JP2011039994A - Component detector, component detection method, program, and recording medium - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a component detector for accurately and easily detecting a component even when there is a part hidden in an image to be identified. <P>SOLUTION: The component detector includes: a target image acquiring means 111 for acquiring an image to be identified, a target region detecting means 121 for detecting a region to be identified from the target image by referring to a region extracting model created from region teacher data 141 obtained in advance, and a component detecting means 122 for detecting a component included in the target region by referring to a component extracting model created from component teacher data 142 obtained in advance. <P>COPYRIGHT: (C)2011,JPO&INPIT

Description

  The present invention relates to a component detection apparatus, a component detection method, a program, and a recording medium.

  A system for detecting a specific facial part from an image is used in various fields such as various image communication such as a video conference and a video phone, a database search, a security system such as building monitoring, and a face matching system. For example, a method for extracting facial parts such as eyes, eyebrows, nostrils, and mouth from a human face image has been proposed (see, for example, Patent Document 1). However, there is a problem that extraction is difficult when the face part is hidden, and erroneous detection is performed when a similar face part exists in the image.

  In recent years, due to security demands, calls and notices have been made to remove sunglasses, masks, hats, etc. in the operation of self-checkout machines equipped with automatic teller machines (ATMs) and point-of-sale information management (POS) systems. However, the effect is not necessarily improved depending on the unmanned case and time zone. Therefore, there is a need for a system that can detect an operator wearing sunglasses or the like, and can perform a warning by voice or limit an operation while wearing. There is also a need for a system that can search for a person who appears to be a criminal wearing sunglasses or the like from a large number of video archives taken by a security camera.

Japanese Patent Laid-Open No. 10-307923

  However, for example, when face detection is performed by extracting eyes, (1) eyes are extracted from the identification target image by image processing, and (2) sunglasses are extracted when eyes cannot be extracted. (3) If sunglasses cannot be extracted, it is necessary to construct an algorithm that searches for possible items for each cause of hiding when the eyes are hidden, such as extracting a hat. Yes, it is complicated.

  Therefore, an object of the present invention is to provide a component detection apparatus, a component detection method, a program, and a recording medium that can detect a component with high accuracy and easily even when there is a component hidden in the identification target image. .

In order to achieve the above object, the component detection apparatus of the present invention includes an identification target image acquisition unit that acquires an identification target image;
An identification target area detecting means for detecting an identification target area from the identification target image with reference to an identification target area extraction model created from the identification target area teacher data acquired in advance;
And a component detection unit that detects a component included in the identification target region with reference to a component extraction model created from component teacher data acquired in advance.

The component detection method of the present invention includes an identification target image acquisition step of acquiring an identification target image,
An identification target area detection step of detecting an identification target area from the identification target image with reference to an identification target area extraction model created from the identification target area teacher data acquired in advance;
A component detection step of detecting a component included in the identification target region with reference to a component extraction model created from component teacher data acquired in advance.

  The program of the present invention causes a computer to execute the component detection method of the present invention.

  The recording medium of the present invention records the program of the present invention.

  According to the present invention, it is possible to provide a component detection apparatus, a component detection method, a program, and a recording medium that can detect a component with high accuracy and easily even when there is a component hidden in the identification target image.

FIG. 1A is a flowchart showing an example of the component detection method of the present invention, and FIG. 1B is a block diagram showing an example of the component detection apparatus of the present invention. FIG. 2A is a flowchart showing another example of the component detection method of the present invention, and FIG. 2B is a block diagram showing another example of the component detection apparatus of the present invention. FIG. 3A is a flowchart showing still another example of the component detection method of the present invention, and FIG. 3B is a block diagram showing still another example of the component detection apparatus of the present invention. FIG. 4 is a diagram for explaining an example of acquisition of identification target area teacher data according to the present invention. FIGS. 5A to 5C are diagrams for explaining an example of acquisition of component teacher data in the present invention. 6A and 6B are diagrams for explaining an example of the identification target region detection step of the present invention. 7A to 7C are diagrams for explaining an example of the component detection step of the present invention. FIG. 8 is a block diagram showing a configuration of an example of a component detection system using the component detection apparatus of the present invention.

