JP2020095756A - Identification device - Google Patents

Abstract

To provide an identification device capable of identifying the posture of a person.SOLUTION: An identification device 100 includes: an input part 101 that receives a piece of image information of a person picked up by a camera; and a control unit 102 that identifies the person on the basis of the image information and detects at least head and hands of the person, and identifies the motion of the person on the basis of the identified person, the detected portion, and a behavior model in which the behavior of the person is registered for each person. The motion model includes a first motion model that is generated on the basis of the motions of the person identified by the control unit and a second motion model which is pre-generated and stored.SELECTED DRAWING: Figure 5

Description

The present invention relates to an identification device.

In recent years, a technique for identifying the posture of a person is known.

For example, in Patent Document 1, a device that detects the positions of the feature points of a person's hand, elbows and shoulders from the image of a camera and corrects the positions of the feature points of the hand, elbows and shoulders so as to match a joint model created in advance. Is disclosed.

JP, 2010-211705, A

However, in the device described in Patent Document 1, when a person wears clothes such as a leather jacket having a low reflectance for near red light, the pixel values of the body/arm portions cannot be normally obtained, and the shoulders and elbows are not obtained. There is a problem that the posture of the person cannot be identified because the joint of the person cannot be detected.

An object of the present invention is to provide an identification device that can identify the posture of a person.

An identification device according to an aspect of the present invention includes an input unit that receives image information of a person photographed by a camera, identification of the person based on the image information, and detection of at least the head and hand parts of the person. The identified person, the detected part, and a control unit that identifies the movement of the person based on a movement model in which the movement of the person is registered for each person; It includes a first motion model generated based on the motion of the person identified by the control unit, and a second motion model generated and stored in advance.

The “person” here is not limited to the meaning of a specific individual, but also a general person of any age, sex, and physique.

According to the present invention, the posture of a person can be identified.

Block diagram showing an example of the configuration of an identification device according to an embodiment of the present invention Block diagram showing an example of the configuration of the control unit Explanatory drawing showing an example of skeleton detection processing Explanatory drawing showing an example of behavior model Explanatory drawing which shows a mode that person information and part information are compared with a motion model in the embodiment of the invention. Flow chart explaining the main operation of the identification device according to an embodiment of the present invention Explanatory drawing which shows a mode that person information and site|part information are compared with a motion model in the modification 2. The figure which shows an example of the hardware constitutions of the computer which implement|achieves the function of each part by the program.

Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings. However, in each embodiment, configurations having the same function are denoted by the same reference numerals, and duplicate description will be omitted.

First, an example of the configuration of the identification device 100 according to the embodiment of the present invention will be described with reference to FIG. 1a. FIG. 1a is a block diagram showing an example of the configuration of the identification device 100 according to the embodiment of the present invention.

The identification device 100, the camera 200, and the device 300 illustrated in FIG. 1a are mounted on a moving body such as an automobile, a railroad vehicle, a ship, or an aircraft. In the present embodiment, a case where the identification device 100, the camera 200, and the device 300 are installed in an automobile will be described as an example.

The camera 200 is, for example, a distance-measuring camera such as a stereo camera or a TOF (Time Of Flight) type camera. The camera 200 captures image information by capturing a range in which a person is present in the vehicle compartment, and the image information is identified by the identification device. Output to 100. The image information includes at least a near-red image that represents the intensity of the reflection intensity of near-infrared light. In the case of a stereo camera or a TOF camera, it may further include a range image showing the distance.

The device 300 is a device used in an automobile. The device 300 is, for example, a car navigation device, an audio device, an air conditioner device, a tablet, a smartphone, a rear camera, a side camera, a front camera, an ETC vehicle-mounted device. Alternatively, the device 300 includes, for example, a handle, a shift lever (gear lever), a winker lever (winker switch), a wiper lever (wiper switch), a door lock switch, a power window up/down switch, an engine start switch, an accelerator pedal, a brake pedal, It may be a clutch pedal or the like.

The device 300 is connected to the identification device 100 via a wired network (for example, CAN: Controller Area Network), a wireless network (for example, wireless LAN: Local Area Network), or a network in which a wired network and a wireless network are mixed. .. Then, the device 300 is controlled based on the control signal received from the identification device 100.

The identification device 100 includes an input unit 101 and a control unit 102.

The input unit 101 receives the image information of a person from the camera 200 and outputs it to the control unit 102.

The control unit 102 identifies the person based on the image information received from the input unit 101 and detects at least the head and hand parts of the person, and identifies the identified person, the detected part, and the person. The action of the person is identified based on the action model registered for each person. With such a configuration, the posture of the person can be identified.

