each pixel constituting each pixel group for each of a plurality of types of pixel groups composed of a combination of a plurality of pixels selected from a sample image to be described later. The identification condition for the combination of the feature amount C1 is defined. In addition, the second reference data R2 is an identification for the combination of the second feature amount C2 in each pixel constituting each pixel group, for each of a plurality of types of pixel groups including a combination of a plurality of pixels selected from the sample image. The conditions are specified.

  A plurality of combinations of the first and second feature values C1 and C2 and identification conditions in each pixel constituting each pixel group in the first and second reference data R1 and R2 are known to be faces. And a sample image group made up of a plurality of sample images that are known to be non-faces.

  In this embodiment, the sample image that is known to be a face has a 30 × 30 pixel size, and as shown in FIG. Pixels, 9 pixels, and 11 pixels, which are rotated stepwise in units of 3 degrees within a range of ± 15 degrees on the plane with respect to a vertically standing face (that is, the rotation angles are −15 degrees, −12 degrees, (-9 degrees, -6 degrees, -3 degrees, 0 degrees, 3 degrees, 6 degrees, 9 degrees, 12 degrees, 15 degrees) sample images shall be used. Therefore, 3 × 11 = 33 sample images are prepared for one face image. Here, in the state where the face stands vertically, the position of the eyes in the vertical direction is the same in all the sample images. In FIG. 6, only sample images rotated at −15 degrees, 0 degrees, and +15 degrees are shown. The center of rotation is the intersection of the diagonal lines of the sample image. As a sample image that is known not to be a face, an arbitrary image having a 30 × 30 pixel size is used.

  Here, as a sample image that is known to be a face, learning is performed using only a center image whose distance between the centers of both eyes is 10 pixels and the rotation angle on the plane is 0 degree (that is, the face is vertical). When it is performed, the face candidate or face that is identified as a face candidate or face with reference to the first and second reference data R1 and R2 is a face candidate or face that is not rotated at all because the distance between the centers of both eyes is 10 pixels. Only. Since the size of a face that may be included in the identification target image S0 is not constant, when identifying whether a face candidate is included or whether a face candidate is a face, identification is performed as described later. The target image S0 is enlarged or reduced so that a face having a size suitable for the size of the sample image can be identified. However, in order to accurately set the distance between the centers of both eyes to 10 pixels, it is necessary to perform identification while gradually enlarging or reducing the size of the identification target image S0 by, for example, 1.1 units. The amount will be enormous.

  Further, the faces that may be included in the identification target image S0 are not only rotated at 0 degrees on the plane as shown in FIG. 7A, but also as shown in FIGS. 7B and 7C. It may be rotating. However, when learning is performed using only a sample image in which the distance between the centers of both eyes is 10 pixels and the rotation angle of the face is 0 degree, FIGS. 7B and 7C are used regardless of the face. As shown in (), the rotated face cannot be identified.

  Therefore, in this embodiment, as a sample image known to be a face, the distance between the centers of both eyes is 9, 10, 11 pixels as shown in FIG. 6, and ± 15 degrees on the plane at each distance. Using the sample image obtained by rotating the face step by step in the range of 3 degrees, the learning of the first and second reference data R1, R2 is allowed. As a result, the identification target image S0 may be scaled up and down in steps of 11/9 as an enlargement rate. For example, the size of the identification target image S0 is scaled up and down in steps of 1.1 units as the enlargement rate, for example. Compared with, the calculation time can be reduced. Further, as shown in FIGS. 7B and 7C, a rotating face can also be identified.

  Hereinafter, an example of a learning method for the sample image group will be described with reference to the flowchart of FIG. Here, learning of the second reference data R2 will be described.

  The group of sample images to be learned includes a plurality of sample images that are known to be faces and a plurality of sample images that are known not to be faces. A sample image known to be a face has 9, 10, 11 pixels at the center of both eyes for each sample image, and is stepped in units of 3 degrees within a range of ± 15 degrees on the plane at each distance. The one with the face rotated is used. Each sample image is assigned a weight or importance. First, the initial value of the weight of all the sample images is set equal to 1 (step S1).

  Next, a discriminator is created for each of a plurality of types of pixel groups in the sample image (step S2). Here, each discriminator provides a reference for discriminating between a face image and a non-face image by using a combination of the second feature amount C2 in each pixel constituting one pixel group. . In the present embodiment, a histogram for a combination of the second feature amount C2 in each pixel constituting one pixel group is used as a discriminator.