  Next, the component detection apparatus, component detection method, program, and storage medium of the present invention will be described with examples. However, the present invention is not limited to the following examples.

(Embodiment 1)
FIG. 1 shows a flowchart of the component detection method according to the present embodiment (FIG. 1A) and a block diagram of the component detection apparatus (FIG. 1B). As shown in FIG. 1B, the component detection apparatus of this embodiment includes an identification target image acquisition unit 111, a calculation unit 120, an output unit 131, and a data storage unit 140 as main components. The data storage unit 140 stores identification target area teacher data 141 and component teacher data 142 acquired in advance, and the calculation unit 120 includes an identification target region detection unit 121 and a component detection unit 122. The identification target area detection unit 121 is connected to the identification target area teacher data 141. The component detection unit 122 is connected to the component teacher data 142. Examples of the identification target image acquisition unit 111 include a CCD (Charge Coupled Device) camera, a CMOS (Complementary Metal Oxide Semiconductor) camera, and an image scanner. Examples of the arithmetic unit 120 include a central processing unit (CPU). Examples of the output means 131 include a monitor that outputs video (for example, various image display devices such as a liquid crystal display (LCD) and a cathode ray tube (CRT) display), a printer that outputs by printing, a speaker that outputs by sound, and the like. can give. The output means 131 is an arbitrary component and may not be included in the component detection apparatus of the present invention, but is preferably included. Examples of the data storage means 140 include a random access memory (RAM), a read-only memory (ROM), a hard disk (HD), an optical disk, a floppy (registered trademark) disk (FD), and the like. The data storage unit 140 may be a device built-in type or an external type such as an external storage device. The identification target image acquisition unit, the calculation unit, the output unit, and the data storage unit are the same in the second, third, and fifth embodiments described later.

  In the present invention, the identification target area is not particularly limited, and may be any person such as a person, a person's head, an animal, and a passenger car.

  In the present invention, the component is not particularly limited as long as it is included in the identification target region, and any component may be used. For example, when the identification target region is a person, examples of the component include a head, a torso, a left hand, a right hand, a left leg, and a right leg. For example, when the identification target region is a human head, examples of the component include a left eye, a right eye, a nose, a mouth, a left eyebrow, a right eyebrow, a left ear, and a right ear. For example, when the identification target region is an animal, examples of the component include a head, a trunk, a leg, and a tail. For example, when the identification target region is a passenger car, examples of the component include a vehicle body and a tire.

  Hereinafter, the present embodiment will be described by taking as an example the case where the identification target area is a human head (hereinafter, simply referred to as “head”) and the parts are four of a left eye, a right eye, a nose, and a mouth. The component detection device and the component detection method of the embodiment will be described in more detail.

  The component detection method of this example is implemented as follows using the component detection apparatus of FIG. 1 (B), for example.

  First, prior to the implementation of the component detection method, learning is performed using a machine learning (pattern recognition) technique. Specifically, first, identification target area teacher data 141 is given to the identification target area (head) of the learning image. In this example, the identification target area teacher data 141 includes teacher data such as the position, size, and orientation of the head. Specifically, for example, as shown in FIG. 4, first, by attaching a labeling 11 to the head of the learning image 10, teacher data of the position and size of the head is given. Furthermore, teacher data of the head orientation such as orientation data in units of 1 degree (for example, 3 degrees to the left) and orientation data in accordance with 8 directions (for example, diagonally left) is given. Teacher data such as the position, size, and orientation of the head is given by a person using a conventionally known input means such as a keyboard and a mouse. In consideration of detection accuracy, the number of learning images 10 is preferably as large as possible, and identification target region teacher data 141 such as the position, size, and orientation of a large number of heads is assigned to a large number of learning images 10 prepared. Is preferred.