Next, an example of the configuration of the control unit 102 will be described with reference to FIG. FIG. 1b is a block diagram showing an example of the configuration of the control unit.

As shown in FIG. 1b, the control unit 102 includes a detection control unit 110, a skeleton detection unit 120, a skeleton motion identification unit 130, a person identification unit 140, a motion model unit 150, a site detection unit 160, a site motion identification unit 170, and a device. The controller 180 is provided.

The detection control unit 110 receives image information from the camera 200, analyzes the image information, and outputs the image information to the skeleton detection unit 120, the person identification unit 140, and the part detection unit 160. Specifically, the detection control unit 110 determines whether or not skeleton detection is possible by analyzing image information, and when it is determined that skeleton detection is possible, the skeleton detection unit 120 and the person identification unit 140 display images. Output information. On the other hand, if it is determined that skeleton detection is not possible, it is further determined whether or not a motion model to be described later is available. If it is determined that the motion model is available, the person identifying unit 140 and the part detection unit. The image information is output to 160. The determination as to whether skeleton detection is possible will be described later.

The skeleton detection unit 120 detects the skeleton position of the person based on the image information received from the detection control unit 110, and outputs the skeleton information indicating the detection result to the skeleton motion identification unit 130. The skeleton position of a person refers to the position of a representative point representing a region such as the head, neck, waist, shoulders, elbows, and hands. The position of the representative point is specified as a two-dimensional coordinate on the image.

FIG. 2 shows an example of the skeleton detection process. An image 1000 is an example of an image of a person, and is a schematic view of an image of a person driving an automobile holding a steering wheel. An image 1001 is an example of an image obtained by detecting the skeleton position of a person in the image 1000, and the positions of nine points that are the representative points of the head, neck, waist, left shoulder, left elbow, left hand, right shoulder, right elbow, and right hand are shown. The detected state is shown. The skeleton information is the position coordinates of the detected representative points, and is obtained as the position coordinates (X1, Y1), (X2, Y2)... (X9, Y9) of the nine representative points on the image. Skeleton detection can be realized by using an existing method as disclosed in Patent Document 1, for example.

Further, in the image 1001, the positions of nine points, which are the representative points of the head, neck, waist, left shoulder, left elbow, left hand, right shoulder, right elbow, and right hand, can be detected. It is determined that detection is possible. On the other hand, an image 1002 shows an example of a result image obtained by detecting the skeleton position of a person obtained when the photographed person is wearing clothes such as a leather jacket having a low reflectance for near red light. In the image 1002, it can be seen that only the positions of the three representative points of the head, left hand, and right hand can be detected. In this case, the detection control unit 110 determines that skeleton detection is not possible.

The skeleton motion identification unit 130 identifies the occupant's motion based on the skeleton information received from the skeleton detection unit 120. The operation of the occupant is, for example, an operation of operating a device such as an air conditioner device, a steering wheel, or an audio device, a search for an object while driving, a gesture of turning and the like.

The operation is identified based on the skeleton information by using an existing method, for example, a so-called random forest method that specifies one of the predetermined patterns from multiple size comparisons according to the tree structure of the comparison target parameter. be able to.

The skeleton motion identification unit 130 outputs, to the motion model unit 150, motion information and skeleton information indicating the above-described occupant motion identification result. Further, the skeleton motion identification unit 130 outputs the motion information to the device control unit 180.

The person identifying unit 140 identifies who the person is based on the image information received from the detection control unit 110, and outputs the person information indicating the identified result to the action model unit 150 and the region action identifying unit 170. To do.

To realize the person identification, it is possible to use the existing face recognition technology, for example. For example, face images of a plurality of persons who drive a car are registered in advance, face images are acquired when those persons actually drive the car, and the acquired face image and the registered face images are registered using existing technology. By matching, it is possible to identify which registrant. As an existing technology for face identification, there is a method of acquiring a feature amount of a face by a method such as LBP (Local Binary Pattern) or HOG (Histogram of Oriented Gradient) and collating the feature amount.

The motion model unit 150 generates and stores a motion model based on the motion information from the skeleton motion identifying unit 130, the skeleton information, and the person information from the person identifying unit 140, and the stored motion model is used as a part motion. Output to the identification unit 170.