The creation of a classifier will be described with reference to FIG. As shown in the sample image on the left side of FIG. 9, each pixel constituting the pixel group for creating the discriminator is a pixel at the center of the right eye on a plurality of sample images that are known to be faces. P1, a pixel P2 on the right cheek, a pixel P3 on the forehead, and a pixel P4 on the left cheek. Then, combinations of the second feature amount C2 in all the pixels P1 to P4 are obtained for all sample images that are known to be faces, and a histogram thereof is created. Here, the second feature amount C2 represents the direction and magnitude of the gradient vector K. Since the gradient vector K has 360 directions from 0 to 359 and the gradient vector K has 256 sizes from 0 to 255, If this is used as it is, the number of combinations is 360 × 256 four pixels per pixel, that is, (360 × 256) ⁴ types, and the number of samples is large for learning and detection. Time and memory are required. For this reason, in this embodiment, the gradient vector directions are 0 to 359, 0 to 44, 315 to 359 (right direction, value: 0), 45 to 134 (upward value: 1), and 135 to 224 (left). Direction, value: 2), 225-314 (downward, value 3), and quaternarization, and the gradient vector magnitude is ternarized (value: 0-2). And the value of a combination is computed using the following formula | equation.

Combination value = 0 (when gradient vector size = 0)
Combination value = ((gradient vector direction + 1) × gradient vector magnitude (gradient vector magnitude> 0)
Thus, since the number of combinations is nine patterns ^4, it can reduce the number of data of the second feature quantity C2.

  Similarly, histograms are created for a plurality of sample images that are known not to be faces. For sample images that are known not to be faces, pixels corresponding to the positions of the pixels P1 to P4 on the sample images that are known to be faces (similarly, reference numerals P1 to P4 are used) are used. It is done. The histogram used as a discriminator shown on the right side of FIG. 9 is a histogram obtained by taking logarithm values of the ratios of the frequency values indicated by these two histograms. The value of each vertical axis indicated by the histogram of the discriminator is hereinafter referred to as an identification point. According to this discriminator, an image showing the distribution of the second feature amount C2 corresponding to the positive discrimination point is highly likely to be a face, and it can be said that the possibility increases as the absolute value of the discrimination point increases. Conversely, an image showing the distribution of the second feature quantity C2 corresponding to the negative identification point is highly likely not to be a face, and the possibility increases as the absolute value of the identification point increases. In step S2, a plurality of classifiers in the above-described histogram format are created for combinations of the second feature amount C2 in each pixel constituting a plurality of types of pixel groups that can be used for identification.

  Subsequently, the most effective classifier for identifying whether or not the image is a face is selected from the plurality of classifiers created in step S2. The most effective classifier is selected in consideration of the weight of each sample image. In this example, the weighted correct answer rates of the classifiers are compared, and the classifier showing the highest weighted correct answer rate is selected (step S3). That is, in the first step S3, since the weight of each sample image is equal to 1, the number of sample images in which the image is correctly identified by the classifier is simply the largest. Selected as a valid discriminator. On the other hand, in the second step S3 after the weight of each sample image is updated in step S5, which will be described later, a sample image with a weight of 1, a sample image with a weight greater than 1, and a sample image with a weight less than 1 The sample images having a weight greater than 1 are counted more in the evaluation of the correct answer rate because the weight is larger than the sample images having a weight of 1. Thereby, in step S3 after the second time, more emphasis is placed on correctly identifying a sample image having a large weight than a sample image having a small weight.

  Next, the correct answer rate of the classifiers selected so far, that is, the result of identifying whether each sample image is a face image using a combination of the classifiers selected so far, is actually It is ascertained whether or not the rate that matches the answer of whether or not the image is a face image exceeds a predetermined threshold (step S4). Here, the sample image group to which the current weight is applied or the sample image group to which the weight is equal may be used for evaluating the correct answer rate of the combination. When the predetermined threshold value is exceeded, learning can be completed because it is possible to identify whether the image is a face with a sufficiently high probability by using the classifier selected so far. If it is equal to or less than the predetermined threshold value, the process proceeds to step S6 in order to select an additional classifier to be used in combination with the classifier selected so far.

  In step S6, the discriminator selected in the most recent step S3 is excluded so as not to be selected again.

