JP2015001871A - Person determination device, person determination method, program and recording medium - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a person determination device capable of, even when a plurality of persons who have not been determinable until that time are present at near distances, or the positions of the plurality of persons are switched, determining whether or not the persons are identical.SOLUTION: A person determination device 100 includes: image acquisition means 101 for acquiring a determination object person image; head/face detection means 102 for detecting at least one of a head area and a face area from the determination object person image; person attribute analysis means 103 for detecting person attributes from the determination object person image; person information storage means 104 for storing the information of the head area, the face area and the person attributes; and person determination means 106 for determining whether or not the person photographed in the determination object person image is identical to the person photographed in the image previously acquired by the image acquisition means 101 by using the person information acquired by the person information storage means 104.

Description

  The present invention relates to a person determination device, a person determination method, a program, and a recording medium.

  A person determination technique for extracting and tracking a specific person from an image is generally popular. For example, in Patent Document 1, a person appearing in a plurality of images arranged in time series is the same person from the amount of change in the position and size of the head measured from an image taken by a monitoring camera provided in a store or the like. There has been proposed a monitoring system for determining whether or not. Also, as described in Patent Document 2, a method for determining a person's attribute from an image taken by a camera or the like is known.

Japanese Patent No. 4939968 International Publication No. 2012/053311

  However, in the monitoring system described in Patent Document 1, when a plurality of persons are present at a short distance, it is determined whether or not each person shown in two images taken in a short time is the same person. In the case where only the position and size of the head of each person in the preceding and following images is used, the same person is different from the same person or different persons are the same when a plurality of persons are close to each other or when the positions are switched. There is a possibility of being mistaken for a person. Further, since the method described in Patent Document 2 is a method for determining the attributes of a person, it cannot be determined whether or not the persons shown in two different images are the same person.

  Therefore, the present invention provides a person determination device, a person, and a person who can determine whether or not they are the same person even when a plurality of persons that could not be determined so far are at close distances or when the positions of a plurality of persons are switched An object is to provide a determination method, a program, and a recording medium.

In order to achieve the above object, the person determination device of the present invention provides:
Image acquisition means for acquiring a determination target person image;
A head / face detection means for detecting at least one of a head region and a face region from the determination target person image;
A person attribute analyzing means for detecting a person attribute from the determination target person image;
Person information storage means for storing information on the head region, the face region and the person attribute;
A person determination that uses the person information acquired from the person information storage means to determine whether the person shown in the determination target person image is the same person as the person shown in the image previously acquired by the image acquisition means Means.

Moreover, the person determination method of the present invention includes:
An image acquisition step of acquiring a determination target person image;
A head / face detection step of detecting at least one of a head region and a face region from the determination target person image;
A person attribute analysis step of detecting a person attribute from the determination target person image;
A person information storing step for storing information on the head region, the face region, and the person attribute;
A person who uses the person information stored in the person information storing step to determine whether the person shown in the determination target person image is the same person as the person shown in the image acquired earlier in the image acquisition step And a determination step.

The program of the present invention is
A person determination method of the present invention is executed by a computer.

The recording medium of the present invention is
The program of the present invention is recorded.

  The present invention relates to a person determination device and a person determination method capable of determining whether or not a plurality of persons that could not be determined so far are close or when the positions of a plurality of persons are switched. A program and a recording medium can be provided.

FIG. 1 is a block diagram illustrating a configuration of an example of a person determination device according to the present invention. FIG. 2 is a flowchart showing an example of the person determination method of the present invention. FIG. 3 is a diagram for explaining an example of head / face detection in the present invention. FIG. 4 is a diagram illustrating another example of head / face detection in the present invention.

  Next, the person determination device, the person determination method, the program, and the recording medium of the present invention will be described with examples. However, the present invention is not limited to the following examples.

  FIG. 1 is a block diagram showing a configuration of an example of a person determination device according to the present invention. As shown in FIG. 1, the person determination apparatus 100 includes an image acquisition unit 101, a head / face detection unit 102, a person attribute analysis unit 103, a person information storage unit 104, and a person determination unit 106. It is included as a component.

  The image acquisition unit 101 acquires a determination target person image. Examples of the image acquisition unit 101 include a CCD (Charge Coupled Device) camera, a CMOS (Complementary Metal Oxide Semiconductor) camera, and an image scanner.