The motion model is stored in a database in which personal information (registration number corresponding to a registered person), skeleton information (positional coordinates of skeleton feature points), and motion information (motion number corresponding to motion type) are stored in association with each other. Form. Here, as the skeleton information, instead of registering all the feature point information output from the skeleton detection unit 120, the head, the left hand, and the left hand, which are detection targets in the site detection processing by the site detection unit 160, which will be described later. Only the information of the feature points on the right hand is registered. As an example, FIG. 3 shows a relational database 1003 of person information, head/left/right hand skeleton information, and motion information.

In the present embodiment, the motion model database becomes available after a certain number or more of skeleton information, person information, and motion information detected in a state where skeleton detection by the skeleton detection unit 120 is possible. For example, in order to be able to use a database of motion models, it is necessary to store 10,000 or more combinations of person information, skeleton information, and motion information.

The behavior model unit 150 outputs state information indicating whether or not the behavior model is available to the detection control unit 110. The detection control unit 110 branches the process depending on whether the behavior model is available.

The part detection unit 160 detects a head position, a left hand position, and a right hand position as the part information of the person based on the image information received from the detection control unit 110, and outputs the detection result part information to the part motion identification unit 170. Output.

The skeleton detection unit 120 described above also detects the position coordinates of a plurality of feature points including the head position, the left-hand position, and the right-hand position, but here, when the image information of the human body part cannot be obtained normally. The detection method in is used.

For example, as a method of realizing the detection of body part information by combining existing methods, the distance information included in the image information is used to detect a group of distance information existing at equal distances, and the position of the group is tracked between a plurality of frames. It is possible to use the method.

Partial movement identification unit 170 identifies the movement of the occupant based on the person information from person identification unit 140, the movement model from movement model unit 150, and the portion information from portion detection unit 160.

A method of part movement identification processing for identifying the movement of the occupant in part movement identification section 170 will be described with reference to FIG.

In the body part movement identification process, the person information and the body part information in the input information 2001 of the body part movement identification unit 170 are compared with the person information and the skeleton information in the registration information 2002 already registered as the motion model, and the registration information 2002 is selected. , Data in which the person information matches the person information in the input information and the skeleton information is closest to the part information in the input information is searched.

As a method of retrieving the registered information having the skeleton information closest to the part information of the input information, for example, the following method can be adopted. First, distances between corresponding parts (for example, head positions, left-hand positions, right-hand positions) between the input information and the registration information are calculated. Then, the registration information having the smallest sum of the calculated distances between the respective parts is determined to be the registration information having the skeleton information closest to the part information of the input information.

When matching the input information and the registered information, considering the identification accuracy of the motion, the data in which the person information matches the person information in the input information and the skeleton information matches the part information in the input information are searched for. Is ideal. However, since the movement of a person when performing a predetermined motion is not always the same, it is conceivable that there is no skeleton information in the registration information that matches the part information in the input information.

In consideration of such a case, in the present embodiment, when matching the input information and the registered information, the person information matches the person information in the input information, and the skeleton information is the site information in the input information. To find the data closest to.

In the case of the example of FIG. 4, since the input information 2001 is the closest to the data of 2003 in the registration information 2002, the part motion identification unit 170 identifies that the motion information is “6”, and outputs the part motion information. It is output to the device control unit 180.

The device control unit 180 generates control information for controlling the device 300 (not shown) based on the motion information from the skeleton motion identification unit 130 or the motion information from the part motion identification unit 170, and outputs the generated control information. Output to the device 300.

The device 300 is controlled based on the control information received from the device control unit 180.

Although the control unit 102 includes the skeleton detection unit 120 and the skeleton motion identification unit 130 in the example of FIG. 1b, the control unit 102 may include the skeleton detection unit 120 and the skeleton motion identification unit 130. Good. Details will be described later.

In the example of FIG. 1b, the control unit 102 includes the device control unit 180, but the device control unit 180 may be provided separately from the control unit 102.

That is, the control unit 102 may be configured to include at least the detection control unit 110, the person identifying unit 140, the motion model unit 150, the part detecting unit 160, and the part motion identifying unit 170.

Next, an example of the operation of the control unit 102 according to this embodiment will be described with reference to FIG. FIG. 5 is a flowchart showing an example of the operation flow of the control unit 102 according to the present embodiment.

First, in step S101, the detection control unit 110 receives image information from the camera 200. As described above, the image information is an image of the occupant.

Subsequently, in step S102, the detection control unit 110 determines whether the image is a skeleton-detectable image based on the image information. Specifically, when all the representative point positions to be detected from the image information can be detected, it is determined that the image can be detected as a skeleton, and one of the representative point positions to be detected from the image information is detected. When only the representative point position of the part can be detected, it is determined that the image is not a skeleton-detectable image.