  Next, the weight of the sample image that could not be correctly identified as a face by the classifier selected in the most recent step S3 is increased, and the sample image that can be correctly identified as whether or not the image is a face is increased. The weight is reduced (step S5). The reason for increasing or decreasing the weight in this way is that in selecting the next discriminator, an image that cannot be discriminated correctly by the already selected discriminator is regarded as important, and whether or not those images are faces can be discriminated correctly. This is to increase the effect of the combination of the discriminators by selecting the discriminators.

  Subsequently, the process returns to step S3, and the next valid classifier is selected based on the weighted correct answer rate as described above.

  As a discriminator suitable for discriminating whether or not a face is included by repeating the above steps S3 to S6, discrimination corresponding to a combination of the second feature amount C2 in each pixel constituting a specific pixel group If the correct answer rate confirmed in step S4 exceeds the threshold when the device is selected, the type of the discriminator used for identifying whether or not a face is included and the identification condition are determined (step S7). Thus, the learning of the second reference data R2 is finished.

  Then, the first reference data R1 is learned by obtaining the classifier type and the identification condition in the same manner as described above.

  In the case of adopting the above learning method, the classifier uses a combination of the first and second feature amounts C1 and C2 in each pixel constituting a specific pixel group, As long as it provides a criterion for identifying the data, it is not limited to the above-described histogram format, and may be any data such as binary data, a threshold value, or a function. Further, even in the same histogram format, a histogram or the like indicating the distribution of difference values between the two histograms shown in the center of FIG. 9 may be used.

  Further, the learning method is not limited to the above method, and other machine learning methods such as a neural network can be used. Note that the first and second reference data R1 and R2 may be empirically determined by a skilled engineer.

  In the above learning method, the first pixel including the left eye and the left cheek in the sample image that is known to be a face as shown in FIG. 10 is used as a pixel to synthesize a pixel group for creating a discriminator. Only the pixels in the area A1 and the second area A2 including the right eye and the right cheek may be used. Further, in addition to the first and second regions A1 and A2, pixels in the third region A3 including both eyes may be used as indicated by broken lines in FIG.

  In this case, the positions of the areas A1, A2, and A3 are the same in all sample images used for learning. That is, in this embodiment, as shown in FIG. 6, the distance between the centers of both eyes is 9, 10, 11 pixels, and the face is stepped in units of 3 degrees within a range of ± 15 degrees on the plane at each distance. The first and second reference data R1 and R2 are learned using the sample image deformed by rotation. The positions of the regions A1, A2, and A3 on the deformed sample image are determined by the distance between the centers of both eyes. Are the same as the positions of the regions A1, A2, and A3 set in the sample image with 10 pixels and the rotation angle of 0 degrees. Also for sample images that are known not to be faces, the positions of the areas A1, A2, and A3 to be set are samples that are known to be faces whose center-to-center distance between both eyes is 10 pixels and whose rotation angle is 0 degrees. Assume that the positions of the regions A1, A2, and A3 set in the image are the same. Therefore, as shown in FIG. 11, the discriminator is created using only the pixels in the areas A1, A2 and further in the area A3 set on all the sample images used for learning.

  Thus, the learning time of the first and second reference data R1 and R2 is created by creating a discriminator using only the pixels in the first to third regions A1 to A3 in the sample image during learning. Can be greatly shortened.

  Further, when identifying whether or not a face is included in the classification target image S0, a classifier created using pixels included in the first and second regions A1 and A2 and further the third region A3. Has been confirmed by experiments of the present applicant to greatly contribute to the improvement of the discrimination performance. For this reason, at the time of learning, a discriminator is created using only the pixels included in the first and second regions A1, A2, and further the third region A3, and the first and second reference data R1, R2 are created. By performing this learning, it is possible to further improve the identification performance as to whether or not a face is included in the identification target image S0.

  The first identification unit 8 refers to the identification conditions learned by the first reference data R1 for all combinations of the first feature amount C1 in each pixel constituting a plurality of types of pixel groups, and each pixel group An identification point for the combination of the first feature amount C1 in each pixel constituting the pixel is obtained, and all the identification points are combined to identify whether or not a face candidate is included in the classification target image S0. At this time, the direction of the gradient vector K, which is the first feature amount C1, is converted into, for example, a quaternary value as in the case of learning the first reference data R1. In the present embodiment, all the identification points are added, and identification is performed based on the positive / negative of the added value. For example, when the sum of the identification points is a positive value, it is determined that a face candidate is included in the identification target image S0, and when it is a negative value, it is determined that no face candidate is included. The identification of whether or not a face candidate is included in the identification target image S0 performed by the first identification unit 8 is referred to as a first identification.

