JP2019139642A - Device, system, and method for detecting locations - Google Patents

Abstract

To provide a location detection device capable of detecting locations of people inside a building without using special equipment for detecting locations.SOLUTION: A location detection device comprises: an image acquisition unit configured to acquire images captured inside a building as target images indicative of targets for detecting shot locations; and a location detection unit configured to detect shot locations where the target images were captured from the target images using a location detection model for estimating the shot locations.SELECTED DRAWING: Figure 1

Description

  The present invention relates to a position detection device, a position detection system, and a position detection method for detecting an imaging position.

  Conventionally, various methods for detecting a human position in a building or the like have been studied. For example, there are many configurations of communication terminal technologies such as beacons, wireless tags, wifi radio waves, geomagnetism, IMES (Indoor MESsaging System), PHS (Personal Handy-phone System), and combinations thereof (for example, Patent Document 1). reference).

JP 2011-214834 A

  However, the above-described methods for detecting the position of a person need to always carry a dedicated device that indicates the position of each person carried by each person, or perform position detection by radio waves or optically. There is a problem that it is necessary to arrange dedicated equipment such as a position detection sensor in the building, which increases the device cost.

  In view of the above-described problems, the present invention provides a position detection device, a position detection system, and a position detection method capable of detecting the position of a person indoors without using a dedicated device for detecting the position. For the purpose.

  In order to solve the above-described problem, a position detection device according to an aspect of the present invention includes an image acquisition unit that acquires an image captured in a building as a target image that indicates a target for detecting an imaging position; And a position detection unit that detects an imaging position where the image is captured using a position detection model that estimates the imaging position from the target image.

  In the position detection device according to one aspect of the present invention, the position detection model is a machine that uses learning data in which information indicating the imaging position at which the learning image is captured is added to the learning image that is captured inside the building. It is a learning model created by executing learning.

  The position detection device according to one aspect of the present invention further includes a position information storage unit that stores information that identifies the target image.

  In the position detection system of one embodiment of the present invention, the above-described position detection device and learning data in which information indicating the imaging position at which the learning image is captured are attached to the learning image in which the inside of the building is captured are used. And a learning server that creates an imaging position detection model for detecting an imaging position at which an image is captured by executing machine learning.

  In the position detection method of one embodiment of the present invention, the image acquisition unit acquires an image captured in the building as a target image indicating a target for detecting the imaging position, and the position detection unit detects that the target image is An imaged imaging position is detected using a position detection model that estimates the imaging position from the target image.

  According to the present invention, it is possible to detect the position of a person indoors without using a dedicated device for detecting the position.

It is a block diagram which shows the structural example of the position detection system 1 of embodiment. It is an image figure which shows the example which the camera 12 of embodiment images indoors. It is a figure which shows an example of the target image imaged with the camera 12 of embodiment. It is a figure which shows an example of the learning data used by the learning server 11 of embodiment. It is a sequence chart explaining the flow of processing in which position detection system 1 of an embodiment detects a position.

Hereinafter, a position detection device, a position detection system, and a position detection method according to embodiments of the present invention will be described.
FIG. 1 is a block diagram illustrating a configuration example of a position detection system 1 according to the embodiment. The position detection system 1 includes a position detection device 10, a learning server 11, and a camera 12. In the present embodiment, the case where the position detection system 1 includes a plurality of cameras 12 is illustrated, but the position detection system 1 only needs to include at least one camera 12.
Each of the position detection device 10, the learning server 11, and the camera 12 is connected to an information communication network 500 including the Internet and a local area network, and transmits and receives data to and from each other. The learning server 11 may be an on-premises server device installed in a building, or an external cloud computing server device connected to the information communication network 500.

In the present embodiment, each of the cameras 12 is a camera that can move in a building. The camera 12 is, for example, an imaging device provided in a portable device (including a smartphone) carried by a person in a building. Alternatively, the camera 12 is an imaging device that is mounted on a moving body (for example, a mobile robot or a human) and moves.
The camera 12 may be, for example, a 360-degree camera provided in a patrol robot that patrols a building, or an imaging device provided in a guard's shoulder or head. The camera 12 captures a captured image showing the scenery in the building imaged from the position where the patrol robot and the security guard are located, via the communication function provided in the imaging device of the patrol robot and the security guard, and the position detection device 10 and the learning server. 11 to send.