  Next, the part teacher data 142 is assigned to the face parts (left eye, right eye, nose and mouth) included in the identification target area (head) to which the identification target area teacher data 141 is assigned. In this example, examples of the component teacher data 142 include teacher data such as the positions of the left eye, the right eye, the nose and the mouth, and the presence / absence of hiding. Specifically, for example, as shown in FIG. 5 (A), first, labeling 12a to 15a is attached with ○ to each of the left eye, right eye, nose and mouth included in the head labeling 11 in FIG. , Teacher data (for example, (x, y) coordinates) of the positions of the left eye, the right eye, the nose and the mouth, and teacher data indicating that they are not hidden. For example, as shown in FIG. 5 (B), when the left eye and the right eye are hidden by sunglasses, the left eye and the right eye are labeled by adding labeling 12b and 13b to each of the left eye and the right eye. Teacher data (for example, (x, y) coordinates) at each position of the right eye and teacher data indicating that they are hidden are assigned. In this case, the labeling of the nose and the mouth and the provision of teacher data are the same as those shown in FIG. Also, for example, as shown in FIG. 5C, when the nose and the mouth are hidden by a mask, the nose and the mouth can be obtained by attaching labeling 14b and 15b with x to the nose and the mouth, respectively. Teacher data (for example, (x, y) coordinates) at each position of the mouth and teacher data indicating that they are hidden are given. In this case, the labeling of the left eye and the right eye and the assignment of teacher data are the same as those shown in FIG. Teacher data such as the position of each of the left eye, the right eye, the nose and the mouth and the presence / absence of hiding are given by a person using a conventionally known input means such as a keyboard and a mouse.

  Next, as shown in FIGS. 1A and 1B, an identification target image is acquired by the identification target image acquisition unit 111 (step S11).

  Next, as shown in FIGS. 1 (A) and 1 (B), the identification target area extraction model created from the identification target area teacher data 141 acquired in advance by the identification target area detection means 121 is referred to. Identification target area data is detected from the identification target image (step S21). In this example, examples of the identification target area data include data such as the position, size, and orientation of the head. Details of the position, size, and orientation data of the head are the same as the teacher data for the position, size, and orientation of the head. Specifically, for example, as shown in FIG. 6A, with reference to the identification target region extraction model created from the identification target region teacher data 141 acquired in advance, from the upper left end of the identification target image 20 The head is searched by a so-called raster scan in which it is searched in the horizontal direction in order toward the lower row. If the head cannot be detected, as shown in FIG. 6B, the identification target area created from the identification target area teacher data 141 acquired in advance by changing the acquisition size of the identification target image 20 With reference to the extracted model, the head is searched by the raster scan. By repeating the change of the acquisition size until the head is detected, data such as the position, size, and orientation of the head is detected.

  Next, as shown in FIGS. 1A and 1B, the component detection unit 122 refers to a component extraction model created from the component teacher data 142 acquired in advance, and is included in the identification target region. The detected face part data is detected (step S22). In this example, the face part data includes, for example, data of the positions of the left eye, right eye, nose and mouth, and the presence / absence of hiding. The details of the data on the positions of the left eye, the right eye, the nose and the mouth and the presence / absence of hiding are the same as the teacher data on the positions of the left eye, the right eye, the nose and the mouth and the presence / absence of the hiding. is there. Specifically, for example, as shown in FIG. 7A, the left eye, the right eye, the nose, and the mouth are detected with reference to the part extraction model created from the part teacher data 142 acquired in advance. Then, the left eye, the right eye, the nose and the mouth are labeled with a circle 22a to 25a, respectively, and the position data of the left eye, the right eye, the nose and the mouth (for example, (x, y ) Coordinates) and data that they are not hidden are detected. For example, as shown in FIG. 7B, when the left eye and the right eye are hidden by sunglasses, the left eye and the right eye are labeled with x and labeled 22b and 23b, respectively. Data of each position (for example, (x, y) coordinates) and data indicating that they are hidden are detected. In this case, the detection of the data of the nose and the mouth is the same as the case shown in FIG. Further, for example, as shown in FIG. 7C, when the nose and the mouth are hidden by a mask, labeling 24b and 25b are attached to the nose and the mouth with x, respectively, and the nose and the mouth Data of each position (for example, (x, y) coordinates) and data indicating that they are hidden are detected. In this case, the detection of the data for the left eye and the right eye is the same as that shown in FIG.