  Here, the size of the identification target image S0 is different from the sample image of 30 × 30 pixels and has various sizes. When a face is included, the rotation angle of the face on the plane is not always 0 degrees. For this reason, as shown in FIG. 12, the first identification unit 8 enlarges or reduces the identification target image S0 stepwise until the vertical or horizontal size becomes 30 pixels and rotates it 360 degrees stepwise on the plane. (In FIG. 12, a reduced state is shown), a mask M having a 30 × 30 pixel size is set on the identification target image S0 enlarged and reduced at each stage, and the identification target image S0 obtained by scaling the mask M is expanded. By identifying whether or not the image in the mask is a face image while moving one pixel at a time, whether or not a face candidate is included in the identification target image S0 is identified.

  Since the sample images learned at the time of generating the first and second reference data R1 and R2 have 9, 10, and 11 pixels at the center position of both eyes, the identification target image S0 and the face The enlargement ratio at the time of candidate enlargement / reduction may be 11/9. In addition, since the sample images learned at the time of generating the first and second reference data R1 and R2 are images obtained by rotating the face within a range of ± 15 degrees on the plane, the identification target image S0 and the face The candidate may be rotated 360 degrees in units of 30 degrees.

  Here, the feature amount calculation unit 4 calculates the first and second feature amounts C1 and C2 at each stage of deformation, that is, enlargement / reduction and rotation of the identification target image S0 and the face candidate.

  In the learning of the first and second reference data R1, R2, as described above, the pixels in the first and second areas A1, A2, and further in the third area A3 set in the sample image. When the discriminator is created using only the feature amount, the feature amount calculation unit 4 uses only the pixels in each region corresponding to the first and second regions A1 and A2 and further the third region A3 in the mask M. The first and second feature amounts C1, C2 are calculated.

  Whether or not a face candidate is included in the identification target image S0 is determined for the identification target image S0 at all stages of enlargement / reduction and rotation, and if it is identified that the face candidate is included even once, the identification target The image S0 is identified as including a face candidate, and the 30 × 30 pixel corresponding to the position of the identified mask M is identified from the size and rotation angle identification target image S0 when the face candidate is identified as being included. Extract regions as face candidates.

  The second identification unit 10 deforms the face candidate extracted by the first identification unit 8 by rotating the face candidate while scaling up or down in the same manner as the first identification unit 8. In each stage of the transformation, each pixel group is referred to by referring to the identification condition learned by the second reference data R2 for all combinations of the second feature amount C2 in each pixel constituting a plurality of types of pixel groups. An identification point for the combination of the second feature amount C2 in each pixel constituting is obtained, and all the identification points are combined to identify whether the face candidate is a face. At this time, the direction of the gradient vector K, which is the second feature amount C2, is quaternized and the magnitude is ternary. In the present embodiment, all the identification points are added, and identification is performed based on the positive / negative of the added value. For example, when the sum of the identification points is a positive value, the face candidate is determined to be a face, and when the sum is negative, the face candidate is determined not to be a face. The identification of whether or not the face candidate performed by the second identification unit 10 is a face is referred to as a second identification.

  The output unit 12 recognizes that the first identification unit 8 identifies that the identification target image S0 does not include a face candidate, and the first identification unit 8 identifies that the identification target image S0 includes a face candidate. When the second identification unit 10 identifies that the face candidate is not a face, the identification result indicating that the identification target image S0 does not include a face is output. On the other hand, when the second identifying unit 10 identifies that the face candidate identified by the first identifying unit 8 is a face, the face extracted and extracted by trimming the identified face from the identification target image S0 The face image data S1 representing the image is output.

  Next, processing performed in the first embodiment will be described. FIG. 13 is a flowchart showing processing performed in the first embodiment. First, the image input unit 2 accepts input of identification target image data S0 (step S11). At this time, a series of image data S0 related to a large number of images may be continuously received. Next, the feature amount calculation unit 4 calculates the direction of the gradient vector K of the identification target image S0 as the first feature amount C1 at each stage of enlargement / reduction and rotation of the identification target image S0 (step S12). Then, the first identification unit 8 reads the first reference data R1 from the memory 6 (step S13), and performs first identification as to whether or not a face candidate is included in the identification target image S0 (step S14).