The position detection device 10 includes a captured image acquisition unit 101, a position detection unit 102, a captured image storage unit 103, and a position information storage unit 104.
The captured image acquisition unit 101 acquires captured images captured by the cameras 12 in the building from the respective cameras 12 via the information communication network 500. Here, the captured image acquisition unit 101 is an example of an “image acquisition unit”. The captured image acquired by the captured image acquisition unit 101 is an image that is a target for detecting the imaging position, and is an example of a “target image”.

  The captured image acquisition unit 101 outputs the captured image acquired from each of the cameras 12 to the position detection unit 102, and also transmits the camera identification information and time stamp (time information) of the camera 12 to the captured image storage unit 103. ) Is added and stored.

The position detection unit 102 detects the imaging position in the captured image supplied from the captured image acquisition unit 101 using a position detection model. For example, when the camera 12 is an imaging device provided in a mobile terminal device, the position detection unit 102 detects the position of the mobile terminal in a building using a position detection model. The position detection model is a learned model generated by the learning server 11 described later. The position detection model will be described in detail later.
The position detection unit 102 writes and stores information indicating the detected imaging position in the position information storage unit 104 by adding information indicating the captured image.

The captured image storage unit 103 stores information in which the captured image acquired by the captured image acquisition unit 101 is associated with camera identification information of the camera 12 and a time stamp.
The position information storage unit 104 stores the imaging position detected by the position detection unit 102 together with information indicating the captured image.

Here, an image acquired by the captured image acquisition unit 101 will be described with reference to FIGS. 2 and 3.
FIG. 2 is an image diagram illustrating an example in which the camera 12 of the embodiment images indoors.
FIG. 3 is a diagram illustrating an example of an indoor image captured by the camera 12 according to the embodiment.
As shown in FIG. 2, a person in the building images a landscape in the building with the camera 12 of the smartphone.

  As illustrated in FIG. 3, the image captured by the camera 12 includes a room number, a floor pattern, and the like as in the signs M <b> 1 to M <b> 4. These signs M can be identifiers that identify locations in the building. In other words, if a learning model is created by performing machine learning using the learning data obtained by adding the imaging position of the image to the image including these signs M, the created model is added to the image. Since the imaging position can be estimated based on this, the imaging position of the camera 12 can be specified. The imaging position of the camera 12 is estimated as a position in a human building having a camera 12 (having a smartphone equipped with the camera 12).

  Returning to FIG. 1, the learning server 11 is a server device that creates a learned model (position detection model) used for position detection by the position detection device 10 based on learning data. The learning data is acquired in advance before the position detection is performed by the position detection system 1, for example, and a learned model (position detection model) is created based on the learning data.

The learning server 11 includes a learning data acquisition unit 111, a model creation unit 112, and a learning model storage unit 113.
The learning data acquisition unit 111 acquires a captured image for learning captured by the learning camera 12A from the learning camera 12A via the information communication network 500. Here, the captured image for learning captured by the learning camera 12A is an example of a “learning image”.

  Here, the learning camera 12A is an imaging device that captures an image used as learning data when creating a position detection model. For example, various scenes in a building captured by a traveling robot with a 360-degree camera are used. Each of the images showing is taken. For example, the learning camera 12 </ b> A images an indoor landscape while traveling around a building, and transmits the captured image to the learning server 11.

  The learning data acquisition unit 111 acquires information indicating the imaging position of the captured image captured by the learning camera 12A. The information indicating the imaging position may be, for example, coordinate information of the world geodetic system as indicated by latitude and longitude, or information that can identify the position in the building such as a guidance display plate in the building. May be.

  For example, the learning data acquisition unit 111 acquires information indicating the imaging position from a traveling robot with a 360-degree camera that acquires a captured image for learning. In this case, for example, the learning data acquisition unit 111 circulates the building with a 360 degree camera-equipped robot with an indoor GPS (Global Positioning System) attached, and obtains position information from the GPS every time a captured image is captured. The information (GPS position information) indicating the imaging position is acquired together with the captured image. Or the learning data acquisition part 111 may acquire an imaging position based on the information input into the input part (not shown) by input operation, such as a manager of a building.

The model creation unit 112 performs machine learning using the learning captured image acquired by the learning data acquisition unit 111 and information indicating the imaging position where the captured image is captured as the learning data, so that A position detection model for detecting the imaging position is created. The machine learning here is, for example, supervised learning.
The learning model storage unit 113 stores the position detection model created by the model creation unit 112.