If it is determined in step S102 that the skeleton can be detected (step S102: YES), the process proceeds to step S103, and in step S103, the skeleton detection unit 120 performs skeleton detection processing.

Subsequently, in step S104, the person identifying unit 140 performs person identifying processing, in step S105, the skeleton action identifying unit 130 performs action identifying processing, and in step S106, the device control unit 180 causes the device control information generation. Perform processing.

In step S107 following step S106, based on the skeleton information identified in the skeleton detection process of step S103, the person information identified in the person identification process of step S104, and the action information identified in the action identification process of step S105, The behavior model unit 150 updates the behavior model. Specifically, the work of adding the obtained skeleton information, person information, and motion information to the motion model database is performed. The order of performing step S106 and step S107 may be reversed, and step S106 and step S107 may be performed simultaneously.

If it is not determined in step S102 that the skeleton can be detected (step S102: NO), the process proceeds to step S108, and in step S108, the detection control unit 110 determines whether the motion model is available.

When it is determined in step S108 that the motion model is available (step S108: YES), the process proceeds to step S109, and in step S109, the part detection unit 160 performs a part detection process.

Subsequently, in step S110, the person identifying unit 140 performs a person identifying process.

In step S111 following step S110, the part motion identification unit 170 is based on the part information identified in the part detection process of step S109, the person information identified in the person identification process of step S110, and the pre-stored motion model. , The part motion identification process is performed.

In step S112 following step S111, the device control section 180 performs a device control information generation process.

When it is not determined that the motion model is available in step S108 (step S108: NO), the process proceeds to step S113, and in step S113, the detection control unit 110 determines that the motion identification process is impossible. The execution of the motion identification process is prohibited, error information is presented, and the process ends.

As described above, in the identification device 100 according to the present embodiment, the person is identified and at least the head and hand parts of the person are detected based on the image information received from the input unit 101, and the person is identified. The motion of the person is identified based on the motion model registered for each person, the detected part, and the motion of the person.

Therefore, if a person wears leather jackets or other clothing with low reflectance to near-infrared light, the distance between the body and arms is not output normally, and the shoulder and elbow joints cannot be detected, only the site information Although there is a possibility that the motion identification may be erroneous, the identification device 100 of the present embodiment can correctly identify the motion of the person even in such a case.

In addition, for example, even if the positions of the head, left hand, and right hand of a person are the same, there may be a case where the type of motion is different due to the habit of the person or the difference in physique. Even in such a case, the identification device 100 of the present embodiment identifies the motion of the person based on the person information, the part information, and the basic motion model. Can be correctly identified.

Although the embodiment of the invention has been described above, the invention is not limited to the description of the embodiment. Hereinafter, each modification will be described.

<Modification 1>
In the above-described embodiment, an example in which the identification device 100, the camera 200, and the device 300 are provided in an automobile has been described, but the present invention is not limited to this. The identification device 100, the camera 200, and the device 300 may be provided in a moving body other than an automobile (for example, a railcar, a ship, an aircraft, etc.) or a place other than the moving body.

<Modification 2>
In the embodiment described above, in the body part movement identification process, the person information is the same as the person information in the input information, and the skeleton information is the movement information in the registration information closest to the body part information in the input information. Although it is identified as information, the invention is not limited to this.

For example, if the person information matches between the input information and the registration information, but there is no match between the site information in the input information and the skeleton information in the registration information, a predetermined number of registration information items are arranged in the order of proximity to the site information in the input information. If all the operation information in the predetermined number of pieces of registration information is searched, the operation information may be used as the operation information in the input information. The number of registered information items to be searched is determined in advance.

Hereinafter, a specific mode of Modification 2 will be described with reference to FIG. It should be noted that in FIG. 6, for ease of explanation, only pieces of registered information whose personal information is “2” are shown. Further, in the following description, an example in which the number of registered information items to be searched is 3 is shown.

In FIG. 6, among the registration information 3002, the registration information close to the part information in the input information 3001 is the data of the closest data 3005, the data closest to the second 3007, and the data closest to the third 3003. The data is retrieved. Then, in this case, since the motion information in all of the retrieved three data is "1", it is determined that the motion information in the input information is "1".

With this, the motion identification accuracy is improved as compared with the case where the skeleton information in the registration information closest to the part information in the input information is simply used. In particular, when there is a small amount of registration information, it is feared that the accuracy of motion identification will decrease if the skeleton information in the registration information that is closest to the part information in the input information is used, so estimation of motion information using the above method is effective. Target.