  Next, as shown in FIGS. 1A and 1B, the output unit 131 outputs at least the face part data among the identification target region data and the face part data (step S31). Examples of the identification target region data to be output include head position (for example, (x, y) coordinates), size, and orientation data. Examples of the face part data to be output include data indicating that the left eye and the right eye are hidden. The output step S31 is an optional step and may not be included in the component detection method of the present invention, but is preferably included.

  As described above, in the component detection apparatus and component detection method of this embodiment, prior to the implementation, the identification target region teacher data and the component teacher data are acquired in advance using a machine learning (pattern recognition) technique. Then, the identification target area data and the face part data are detected with reference to the identification target area extraction model and the part extraction model created from them. For this reason, according to the component detection apparatus and component detection method of the present embodiment, it is possible to easily detect the identification target area data and the facial component data with high accuracy. As described above, in the present embodiment, since the part teacher data includes a face part such as a left eye that is hidden, the face part is hidden even when the face part is hidden in the identification target image. Data can be detected with high accuracy and ease. The same applies to later-described Embodiments 2 and 3.

  In the present embodiment, an image of one person is used as the learning image and the identification target image, but the present invention is not limited to this. By including images of two or more persons in the learning image and using the machine learning (pattern recognition) technique, the identification target region teacher data and the component teacher data are acquired in advance, thereby the identification target image. Even if it is an image of two or more persons, it is possible to easily detect the identification target area data and the component data of each person with high accuracy. The same applies to later-described Embodiments 2 and 3.

  As described above, in the present invention, for example, a person image may be used as the learning image and the identification target image. The learning image includes an image in which two persons are overlapped and three legs are seen, and identification target region teacher data and component teacher data are acquired in advance using a machine learning (pattern recognition) technique. Thus, even if the identification target image is an image in which two persons overlap and the legs look three, one of the three legs is erroneously detected as the tail of an animal. There is nothing. The same applies to later-described Embodiments 2 and 3.

(Embodiment 2)
Next, FIG. 2 shows a flowchart (FIG. 2A) and a block diagram (FIG. 2B) of an example including attribute recognition. In this figure, the same parts as those in FIG. As shown in FIG. 5B, in the component detection apparatus of the present embodiment, attribute recognition teacher data 143 is further stored in the data storage unit 140, and the calculation unit 120 includes an attribute recognition unit 123. Except for this point, the configuration is the same as that of the component detection apparatus shown in FIG. The attribute recognition unit 123 is connected to the attribute recognition teacher data 143.

  Hereinafter, in the same manner as in the first embodiment, the component detection apparatus and the component detection method according to the present embodiment are described by taking, as an example, the case where the identification target region is the head and the components are the left eye, right eye, nose, and mouth. Will be described in more detail.

  The component detection method of this example is implemented as follows using the component detection apparatus of FIG. 2 (B), for example.

  First, prior to the implementation of the component detection method, learning is performed using a machine learning (pattern recognition) technique. Specifically, first, similarly to the first embodiment, the identification target area teacher data 141 is given to the identification target area (head) of the learning image.

  Next, in the same manner as in the first embodiment, the part teacher data 142 is assigned to the facial parts (left eye, right eye, nose and mouth) included in the identification target area (head) to which the identification target area teacher data 141 is assigned. .

  Next, attribute recognition teacher data 143 is acquired by assigning an attribute to the identification target region teacher data 141. In this example, the attributes include, for example, the presence / absence of a cover (eg, hat, hood, etc.), the presence / absence of glasses, the presence / absence of sunglasses, the presence / absence of a mask, gender, estimated age, and the like. The attribute is given by a person using conventionally known input means such as a keyboard and a mouse.

