DETECTION APPARATUS, DETECTION METHOD, AND PROGRAM

Abstract

The present invention provides a detection device (10) including: an equipment information acquisition unit (12) that acquires equipment information indicating the position of equipment present in the interior of a moving body; a person information generation unit (11) that identifies the position of a person present in the interior of the moving body on the basis of an image representing the interior of the moving body, and generates person information indicating the position of the person; and a detection unit (13) that detects a danger state occurring in the interior of the moving body on the basis of the equipment information and the person information.

Description

[Technical field]

[0001] The present invention relates to a detection device, a detection method, and a program. [Background technology]

[0002] By detecting dangerous situations (situations where an accident may occur or has occurred) occurring inside a moving body early and taking necessary measures, it is possible to prevent accidents (such as falls) inside the moving body or prevent accidents from becoming serious. Patent Document 1 discloses a technology for preventing passengers from falling, etc. inside a moving body. Specifically, it discloses detecting passengers' postures, positions, actions, gripping of straps or bars, etc., and executing processing according to the detection results. Non-Patent Document 1 discloses a technology for detecting a person's skeletal structure. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] JP 2020-003936 A [Non-patent literature]

[0004] [Non-Patent Document 1] Zhe Cao, Tomas Simon, Shih-En Wei, Yaser Sheikh, "Realtime Multi-Person 2D Pose Estimation using Part Affinity Fields", The IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2017, P. 7291-7299 Summary of the Invention [Problem to be solved by the invention]

[0005] Patent Document 1 does not disclose a specific means for detecting passengers' posture, position, behavior, gripping of straps or handrails, etc. In order to accurately prevent accidents (such as falls) inside a moving object or to prevent the accident from becoming serious, it is necessary to accurately detect the detection target.

[0006] An object of the present invention is to accurately detect a dangerous state occurring inside a moving body. [Means for solving the problem]

[0007] According to the present invention, A facility information acquisition means for acquiring facility information indicating a location of a facility present inside the mobile body; a person information generating means for identifying a position of a person present inside the moving body based on an image showing the inside of the moving body and generating person information showing the position of the person; a detection means for detecting a dangerous state occurring inside the moving object based on the facility information and the person information; A detection device is provided having the following:

[0008] Further, according to the present invention, The computer Acquire facility information indicating the location of the facility inside the mobile object; Identifying a position of a person present inside the moving body based on an image showing the inside of the moving body, and generating person information showing the position of the person; A detection method is provided for detecting a dangerous state occurring inside the mobile object based on the facility information and the person information.

[0009] Further, according to the present invention, Computer, A facility information acquisition means for acquiring facility information indicating a location of a facility present inside the mobile object; a person information generating means for identifying a position of a person present inside the moving body based on an image showing the inside of the moving body, and generating person information showing the position of the person; a detection means for detecting a dangerous state occurring inside the moving object based on the facility information and the person information; A program is provided to function as a Effect of the Invention

[0010] According to the present invention, it is possible to accurately detect a dangerous state occurring inside a moving body. [Brief description of the drawings]

[0011] [Figure 1] FIG. 2 is a diagram illustrating an example of a hardware configuration of a detection device. [Diagram 2] FIG. 2 is a functional block diagram of a detection device. [Diagram 3] FIG. 2 is a diagram illustrating an example of a human body model. [Figure 4] FIG. 13 is a diagram showing an example of detection of a skeletal structure. [Diagram 5] FIG. 13 is a diagram showing an example of detection of a skeletal structure. [Figure 6] FIG. 13 is a diagram showing an example of detection of a skeletal structure. [Figure 7] 10 is a flowchart showing an example of a process flow of the detection device. [Figure 8] 10 is a flowchart showing an example of a process flow of a detection device. [Figure 9] 10 is a flowchart showing an example of a process flow of a detection device. [Figure 10] FIG. 2 is a diagram illustrating an example of a human body model. [Figure 11] FIG. 13 is a diagram showing an example of detection of a skeletal structure. [Figure 12] FIG. 13 is a diagram showing an example of detection of a skeletal structure. [Figure 13] FIG. 13 is a diagram showing an example of detection of a skeletal structure. [Figure 14] FIG. 2 is a diagram illustrating an example of a human body model. [Figure 15] FIG. 13 is a diagram showing an example of detection of a skeletal structure. [Figure 16] 13 is a histogram for explaining a height pixel number calculation method. [Figure 17]FIG. 13 is a diagram showing an example of detection of a skeletal structure. [Figure 18] FIG. 1 is a diagram showing a three-dimensional human body model. [Figure 19] FIG. 13 is a diagram for explaining a height pixel number calculation method. [Figure 20] FIG. 13 is a diagram for explaining a height pixel number calculation method. [Figure 21] FIG. 13 is a diagram for explaining a height pixel number calculation method. [Figure 22] FIG. 13 is a diagram for explaining a normalization method. [Figure 23] FIG. 13 is a diagram for explaining a normalization method. [Figure 24] FIG. 13 is a diagram for explaining a normalization method. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0012] Hereinafter, embodiments of the present invention will be described with reference to the drawings. In all the drawings, the same components are given the same reference numerals and the description will be omitted as appropriate.