<Modification 3>
In the above-described embodiment, the behavior model database cannot be used until, for example, 10,000 or more combinations of person information, skeleton information, and movement information are accumulated, but the use restriction of the behavior model is not limited to this.

Even if the accumulated amount of combination of person information, skeleton information, and motion information is less than 10000, it is possible to use the motion model database at the stage when the information that can secure the reliability is accumulated to some extent. is there. Further, even when the amount of accumulated combination of person information, skeleton information, and motion information is small, it is possible to use the motion model database only for specific motions.

In particular, by combining with the method shown in the second modification, it is possible to use the motion model database while ensuring the motion identification accuracy even in the case where the accumulated amount of the combination of person information, skeleton information, and motion information is small. ..

<Modification 4>
In the above-described embodiment, the behavior model is created and stored based on the image capable of skeleton detection, but the manner in which the behavior model is stored is not limited to this. It is possible to store a basic behavior model (hereinafter, referred to as “basic behavior model”) in the identification device 100 in advance when the identification device 100 is shipped.

The basic motion model is a model in which standard skeleton information and motion information are registered for each predetermined age, sex, and physique. This basic motion model is created in advance from data obtained by various experiments and/or calculations.

When using the basic motion model, the age, sex, physique, etc. of the occupant are identified based on the image information, and an appropriate basic motion model is selected from the stored basic motion models.

In addition, it is preferable to store a large number of basic motion models having different ages, sexes, and physiques in the identification device 100 in order to improve identification accuracy.

<Modification 5>
In the modified example 5, in the modified example 4, a basic motion model database and a motion model database created based on an image capable of skeleton detection are separately prepared, and the basic motion model and skeleton detection are possible. The behavior model created based on such an image is switched and used.

Specifically, in a state in which the amount of accumulated motion models created based on images capable of skeleton detection is low, the amount of motion models created based on images capable of skeleton detection is stored using the basic motion model. When the amount exceeds a predetermined amount, the motion model to be used is switched from the basic motion model to the motion model created based on the skeleton-detectable image.

With this configuration, the motion model database can be used while ensuring the accuracy of motion identification even when the amount of motion models created based on the skeleton-detectable image is small.

<Modification 6>
Modification 6 is such that, in Modification 4, the basic motion model and the motion model created based on the skeleton-detectable image are used in combination.

Specifically, when the basic motion model is used and the motion model is created based on the skeleton-detectable image in a state in which the amount of the motion model created based on the skeleton-detectable image is small. Then, the basic behavior model database is updated using the created behavior model. Specifically, the work of adding the obtained skeleton information, person information, and motion information to the basic motion model database is performed.

By doing so, even when the amount of accumulated motion models created based on the skeleton-detectable images is small, the motion model database can be used while ensuring motion identification accuracy. Furthermore, the reliability of the motion model can be further improved by incorporating the data of the motion model with high reliability created based on the image capable of detecting the skeleton into the basic motion model.

<Modification 7>
In the above-described embodiment, the behavior model is created based on the skeleton-detectable image, but the creation of the behavior model is not limited to this. For example, when motion information is identified by the part motion identification unit 170 using a motion model, the person information, part information, and motion information are fed back to the motion model unit 150.

Specifically, in the flowchart of FIG. 5, after step S111, the person information and the body part information used in the body part motion identification process and the motion information identified in the part motion identification process are output to the motion model unit 150. The behavior model is updated in the behavior model unit 150.

By doing so, in addition to creating a motion model based on an image capable of detecting a skeleton, a motion model can be created from motion information identified based on person information, body part information, and motion model. The amount of accumulated electricity increases dramatically. Of the data output from the body part motion identification unit, only the data with high reliability of motion identification, such as the part information in the input information and the plurality of skeleton information in the registration information, are fed back. As a result, it is possible to prevent the reliability of the behavior model from decreasing.

<Modification 8>
In the embodiment described above, it is assumed that the driver, who is an occupant, wears long-sleeved clothes having a low reflectance for near-red light, and the part detected by the part detection unit 160 is exposed to the head. , The left hand, and the right hand, but the site detected by the site detection unit 160 is not limited to this. The number of sites to be detected may be less than three or more than three. Further, instead of or in addition to the hand (from the wrist to the fingertip), a predetermined portion of the upper limb such as an elbow or a shoulder may be detected.