  The head / face detection unit 102 detects at least one of a head region and a face region from the determination target person image. The detection of the head region and the face region will be described with reference to FIGS. FIG. 3 shows an example in which the determination target person image 30 includes two persons, a male whose top is reflected from the chest facing the left side and a male whose whole body is captured facing the front side on the right side. Labeling 31a and 31b are attached to the head region of the determination target person image 30, and the head region (for example, the coordinates and size of the head region) is detected. Also, labeling 32a and 32b are attached to the face area of the determination target person image 30, and the face area (for example, the coordinates and size of the face area) is detected. FIG. 4 shows an example in which the determination target person image 40 includes two persons, a male whose left side is facing backwards and a male who is wearing the right hat and whose whole body is facing forward. It is. A labeling 41c is attached to the head region of the determination target person image 40, and a head region (for example, coordinates and size of the head region) is detected. Further, a labeling 42d is attached to the face area of the determination target person image 40 to detect the face area (for example, the coordinates and size of the face area). The method for detecting the head region is not particularly limited. For example, a template (mask pattern) indicating the outline of a general head of a person is prepared, and this template is shifted and matched with a determination target person image to determine the head region. Various known head region detection algorithms, such as template matching that detects the head region present in the target person image, and detection using the feature information of the head region such as brightness, hue, shape, and ear of the target person image (Image processing technology) can be used. Also, the detection method of the face area is not particularly limited, for example, by preparing a template (mask pattern) indicating the outline of a general face area of a person, and shifting and matching the template to the determination target person image, Various known face area detection algorithms (images) such as template matching for detecting a face area existing in a determination target person image, detection using feature information of the face area such as brightness, hue, shape, and eyes of the determination target person image Processing technology). In the present invention, it is preferable to detect both the head region and the face region, but only one of the head region and the face region may be detected.

  The person attribute analysis means 103 detects a person attribute from the determination target person image. In the present invention, the personal attributes include, for example, gender, age, race, hair length, presence / absence of a hat, hair color, hair style, presence / absence of glasses or sunglasses, presence / absence of a mask, presence / absence of wrinkles, etc. It is done. The method for detecting a person attribute is not particularly limited. For example, a person attribute is detected from the determination target person image using a person attribute determination model created by machine learning (pattern recognition). When the person attribute is gender, for example, it can be detected based on gender degree (for example, 0 to 1). The gender degree can be calculated based on the determination target person image. For example, if the gender degree is “0 to less than 0.5”, the standard is set to detect “female”, and if the gender degree is “0.5 to 1”, the standard is to detect “male”. In addition, the gender is detected from the calculated gender degree value. Note that the criteria for determining gender from the gender degree are not limited to this example. For example, if the gender degree is “0 to 0.3”, “female” is detected, and the gender degree is “0. 7 to 1 may be detected as “male”, and if the gender degree is “greater than 0.3 and less than 0.7”, “gender unknown” may be used. As for the age and the like, for example, a predetermined standard is set, and the age and the like are determined from values calculated based on the determination target person image.

  The person information storage means 104 stores the head area, the face area, and the person attribute information (person information). Examples of the person information storage means 104 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 person information storage unit 104 may be, for example, a device built-in type or an external type such as an external storage device.

  The person determination unit 106 uses the person information acquired from the person information storage unit 104 to determine whether the person shown in the determination target person image is the same person as the person previously acquired by the image acquisition unit 101. . The determination method is not particularly limited, and examples thereof include the following method. That is, first, a threshold value for determining whether or not they are the same person is set as individual information stored in the person information storage means 104 (for example, the ordinate / abscissa / size of the head region, gender of the person, Age, presence / absence of hat, presence / absence of glasses or sunglasses, presence / absence of mask, etc.). Further, of the person information stored in the person information storage unit 104, the number of person information that must be equal to or lower than the threshold and the number of person information that should be equal to or lower than the threshold are also determined in advance. If the amount of change of individual information is obtained and the amount of change is less than or equal to the threshold, it is determined that the condition is satisfied for the person information, and the number of pieces of person information satisfying the condition is greater than or equal to a predetermined number. If there is a condition that satisfies the condition for the person information that is required to be equal to or less than the threshold, it is determined that the person is the same person, and if not, the person information is determined not to be the same person (different person). To do.