  If step S14 is positive, the first identification unit 8 extracts face candidates from the identification target image S0 (step S15). A plurality of face candidates may be extracted. Next, the feature quantity calculation unit 4 calculates the second feature quantity C2 from the face candidate at each stage of the face candidate enlargement / reduction and rotation (step S16), and normalizes the second feature quantity C2 (step S17). . Then, the second identification unit 10 reads the second reference data R2 from the memory 6 (step S18), and performs second identification as to whether or not the face candidate is a face (step S19).

  If step S19 is positive, the output unit 12 extracts the identified face from the identification target image S0, outputs face image data S1 representing the extracted face image (step S20), and ends the process.

  If step S14 and step S19 are negative, the output unit 12 outputs an identification result indicating that the identification target image S0 does not include a face (step S21), and the process ends.

  As described above, the first identification unit 8 of the object identification device 1 according to the first embodiment uses the first feature amount C1 that is the gradient of the gradient vector K that does not require normalization, and thus the identification target. Even if it is identified whether or not a face candidate is included in the entire image S0, the amount of calculation is not so large, and as a result, it can be determined whether or not a face candidate is included in the identification target image S0 at a relatively high speed. On the other hand, since the second identification unit 10 normalizes the second feature amount C2 such as the gradient and the magnitude of the gradient vector K to identify whether the face candidate is a face, the identification accuracy is high. Is expensive, but the amount of computation increases. However, in the present embodiment, the second feature amount is normalized and the second identification is performed only for the face candidate portion extracted from the identification target image S0. As a result, the time required for the identification process is short. Therefore, according to the present embodiment, whether or not a face is included in the identification target image S0 can be identified at high speed and with high accuracy.

  In the identification target image S0, the first and second feature amounts C1 and C2 are calculated from the regions corresponding to the first and second regions A1 and A2 and further the third region A3 in the sample image. Since the range for calculating the first and second feature values C1 and C2 is smaller than when the first and second feature values C1 and C2 are calculated from the entire identification target image S0, the calculation time is shortened. be able to.

  Next, a second embodiment of the present invention will be described. FIG. 14 is a schematic block diagram showing a configuration of an object identification device 1 ′ according to the second embodiment of the present invention. In the second embodiment, the same components as those in the first embodiment are denoted by the same reference numerals, and detailed description thereof is omitted here. In the second embodiment, the image input unit 2, the feature amount calculation unit 4, the memory 6, the first identification unit 8, the second identification unit 10, and the output that constitute the object identification device according to the first embodiment. In addition to the unit 12, it is determined whether or not the identification result by the first identification unit 8 satisfies a predetermined request. If this determination is affirmative, only the first feature amount C1 is calculated, and the first The feature amount calculation unit 4, the first identification unit 8, and the second identification unit 10 are used so that the face candidate identified by the identification unit 8 is identified as a face and the second identification unit 10 does not perform the second identification. The point provided with the control part 14 which controls is different from the first embodiment.

  The control unit 14 starts the first identification and the second identification with respect to the identification target image data S0, and when the number of identification target image data S0 that has been identified reaches a predetermined number, The number of times that the unit 8 identifies that the face candidate is included in the identification target image S0 (N1) is compared with the number of times that the second identification unit 10 identifies that the face candidate is a face (N2). Whether the identification result of the first identification unit 8 satisfies a predetermined request by determining whether the ratio N2 / N1 of the number N2 to the number N1 is, for example, a predetermined ratio (for example, 0.95) or more. Determine whether or not. If this determination is affirmative, it is determined that the identification accuracy of the face candidate by the first identification unit 8 is very high, and the feature amount calculation unit 4 performs first identification on the target image data S0 to be identified. Only the feature amount C1 is calculated, and only the first identification unit 8 identifies whether or not a face candidate is included in the identification target image S0. If it is identified that the face candidate is included, the face candidate is identified. The feature amount calculation unit 4, the first identification unit 8, and the second identification unit 10 are identified so that the identification target image S 0 is identified as including a face and the identification result is output to the output unit 12. Control.

  Next, processing performed in the second embodiment will be described. In the second embodiment, the processing performed in the image input unit 2, the feature amount calculation unit 4, the memory 6, the first identification unit 8, the second identification unit 10, and the output unit 12 is the first implementation. Since the process is the same as that performed in the embodiment, only the process performed by the control unit 14 will be described here.