  Next, as shown in FIGS. 2A and 2B, the identification target image is acquired by the identification target image acquisition unit 111 in the same manner as in the first embodiment (step S11).

  Next, as shown in FIGS. 2A and 2B, as in the first embodiment, the identification target created from the identification target area teacher data 141 acquired in advance by the identification target area detection unit 121. With reference to the region extraction model, identification target region data is detected from the identification target image (step S21).

  Next, as shown in FIGS. 2A and 2B, the component extraction model created from the component teacher data 142 acquired in advance by the component detection unit 122 is referred to as in the first embodiment. Then, the data of the face part included in the identification target area is detected (step S22).

  Next, as shown in FIGS. 2A and 2B, the attribute of the face part is recognized with reference to the attribute extraction model created from the attribute recognition teacher data 143 acquired in advance (step S23). ). The attribute is the same as that given to the identification target area teacher data 141.

  Next, as shown in FIGS. 2A and 2B, the output unit 131 outputs at least the face part data among the identification target area data, the face part data, and the attribute (step S31). ). The output identification target area data and face part data are the same as those in the first embodiment. Examples of the output attribute include attributes such as no cover, no glasses, and no mask. The output step S31 is an optional step and may not be included in the component detection method of the present invention, but is preferably included.

  Whether or not each face part is hidden in the first embodiment may be determined by adding the attribute and recognizing the attribute in the present embodiment in addition to or instead of acquiring the teacher data and detecting the data. The same applies to the third embodiment described later.

  As described above, in the component detection apparatus and the component detection method of the present embodiment, the attribute recognition teacher in which attributes are added to the identification target region teacher data acquired using the machine learning (pattern recognition) technique prior to the implementation. Data is acquired in advance, and attributes are recognized with reference to an attribute extraction model created therefrom. For this reason, according to the component detection apparatus and component detection method of this embodiment, an attribute can be recognized with high accuracy and easily. The same applies to the third embodiment described later.

  Using the component detection apparatus and the component detection method of this embodiment, for example, when face component data indicating that there is a hiding in the left eye and the right eye is detected, and an attribute indicating that sunglasses is present is recognized, the sunglasses are removed. It is possible to construct a system that can warn by voice and limit operations while wearing. Further, when the component detection device and the component detection method of the present embodiment are used, for example, face component data indicating that there is no hiding in the nose and hiding in the mouth is detected, and an attribute that there is a mask is recognized. In addition, it is possible to construct a system capable of warning by voice so as to cover the nose with a mask. Furthermore, using the component detection apparatus and the component detection method of the present embodiment, for example, facial component data indicating that the left eye, right eye, nose and mouth are hidden is detected, and the attribute indicating that sunglasses and a mask are present is recognized. A system for searching for a person who seems to be a criminal can be constructed by extracting an identification target image. The same applies to Embodiment 3 described later.

  In this embodiment, the acquisition of attribute recognition teacher data and the recognition of attributes are performed after the acquisition of component teacher data and the detection of face component data, respectively, but the present invention is not limited to this, and the teacher for attribute recognition Data acquisition and attribute recognition may be performed simultaneously with acquisition of part teacher data and detection of face part data, respectively.

(Embodiment 3)
Next, FIG. 3 shows a flowchart (FIG. 3A) and a block diagram (FIG. 3B) of an example including alignment (alignment). In this figure, the same parts as those in FIGS. 1 and 2 are denoted by the same reference numerals. As shown in FIG. 2B, the component detection apparatus of the present embodiment further has the same configuration as the component detection apparatus shown in FIG. 2B except that the calculation unit 120 includes an alignment unit 124. It is.

  Hereinafter, as in Embodiments 1 and 2, the case where the identification target region is the head, and there are four parts, ie, the left eye, the right eye, the nose, and the mouth, is taken as an example. The detection method will be described in further detail.

  The component detection method of this example is implemented as follows using the component detection apparatus of FIG. 3 (B), for example.

  First, prior to the implementation of the component detection method, learning is performed using a machine learning (pattern recognition) technique. Specifically, first, similarly to the first and second embodiments, the identification target area teacher data 141 is assigned to the identification target area (head) of the learning image.