FIG. 4 is a diagram illustrating an example of learning data used by the learning server 11 of the embodiment. 4A to 4E include an identifier (for example, sign M) that can identify a position in the building, such as a sign including a sign such as a signboard or a landmark in the building, a floor, a ceiling, a wall, or a pillar. This is an image.
The images shown in FIGS. 4A to 4E are attached with information indicating the imaging position where the captured image is captured, and are used as learning data used for machine learning.

  For example, in FIG. 4A, a plate marked “Meeting room 2-A” is placed on the wall, and a dot pattern carpet is laid on the floor at that position. In addition, a plate marked “Meeting room 2-B” is lifted on the wall, and an image of a hatched carpet on the floor at that position is captured. In this case, the image includes a plate of “conference room 2-A” and “conference room 2-B” from the corridor side of the floor where “conference room 2-A” and “conference room 2-B” are provided. A position overlooking can be attached as information indicating the imaging position. Or the information which shows the predetermined position coordinate of the corridor of the floor in which "Conference room 2-A" and "Conference room 2-B" were provided is attached to the said image.

  Further, for example, in FIG. 4B, the plate marked “Conference room 2-A” is imaged. The position where the plate of “conference room 2-A” can be imaged from the front in the horizontal direction is attached to the image as information indicating the imaging position. Further, for example, in FIG. 4C, the plate marked “Conference room 2-B” is imaged. The position where the plate of “conference room 2-B” can be imaged from the front in the horizontal direction is attached to the image as information indicating the imaging position. In addition, for example, in FIG. 4D, a diagonal carpet pattern is captured. In the image, a position where the dot-patterned carpet can be imaged from the vertical direction on the front is attached as information indicating the imaging position. Further, for example, in FIG. 4 (e), a hatched carpet is captured. In the image, a position where a diagonal carpet pattern can be imaged from the vertical direction in front of the carpet is attached as information indicating the imaging position.

  As already described, the model creation unit 112 executes machine learning using learning data in which information indicating the imaging position where the learning image is captured is added to the learning image captured within the building, A learning model for detecting an imaging position where the indoor image is captured is created.

  The position detection model is obtained by modeling a specific factor for specifying a position in a building, for example. In this position detection model, specific factors are modeled for various places in the building. For example, when the position in the building is “conference room 2-A”, the specific factor is the plate of “conference room 2-A” or a carpet with a dot pattern. In addition, when the position in the building is “conference room 2-B”, the specific factor is the plate of “conference room 2-B” or the diagonally-patterned carpet.

FIG. 5 is a sequence chart for explaining the flow of processing in which the position detection system 1 of the embodiment detects a position.
First, the learning camera 12A captures a captured image for learning (step S101). The learning camera 12A, for example, circulates in a building and images various places in the building.
Next, the learning camera 12A associates information indicating the imaging position with the captured image (step S102). The learning camera 12A may acquire position information using an indoor GPS or the like each time an image is captured, or acquire position information corresponding to a captured image instructed by an input operation by a building manager or the like. Also good.

Next, the learning camera 12A transmits information associating the captured position with the captured image to the learning server 11 as learning data (step S103).
The learning data acquisition unit 111 acquires learning data (step S104). The learning data acquisition unit 111 outputs the acquired learning data to the model creation unit 112.
The model creation unit 112 creates a position detection model by performing machine learning using the learning data supplied from the learning data acquisition unit 111 (step S105). The model creation unit 112 stores the created position detection model in the learning model storage unit 113.

On the other hand, the camera 12 images the landscape in the building (step S106).
The camera 12 transmits the captured image to the position detection device 10 (step S107).
The captured image acquisition unit 101 acquires a captured image captured by the camera 12 (step S108). The captured image acquisition unit 101 outputs the acquired image to the position detection unit 102. The position detection unit 102 transmits the captured image supplied from the captured image acquisition unit 101 to the learning server 11 (step S109).

The learning server 11 receives the captured image from the position detection device 10 (step S110), and obtains information indicating the imaging position from the position detection model by inputting the captured image into the position detection model created by the model creation unit 112 ( Step S111).
The learning server 11 transmits information indicating the acquired imaging position to the position detection device 10 (step S112).