  FIG. 15 is a flowchart showing processing performed in the second embodiment. When the process of identifying whether or not a face is included in the identification target image S0 is started, the control unit 14 starts the process, and the first identification unit 8 has identified that a face candidate is included in the identification target image S0. The number of times N1 is counted (step S31). On the other hand, the number N2 of times when the second identification unit 10 identifies that the face candidate is a face is counted (step S32).

  Next, the control unit 14 determines whether or not the number of identification target images S0 that have been identified (that is, the number of identifications) has reached a predetermined number (step S33). If step S33 is negative, the process returns to step S31, and the processing from step S31 to step S33 is repeated until step S33 is positive. If step S33 is affirmed, it is determined whether or not the ratio N2 / N1 of the number of times N2 to the number of times N1 is equal to or greater than a predetermined ratio (step S34).

  When step S34 is affirmed, the feature quantity calculation unit 4 calculates only the first feature quantity C1 and uses the first identification unit 8 alone to identify whether or not the identification target image S0 includes a face. Then, the first identification unit 8 and the second identification unit 10 are controlled (step S35), and the process ends. On the other hand, if step S34 is negative, the first and second feature amounts C1 and C2 are continuously calculated, and whether the identification target image S0 includes a face using the first and second identification units 8 and 10 or not. The feature amount calculation unit 4, the first identification unit 8, and the second identification unit 10 are controlled so as to identify whether or not (step S36), and the process ends.

  As described above, in the second embodiment, it is determined whether or not the identification result by the first identification unit 8 satisfies a predetermined request. If this determination is affirmative, Only one feature quantity C1 is calculated, and only the first identification unit 8 is used to identify the face candidate identified by the first identification unit 8 as a face. For this reason, when the 1st discriminating part 8 is discriminating with sufficient precision, the calculation of the normalized 2nd feature-value C2 and the 2nd discriminating which the 2nd discriminating part 10 performs are abbreviate | omitted. Accordingly, it is possible to identify whether or not a face is included in the identification target image S0 at a higher speed.

  In the first and second embodiments, the first and second reference data R1 and R2 are stored in the memory 6 in the apparatus 1. However, the feature amount calculation unit 4, As long as the discriminating unit 8 and the second discriminating unit 10 can access the first and second reference data R1 and R2, the first and second reference data R1 and R2 are separated from the device 1 by a device or CD. It may be stored in a replaceable medium such as a ROM.

  In the first and second embodiments, the gradient of the gradient vector K is used as the first feature amount C1 that does not require normalization. However, color information such as hue and saturation of the identification target image S0 is used. Similarly to the gradient vector K, the color information of the identification target image S0 may be used as the first feature amount because it remains unchanged even if the brightness and contrast of the identification target image S0 change.

  In the first and second embodiments, a face is identified as an identification target, and whether or not the identification target image S0 includes a face is identified. Automobiles, road signs, and the like that can be matched in size when learning may be used as identification objects.

  In the first and second embodiments, the output unit 12 extracts a face from the identification target image S0. However, face position information (for example, an identified face) indicating the position of the face in the identification target image S0 is used. (Coordinates of the four corners of the rectangular area surrounding the image) may be added to the identification target image data S0, and the identification target image data S0 to which the face position information is added may be output. Here, in order to add the face position information to the identification target image data S0, the face position information is described in the header or tag of the identification target image data S0, or the file name has the same file name as the identification target image data S0. For example, it is possible to use a method in which face position information is described in a different text file and the identification target image data S0 and the text file are inseparably integrated. When it is identified that the identification target image S0 does not include a face, identification information indicating the identification result may be added to the identification target image data S0 and output.

  In the first and second embodiments, the feature amount calculation unit 4 calculates the first and second feature amounts C1 and C2. However, the first feature amount C1 and the second feature amount are calculated. A dedicated feature amount calculation unit for calculating C2 may be provided.

  Further, in the first and second embodiments, two identification units, the first and second identification units 8 and 10, are used, but the object according to the third embodiment of the present invention shown in FIG. As in the object identification device 1 ″, a third identification unit 16 may be further provided.