  As shown in FIG. 1, the person determination device 100 of the present invention preferably further includes a person information acquisition unit 105. The person information acquisition unit 105 sends the determination target person image from the image acquisition unit 101 to the head / face detection unit 102 and the person attribute analysis unit 103. However, the person information acquisition unit 105 is an arbitrary component, and the person determination apparatus 100 of the present invention may not include the person information acquisition unit 105.

  Although not shown, it is preferable that the person determination device 100 of the present invention further includes an identification number assigning unit. When the person determination means 106 determines that the person is the same person, the identification number assigning means assigns the same identification number as the person shown in the previously acquired image to the person shown in the determination target person image. When the determination unit 106 determines that the person is different, a new identification number is assigned to the person shown in the determination target person image. When the person determination device 100 of the present invention includes the identification number giving means, the person information storage means 104 records the identification number. In a retail store, information on how a person who visits the store moves through the store, sees the product, and decides to purchase can be important information for the store to determine the product arrangement and the sales strategy. However, in order to collect this information, it is necessary to track a person, but it is almost impossible to track a large number of visitors one by one with a limited number of shop assistants. On the other hand, by assigning the same identification number when it is determined that they are the same person, it is possible to track the same person from a plurality of images acquired by the image acquisition means 101 over time. In recent years, many stores have installed surveillance cameras for the purpose of crime prevention. Therefore, if a person can be determined from an image captured by the monitoring camera, it is possible to collect such information. However, the identification number giving means is an arbitrary constituent member, and the person determination device 100 of the present invention may not include the identification number giving means.

  As shown in FIG. 1, in the person determination device 100 of the present invention, it is preferable that the head / face detection means 102 and the person attribute analysis means 103 are in the same server 107. As shown in FIG. 1, in the person determination device 100 of the present invention, it is preferable that the person information acquisition unit 105 and the person determination unit 106 are in the same server 108. However, in the person determination apparatus 100 of the present invention, each unit does not necessarily have to be in the same server, and the person image to be determined by a wired or wireless communication line network such as a LAN (Local Area Network), the Internet, or a cloud system. And personal information can be transmitted and received.

  Next, a person determination method using the person determination apparatus 100 will be described with reference to FIG.

  First, the person information acquisition unit 105 acquires a determination target person image through the image acquisition unit 101 (step 201 (S201)).

  Next, the person information acquisition unit 105 sends the determination target person image to the head / face detection unit 102 and the person attribute analysis unit 103 (step 202 (S202)).

  Next, the head / face detection means 102 determines a head region (for example, coordinates and size of the head region) and a face region (for example, coordinates and size of the face region) from the determination target person image. At least one of them is detected (step 203 (S203)).

  Next, the person attribute analysis means 103 detects person attributes (for example, the gender of the person, the age, the presence / absence of a hat, the presence / absence of glasses or sunglasses, the presence / absence of a mask, etc.) from the determination target person image (step 204 ( S204)).

  In FIG. 2, detection by the person attribute analysis unit 103 is performed after detection by the head / face detection unit 102, but the person determination method of the present invention is not limited to this. In the person determination method of the present invention, detection by the head / face detection unit 102 may be performed after detection by the person attribute analysis unit 103, or both may be performed simultaneously.

  Next, the detected head area, face area, and person attribute information (person information) is stored in the person information storage means 104 (step 205 (S205)).

  Next, the person determination unit 106 compares the person in the determination target person image to determine whether or not the person in the determination target person image is the same person as the person in the image previously acquired by the image acquisition unit 101. Information is acquired from the person information storage means 104 (step 206 (S206)).

  Next, the person determination means 106 calculates the amount of change of each person information (step 207 (S207)).

  Next, the person determination means 106 compares the amount of change of each person information with a threshold set in advance for each person information (step 208 (S208)).

  Next, the person determination unit 106 determines whether the number of pieces of person information whose amount of change is equal to or less than the threshold is greater than a predetermined set value and whether the amount of change of the required person information is equal to or less than the threshold (step 209). (S209)) If the condition is satisfied (Yes), the process proceeds to step 210 (S210), and the identification number assigning means is reflected in the previously acquired image on the person reflected in the determination target person image. If the same identification number as that of the person is assigned and the condition is not satisfied (No), the process proceeds to step 211 (S211), and a new identification number is assigned to the person shown in the determination target person image. The person information storage means 104 records the identification number.