The position detection unit 102 detects information indicating the imaging position received from the learning server 11 as the imaging position of the captured image (step S113).
The position detection unit 102 stores the detected imaging position in the position information storage unit 104 together with information for identifying the captured image (step S114).

  In the above-described example, the position detection unit 102 has exemplified and described the case where the information indicating the imaging position is acquired from the learning server 11 by transmitting the captured image to the learning server 11, but is not limited thereto. .

  For example, the position detection device 10 may acquire the position detection model created by the model creation unit 112 of the learning server 11. In this case, the position detection device 10 includes a storage unit (not shown) that stores the position detection model, and stores the position detection model received from the learning server 11.

  And the position detection apparatus 10 detects an imaging position by inputting the captured image supplied from the captured image acquisition part 101 to the position detection model memorize | stored in the memory | storage part which is not illustrated of the position detection apparatus 10. FIG. Thereby, since the position detection apparatus 10 can detect an imaging position without going through the information communication network 500, it is possible to reduce a communication burden.

  As described above, the position detection apparatus 10 according to the embodiment can detect the imaging position by the position detection unit 102 using the position detection model. For this reason, without using a dedicated device for detecting the position of a beacon, wireless tag, wifi radio wave, geomagnetism, IMES, PHS or the like, the person who is indoors can simply determine the position where the person is located Can be detected. In addition, since the position detection model is a learning model created by executing machine learning, it is not necessary to perform complicated processing such as image analysis of the captured image in order to specify the captured position. It is possible to easily detect the imaging position only by executing machine learning in advance with learning data in which the imaging position is associated.

  In addition, the position detection device 10 according to the embodiment includes a position information storage unit 104 that stores information indicating an imaging position in association with a captured image. Thereby, it is possible to grasp the movement history of a specific camera 12 (the person carrying it) by arranging the captured images captured by the same camera 12 and their imaging positions in time series.

  Further, the program for realizing the functions of the position detection system 1 and the position detection apparatus 10 in FIG. 1 is recorded on a computer-readable recording medium, and the program recorded on the recording medium is read by the computer system. By executing the control, the control of detecting the imaging position of the camera 12 may be performed. Here, the “computer system” includes an OS and hardware such as peripheral devices.

Further, the “computer system” includes a homepage providing environment (or display environment) if a WWW system is used.
The “computer-readable recording medium” refers to a storage device such as a flexible medium, a magneto-optical disk, a portable medium such as a ROM and a CD-ROM, and a hard disk incorporated in a computer system. Furthermore, the “computer-readable recording medium” dynamically holds a program for a short time like a communication line when transmitting a program via a network such as the Internet or a communication line such as a telephone line. In this case, a volatile memory in a computer system serving as a server or a client in that case, and a program that holds a program for a certain period of time are also included. The program may be a program for realizing a part of the functions described above, and may be a program capable of realizing the functions described above in combination with a program already recorded in a computer system.

  The embodiment of the present invention has been described in detail with reference to the drawings. However, the specific configuration is not limited to this embodiment, and includes designs and the like that do not depart from the gist of the present invention.

DESCRIPTION OF SYMBOLS 1 ... Position detection system 10 ... Position detection apparatus 11 ... Learning server 12 ... Camera 101 ... Captured image acquisition part 102 ... Position detection part 103 ... Captured image storage part 104 ... Position information storage part 111 ... Learning data acquisition part 112 ... Model creation Part

Claims (5)

An image acquisition unit that acquires an image captured in a building as a target image indicating a target for detecting an imaging position;
A position detection device comprising: a position detection unit that detects an image pickup position at which the target image is picked up using a position detection model that estimates the image pickup position from the target image. The position detection model is a learning model created by performing machine learning using learning data in which information indicating the imaging position where the learning image is captured is added to the learning image capturing the inside of the building. The position detection device according to claim 1. The position detection device according to claim 1, further comprising a position information storage unit that stores information indicating the imaging position detected by the position detection unit together with information for identifying the target image. . The position detection device according to claim 1 or 2,
In order to detect the imaging position where the image was captured by performing machine learning using learning data in which information indicating the imaging position where the learning image was captured is added to the learning image captured within the building A position detection system comprising: a learning server that creates a position detection model of An image acquisition unit acquires an image captured in a building as a target image indicating a target for detecting an imaging position,
A position detection method, wherein a position detection unit detects an imaging position where the target image is captured using a position detection model that estimates the imaging position from the target image.

Images