  Next, in the same manner as in the first and second embodiments, the part teacher data 142 is applied to the face parts (left eye, right eye, nose and mouth) included in the identification target area (head) to which the identification target area teacher data 141 is assigned. Give.

  Next, after selecting a reference face part (for example, left eye and right eye) as a reference for alignment from the face parts, the reference face part is fixed at a predetermined position, and alignment teacher data is acquired. The reference face part may be at least one selected from the face parts, and may be, for example, a nose or a mouth.

  Next, attribute recognition teacher data 143 is acquired by assigning an attribute to the alignment teacher data. The attributes and the method for giving them are the same as those in the second embodiment.

  Next, as shown in FIGS. 3A and 3B, the identification target image is acquired by the identification target image acquisition unit 111 in the same manner as in the first and second embodiments (step S11).

  Next, as shown in FIGS. 3A and 3B, similarly to Embodiments 1 and 2, the identification target area detection unit 121 creates the identification target area teacher data 141 acquired in advance. With reference to the identification target region extraction model, identification target region data is detected from the identification target image (step S21).

  Next, as shown in FIGS. 3A and 3B, in the same manner as in the first and second embodiments, a component extraction model created from the component teacher data 142 acquired in advance by the component detector 122 is obtained. Referring to the facial part data included in the identification target area, the data is detected (step S22).

  Next, as shown in FIGS. 3A and 3B, after selecting a reference face part (for example, left eye and right eye) as a reference for alignment from the face part, the reference face part is moved to a predetermined position. To create an alignment image (step S24). The reference face part may be at least one selected from the face parts, and may be, for example, a nose or a mouth. In consideration of detection accuracy and detection speed, it is preferable to create the alignment image under the same conditions as the alignment teacher data acquired in advance.

  Next, as shown in FIGS. 3A and 3B, the attribute of the face part is recognized with reference to the attribute extraction model created from the attribute recognition teacher data 143 acquired in advance (step S23). ). The attribute is the same as that given to the alignment teacher data.

  Next, as shown in FIGS. 3A and 3B, the output unit 131 outputs at least the face part data among the identification target area data, the face part data, and the attribute (step S31). ). The output identification target area data, face part data, and attributes are the same as those in the first and second embodiments. The output step S31 is an optional step and may not be included in the component detection method of the present invention, but is preferably included.

  As described above, in the component detection apparatus and component detection method of the present embodiment, prior to the implementation, alignment (alignment) is performed on the identification target region teacher data acquired using the machine learning (pattern recognition) technique. After that, attribute recognition teacher data to which the attribute is given is acquired in advance, and the attribute is recognized with reference to the attribute extraction model created therefrom. For this reason, according to the component detection apparatus and component detection method of this embodiment, an attribute can be recognized more accurately and rapidly.

(Embodiment 4)
The program of this embodiment is a program that can execute the above-described component detection method on a computer. Or the program of this embodiment may be recorded on a recording medium, for example. The recording medium is not particularly limited, and examples thereof include a read only memory (ROM), a hard disk (HD), an optical disk, a floppy (registered trademark) disk (FD), and the like.

(Embodiment 5)
FIG. 8 shows a configuration of an example of a component detection system using the component detection apparatus of the present invention. As illustrated, the component detection system includes identification target image acquisition units 111a, 111b, and 111c, output units 131a, 131b, and 131c, communication interfaces 150a, 150b, and 150c, and a server 170. The identification target image acquisition unit 111a and the output unit 131a are connected to a communication interface 150a. The identification target image acquisition unit 111a, the output unit 131a, and the communication interface 150a are installed in the place X. The identification target image acquisition unit 111b and the output unit 131b are connected to a communication interface 150b. The identification target image acquisition unit 111b, the output unit 131b, and the communication interface 150b are installed in the place Y. The identification target image obtaining unit 111c and the output unit 131c are connected to a communication interface 150c. The identification target image obtaining unit 111c, the output unit 131c, and the communication interface 150c are installed in the place Z. The communication interfaces 150a, 150b, and 150c are connected to the server 170 via the line network 160.