  The third identification unit 16 of the object identification device 1 ″ according to the third embodiment learns about a feature quantity (referred to as a third feature quantity C3) different from the first and second feature quantities C1 and C2. The face identified by the second identification unit 10 based on the third feature amount C3 calculated from the identification target image S0 with reference to another reference data (referred to as third reference data R3) that has been subjected to In this manner, the third identification unit 16 is further provided to further improve the identification accuracy of whether or not the identification target image S0 includes a face. In the third embodiment, one identification unit called the third identification unit 16 is added to the first and second identification units 8 and 10, but a plurality of identification units are further added. May be.

  In the first to third embodiments, the object identification device according to the present invention is used as a single unit. However, the object identification device according to the present invention is image data obtained by photographing with a digital camera, a camera-equipped mobile phone, or the like. You may make it provide in the imaging device which acquires. Thus, in the imaging apparatus, when performing face detection, red-eye correction, or detection of whether or not eyes are closed on the image represented by the image data, the face and the eye position are recognized. Can do.

  As described above, the devices according to the first to third embodiments of the present invention have been described. However, the computer includes the image input unit 2, the feature amount calculation unit 4, the memory 6, the first identification unit 8, and the second. A program that functions as means corresponding to the identification unit 10, the output unit 12, the control unit 14, and the third identification unit 16 and that performs processing for identifying whether or not a face is included in the identification target image S0 is also included in the present invention. This is one of the embodiments. A computer-readable recording medium that records such a program is also one embodiment of the present invention. In these cases, the reference data may be included in the program or the same recording medium, or may be provided from an external device or a separate medium.

1 is a schematic block diagram showing the configuration of an object identification device according to a first embodiment of the present invention. (A) is a diagram showing a horizontal edge detection filter, (b) is a diagram showing a vertical edge detection filter Diagram for explaining calculation of gradient vector (A) is a figure which shows a person's face, (b) is a figure which shows the gradient vector of eyes and mouth vicinity of the person's face shown to (a). (A) is a diagram showing a histogram of the magnitude of a gradient vector before normalization, (b) is a diagram showing a histogram of the magnitude of a gradient vector after normalization, and (c) is a magnitude of a gradient vector obtained by quinarization. The figure which shows the histogram of the length, (d) is a figure which shows the histogram of the magnitude | size of the quinary gradient vector after normalization Figure showing an example of a sample image that is known to be a face Illustration for explaining face rotation Flow chart showing learning method of reference data Diagram showing how to derive a classifier The figure which shows the state which set the 1st area | region containing a left eye and a left cheek, the 2nd area | region containing a right eye and a right cheek, and also the 3rd area | region containing both eyes in the sample image. The figure which shows the state which set the 1st-3rd area | region to the deformed sample image The figure for demonstrating the stepwise deformation | transformation of an identification object image The flowchart which shows the process performed in 1st Embodiment. The schematic block diagram which shows the structure of the target object identification apparatus by the 2nd Embodiment of this invention. The flowchart which shows the process which the control part of 2nd Embodiment performs The schematic block diagram which shows the structure of the target object identification apparatus by the 3rd Embodiment of this invention.

Explanation of symbols

1, 1 ′, 1 ″ Object identification device 2 Image input unit 4 Feature quantity calculation unit 6 Memory 8 First identification unit 10 Second identification unit 12 Output unit 14 Control unit 16 Third identification unit

Claims (16)