For example, considering the case where the driver wears short-sleeved clothes and gloves having a low reflectance for near red light, in the above-described embodiment, the head can be detected but both hands can be detected. No, it is not possible to perform the part motion identification process. However, if the elbow is detected instead of the hand, the part motion identification process can be performed.

When the driver's seat is on the right side in the vehicle traveling direction, the driver often operates the device 300 with the left hand. Therefore, if the parts to be detected are the head and one or more parts on the upper left limb, even if the number of parts to be detected is small, the movement of the parts to be detected is large, and the motion of the driver can be correctly identified.

<Modification 9>
In the above-described embodiment, as the skeleton information stored in the motion model, only the skeleton information of the part to be detected by the part detection process in the part detection unit 160 is registered. However, the skeleton to be stored in the motion model The type of information is not limited to this.

For example, all skeleton information may be stored in the motion model. In that case, in the body part motion identifying section 170, the part information of the part that can be detected from the image information and the skeleton information of the part corresponding to the part that can be detected from the image information among the skeleton information in the motion model. To match.

By doing so, the part motion identification processing can be performed even when the combination of detectable parts changes.

The embodiments and modifications of the present invention have been described above.

FIG. 7 is a diagram showing a hardware configuration of a computer that realizes the functions of the respective units in the above-described embodiments and modifications by a program.

As shown in FIG. 7, the computer 2100 includes an input device 2101 such as an input button and a touch pad, an output device 2102 such as a display and a speaker, a CPU (Central Processing Unit) 2103, a ROM (Read Only Memory) 2104, and a RAM (Random). Access Memory) 2105. Also, the computer 2100 is a reading device that reads information from a storage device 2106 such as a hard disk device or an SSD (Solid State Drive), a recording medium such as a DVD-ROM (Digital Versatile Disk Read Only Memory), or a USB (Universal Serial Bus) memory. 2107, and a transmission/reception device 2108 that communicates via a network. The above-mentioned units are connected by a bus 2109.

Then, the reading device 2107 reads the program from the recording medium in which the program for realizing the functions of the respective units is recorded, and stores the program in the storage device 2106. Alternatively, the transmission/reception device 2108 communicates with a server device connected to the network, and causes the storage device 2106 to store a program downloaded from the server device for implementing the functions of the above-described units.

Then, the CPU 2103 copies the program stored in the storage device 2106 to the RAM 2105, sequentially reads out the instructions included in the program from the RAM 2105, and executes the instructions, whereby the functions of the above-described units are realized. Further, when the program is executed, the RAM 2105 or the storage device 2106 stores the information obtained by the various processes described in the embodiments and appropriately uses it.

The identification device according to the present invention is useful for identifying a person who operates a predetermined device.

100 identification device 101 input unit 102 control unit 110 detection control unit 120 skeleton detection unit 130 skeleton motion identification unit 140 person identification unit 150 motion model unit 160 site detection unit 170 site operation identification unit 180 device control unit 200 camera 300 device 1000, 1001 , 1002 image 1003 database 2001 input information 2002 registration information 2003 data 2100 computer 2101 input device 2102 output device 2103 CPU
2104 ROM
2105 RAM
2106 storage device 2107 reading device 2108 transmitting/receiving device 2109 bus 3001 input information 3002 registration information 3003, 3004, 3005, 3006, 3007 data

Claims (6)

An input unit that receives the image information of the person photographed by the camera,
An operation in which the person is identified and at least the head and hand parts of the person are detected based on the image information, and the identified person, the detected part, and the person's motion are registered for each person. A control unit for identifying the motion of the person based on a model,
The identification device includes: a first action model generated based on the action of the person identified by the control unit; and a second action model generated and stored in advance. The identification device according to claim 1, wherein the control unit switches and uses the first behavior model and the second behavior model based on an accumulated amount of the first behavior model. The control unit uses a second behavior model when the accumulated amount of the first behavior model is less than a predetermined amount, and uses a behavior model to be used when the accumulated amount of the first behavior model is equal to or more than the predetermined amount. The identification device according to claim 2, wherein the second behavior model is switched to the first behavior model. The identification device according to claim 1, wherein the control unit uses the first behavior model and the second behavior model in combination based on an accumulated amount of the first behavior model. The control unit uses the second behavior model when the accumulated amount of the first behavior model is less than a predetermined amount, and uses the generated behavior model when the first behavior model is generated. The identification device according to claim 4, wherein the second behavior model is updated. The control unit, when identifying the action of the person, feeds back the person information of the person, part information of the person, or action information indicating an identification result of the action of the person to the first action model. The identification device according to 1.

Images