  In this component detection system, the server 170 side has an identification target region acquisition unit and a component detection unit, and the identification target region teacher data and the component teacher data are stored in the server 170. For example, the identification target image acquired by using the identification target image acquisition unit 111a at the location X is transmitted to the server 170, and the identification target region and parts are detected on the server 170 side. In addition, at least the component of the detected identification target area and component is output by the output unit 131a.

  According to the component detection system of the present embodiment, the identification target image acquisition unit and the output unit are installed in the field, and the server or the like is installed in another place, so that the identification target region and the component can be detected online. . Therefore, the installation of the apparatus does not take a place, and maintenance is easy. Further, even when the installation locations are separated, centralized management and remote operation can be performed at one location. The partial detection system of the present embodiment may correspond to the attribute recognition of the second embodiment and the alignment (alignment) of the third embodiment.

10 Learning images 11, 12a, 13a, 14a, 15a, 12b, 13b, 14b, 15b, 22a, 23a, 24a, 25a, 22b, 23b, 24b, 25b Labeling 20 Discrimination target image 111 Discrimination target image acquisition means 120 Calculation Means 121 Identification target area detection means 122 Component detection means 123 Attribute recognition means 124 Alignment means 131 Output means 140 Data storage means 141 Identification target area teacher data 142 Part teacher data 143 Attribute recognition teacher data

Claims (16)

An identification target image acquisition means for acquiring an identification target image;
An identification target area detecting means for detecting an identification target area from the identification target image with reference to an identification target area extraction model created from the identification target area teacher data acquired in advance;
A component detection apparatus comprising: a component detection unit configured to detect a component included in the identification target region with reference to a component extraction model created from component teacher data acquired in advance. further,
2. The component detection apparatus according to claim 1, further comprising attribute recognition means for recognizing an attribute of the component by referring to an attribute extraction model created from attribute recognition teacher data acquired in advance. further,
After selecting a reference part as a reference for alignment from the parts, including an alignment means for creating an alignment image by fixing the reference part at a predetermined position,
The component detection apparatus according to claim 2, wherein the attribute of the component is recognized by the attribute recognition unit using the alignment image. The component detection apparatus according to claim 3, wherein the alignment image is created under the same conditions as the alignment teacher data acquired in advance. The component detection apparatus according to claim 1, wherein the identification target region is a person. The component detection apparatus according to claim 1, wherein the identification target region is a human head. The component detection apparatus according to claim 6, wherein the component includes at least one selected from the group consisting of a left eye, a right eye, a nose, a mouth, a left eyebrow, a right eyebrow, a left ear, and a right ear. An identification target image acquisition step of acquiring an identification target image;
An identification target area detection step of detecting an identification target area from the identification target image with reference to an identification target area extraction model created from the identification target area teacher data acquired in advance;
A component detection method comprising: detecting a component included in the identification target region with reference to a component extraction model created from component teacher data acquired in advance. further,
9. The component detection method according to claim 8, further comprising an attribute recognition step of recognizing the attribute of the component with reference to an attribute extraction model created from attribute recognition teacher data acquired in advance. further,
After selecting a reference component that is a reference for alignment from the component, including an alignment step of fixing the reference component at a predetermined position and creating an alignment image,
The component detection method according to claim 9, wherein the attribute of the component is recognized in the attribute recognition step using the alignment image. The component detection method according to claim 10, wherein the alignment image is created under the same conditions as the alignment teacher data acquired in advance. The component detection method according to claim 8, wherein the identification target region is a person. The component detection method according to claim 8, wherein the identification target region is a human head. The component detection method according to claim 13, wherein the component includes at least one selected from the group consisting of a left eye, a right eye, a nose, a mouth, a left eyebrow, a right eyebrow, a left ear, and a right ear. A program for causing a computer to execute the component detection method according to any one of claims 8 to 14. A recording medium on which the program according to claim 15 is recorded.

Images