Image input means for receiving input of an image to be identified;
First feature amount calculating means for calculating a first feature amount that is not required for normalization used for identifying a predetermined object from the image of the identification target;
The first reference data that predefines the first feature quantity and the identification condition corresponding to the first feature quantity are referred to based on the first feature quantity calculated from the image to be identified. First identifying means for identifying whether or not a predetermined object candidate is included in the image to be identified;
When the first identifying means identifies that the predetermined object candidate is included, a second normalized feature value used for identifying the predetermined object is calculated from the predetermined object candidate. A feature amount calculating means;
Based on the normalized second feature amount calculated from the predetermined object candidate, second reference data that predefines the second feature amount and an identification condition corresponding to the second feature amount. And a second identification means for identifying whether or not the predetermined object candidate is the predetermined object.   The first reference data is included in a number of sample image groups including a plurality of sample images known to be the predetermined object and a plurality of sample images known to be not the predetermined object. The object identifying apparatus according to claim 1, wherein the first feature amount is obtained by learning in advance by a machine learning method. When the predetermined object is a face, the first reference data includes a first region of a predetermined range including a left eye and a left cheek in a sample image known to be the predetermined object, and a right eye and a right The first feature amount included in the second region of the predetermined range including the cheek, and each region corresponding to the first and second regions in the sample image that is known not to be the predetermined object It is obtained by learning the first feature amount included,
The first feature amount calculating means is means for calculating the first feature amount from each region corresponding to the first and second regions in the image to be identified. 2. The object identification device according to 2. The first reference data is not the predetermined feature and the first feature amount included in the third region of the predetermined range including both eyes in the sample image that is known to be the predetermined target. Obtained by further learning the first feature amount included in a region corresponding to the third region in the known sample image,
The first feature amount calculating unit is a unit that calculates the first feature amount from each region corresponding to the first to third regions in the image to be identified. 3. The object identification device according to 3.   The second reference data is included in a number of sample image groups including a plurality of sample images that are known to be the predetermined object and a plurality of sample images that are known not to be the predetermined object. 5. The object identification device according to claim 1, wherein the second feature amount is obtained by learning in advance by a machine learning method. When the predetermined object is a face, the second reference data includes a first region of a predetermined range including a left eye and a left cheek and a right eye and a right in a sample image that is known to be the predetermined object. The second feature amount included in the second region of the predetermined range including the cheek, and each region corresponding to the first and second regions in the sample image that is known not to be the predetermined object It is obtained by learning the second feature amount included,
The second feature quantity calculating means is a means for calculating the second feature quantity from each area corresponding to the first and second areas in the image to be identified. 5. The object identification device according to 5. The second reference data is not the predetermined feature and the second feature amount included in the third region of the predetermined range including both eyes in the sample image that is known to be the predetermined target. Obtained by further learning the second feature amount included in a region corresponding to the third region in the known sample image,
7. The second feature quantity calculating means is a means for calculating the second feature quantity from each area corresponding to the first to third areas in the image to be identified. The object identification device described.   The object identification device according to claim 1, wherein the first feature amount is a direction of a gradient vector or color information at each pixel on the image.   9. The object identification device according to claim 1, wherein the second feature amount is a direction and a magnitude of a gradient vector in each pixel on the image.   It is determined whether or not the identification result by the first identification means satisfies a predetermined request, and when the determination is affirmative, only the first feature amount is calculated from the image to be identified, Control for controlling the first and second feature quantity calculating means and the first and second identifying means so as to identify the predetermined object candidate identified by the first identifying means from the predetermined object. The object identification device according to claim 1, further comprising means.   Based on yet another feature amount calculated from the predetermined object candidate, it is identified whether or not the predetermined object included in the image identified by the second identifying means is truly the predetermined object. The object identifying device according to any one of claims 1 to 9, further comprising at least one other identifying means.   The target object identification apparatus according to claim 1, further comprising an extraction unit configured to extract the predetermined target object from the identification target image.   12. The output device according to claim 1, further comprising an output unit configured to output information indicating the position of the predetermined object in the identification target image by adding the information to the identification target image. Object identification device.   An imaging apparatus comprising the object identification device according to claim 1. Accept input of image to be identified,
Calculating a first feature amount that is not required for normalization used for identification of a predetermined object from the image of the identification object;
The first reference data that predefines the first feature quantity and the identification condition corresponding to the first feature quantity are referred to based on the first feature quantity calculated from the image to be identified. , Identifying whether a predetermined target candidate is included in the image to be identified,
When the first identification means identifies that the predetermined object candidate is included, a normalized second feature value used for identification of the predetermined object is calculated from the predetermined object candidate,
Based on the normalized second feature amount calculated from the predetermined object candidate, second reference data that predefines the second feature amount and an identification condition corresponding to the second feature amount. The object identification method is characterized by identifying whether or not the predetermined object candidate is the predetermined object. A procedure for accepting input of an image to be identified;
A procedure for calculating a first feature amount used for identification of a predetermined object that does not require normalization from the image of the identification object;
The first reference data that predefines the first feature quantity and the identification condition corresponding to the first feature quantity are referred to based on the first feature quantity calculated from the image to be identified. , A procedure for identifying whether or not a predetermined object candidate is included in the image to be identified;
A step of calculating, from the predetermined object candidate, a normalized second feature amount used for identifying the predetermined object when the first identifying means identifies that the predetermined object candidate is included;
Based on the normalized second feature amount calculated from the predetermined object candidate, second reference data that predefines the second feature amount and an identification condition corresponding to the second feature amount. A program for causing a computer to execute an object identification method, comprising: a step of identifying whether or not the predetermined object candidate is the predetermined object.

Images