  Thereafter, the same processing is repeated for the newly acquired determination target person image. At this time, the identification number assigned earlier is used to determine whether the newly acquired determination target person image and the person shown in the previously determined image are the same person. However, in determining whether or not they are the same person, only the identification number of the image determined immediately before is used, and the identification number assigned to the image determined two or more times before is not used. However, it is possible to collect the identification numbers after performing person determination on all images obtained by shooting a series of actions and use them to track a series of movements of the person during the shooting period.

  According to the person determination device and the person determination method of the present invention, the head region and the face region detected from the determination target person image when determining whether or not the persons shown in different images acquired by the image acquisition unit are the same person. In addition to at least one of the above, by using the person attribute, even when a plurality of persons are close to each other or the positions of a plurality of persons are switched, which cannot be determined only by the head area and the face area It is possible to determine whether or not they are the same person.

  In the above-described example, the determination target person image is an image acquired by one image acquisition unit 101 over time. However, according to the person determination apparatus and the person determination method of the present invention, the determination target person image is acquired by a different image acquisition unit. It is also possible to perform person determination using a plurality of images.

  The program of the present invention is a program capable of executing the person determination method described above on a computer. The program of the present invention may be recorded on a recording medium, for example. The recording medium is not particularly limited, and examples thereof 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 person determination device, the person determination method, the program, and the recording medium of the present invention can be widely used, for example, when performing a movement route, a behavior analysis, or the like of a person in a limited space such as a store.

30, 40 Determination target person images 31a, 31b, 32a, 32b, 41c, 42d Labeling 100 Person determination apparatus 101 Image acquisition means 102 Head / face detection means 103 Person attribute analysis means 104 Person information storage means 105 Person information acquisition means 106 Person determination means 107, 108 server

Claims (10)

Image acquisition means for acquiring a determination target person image;
A head / face detection means for detecting at least one of a head region and a face region from the determination target person image;
A person attribute analyzing means for detecting a person attribute from the determination target person image;
Person information storage means for storing information on the head region, the face region and the person attribute;
A person determination that uses the person information acquired from the person information storage means to determine whether the person shown in the determination target person image is the same person as the person shown in the image previously acquired by the image acquisition means Means for determining a person. Furthermore, it includes person information acquisition means,
The person determination apparatus according to claim 1, wherein the person information acquisition unit sends the determination target person image from the image acquisition unit to the head / face detection unit and the person attribute analysis unit. Furthermore, including an identification number giving means,
The identification number giving means is
When the person determination unit determines that the person is the same person, the person identified in the determination target person image is assigned the same identification number as the person reflected in the previously acquired image.
When the person determination means determines that the person is different, a new identification number is assigned to the person shown in the determination target person image,
The person determination apparatus according to claim 1, wherein the person information storage unit records the identification number. The person determination apparatus according to any one of claims 1 to 3, wherein the head / face detection unit and the person attribute analysis unit are in the same server. 5. The person determination device according to claim 2, wherein the person information acquisition unit and the person determination unit are in the same server. An image acquisition step of acquiring a determination target person image;
A head / face detection step of detecting at least one of a head region and a face region from the determination person target image;
A person attribute analysis step of detecting a person attribute from the determination target person image;
A person information storing step for storing information on the head region, the face region, and the person attribute;
A person who uses the person information stored in the person information storing step to determine whether the person shown in the determination target person image is the same person as the person shown in the image acquired earlier in the image acquisition step A person determination method comprising: a determination step. Furthermore, a person information acquisition step is included,
In the person information acquisition step,
The person determination method according to claim 6, wherein the determination target person image is sent to the head / face detection step and the person attribute analysis step. Furthermore, an identification number assigning step is included,
In the identification number assigning step,
When it is determined that the person is the same person in the person determination step, the same identification number as the person shown in the previously acquired image is given to the person shown in the determination target person image,
When it is determined that the person is different in the person determination step, a new identification number is given to the person shown in the determination target person image,
8. The person determination method according to claim 6, wherein the identification number is recorded in the person information storing step. A program for causing a computer to execute the person determination method according to any one of claims 6 to 8. A recording medium on which the program according to claim 9 is recorded.

Images