[0013] <First embodiment> "overview" The detection device of this embodiment detects a dangerous state occurring inside a moving body based on a relative positional relationship between the position of the equipment and the position of the person calculated based on equipment information indicating the position of the equipment present inside the moving body and person information indicating the position of the person present inside the moving body. By making a judgment based on the relative positional relationship between the position of the equipment and the position of the person, it becomes possible to accurately detect a dangerous state occurring inside the moving body.

[0014] In addition, the detection device of this embodiment detects the skeletal structure of a person included in an image (image generated using an image sensor) captured inside a moving body by a camera or the like, and identifies matters related to the person, such as the person's position and posture, based on the detected skeletal structure of the person. As a method for detecting the position, posture, etc. of a person based on the skeletal structure of a person included in an image generated by a camera or the like, several methods that are robust against changes in lighting conditions are known. As with the detection device of this embodiment, by detecting the skeletal structure of a person included in an image generated by a camera or the like and detecting the position, posture, etc. of the person based on the skeletal structure of the person, analysis that is robust against changes in lighting conditions is realized. It is then possible to accurately detect a dangerous situation occurring inside a moving body regardless of the lighting conditions.

[0015] "composition" Next, the configuration of the detection device will be described. First, an example of the hardware configuration of the detection device will be described. Each functional unit of the detection device is implemented by a CPU (Central Processing Unit) of an arbitrary computer. It is realized by any combination of hardware and software, centered on memory, a program loaded into the memory, a storage unit such as a hard disk that stores the program (in addition to programs that are stored before the device is shipped, it can also store programs downloaded from storage media such as CDs (Compact Discs) or servers on the Internet), and a network connection interface. It will be understood by those skilled in the art that there are many variations in the method of realizing this and the device.

[0016] FIG. 1 is a block diagram illustrating a hardware configuration of a detection device. As shown in FIG. 1, the detection device has a processor 1A, a memory 2A, an input / output interface 3A, a peripheral circuit 4A, and a bus 5A. The peripheral circuit 4A includes various modules. The detection device does not need to have the peripheral circuit 4A. The detection device may be composed of a plurality of physically and / or logically separated devices, or may be composed of a single physically and / or logically integrated device. When the detection device is composed of a plurality of physically and / or logically separated devices, each of the plurality of devices can have the above hardware configuration.

[0017] The bus 5A is a data transmission path for the processor 1A, memory 2A, peripheral circuit 4A, and input / output interface 3A to transmit and receive data to each other. The processor 1A is an arithmetic processing device such as a CPU or a GPU (Graphics Processing Unit). The memory 2A is a memory such as a RAM (Random Access Memory) or a ROM (Read Only Memory). The input / output interface 3A includes an interface for acquiring information from an input device, an external device, an external server, an external sensor, a camera, etc., and an interface for outputting information to an output device, an external device, an external server, etc. Examples of the input device include a keyboard, a mouse, a microphone, a physical button, a touch panel, etc. Examples of the output device include a display, a speaker, a printer, a mailer, etc. The processor 1A can issue commands to each module and perform calculations based on the results of those calculations.

[0018] Next, the functional configuration of the detection device will be described. The detection device is a device mounted on a moving body. Examples of moving bodies include, but are not limited to, vehicles (such as buses), trains, ropeways, elevators, ships, and airplanes. The detection device is suitable for mounting and use on moving bodies in which a person inside may move while standing. The moving body may have a function (automatic driving function) in which a system controls the moving operation of the moving body (starting, stopping, accelerating, decelerating, changing the direction of travel, etc.), or may have a function (manual driving function) in which a person such as a driver controls the moving operation, or may have both functions.

[0019] 2 shows an example of a functional block diagram of the detection device 10. As shown in the figure, the detection device 10 includes a person information generation unit 11, a facility information acquisition unit 12, a detection unit 13, and an output unit 14.

[0020] The equipment information acquisition unit 12 acquires equipment information indicating the position of equipment present inside the moving body. The equipment whose position is indicated by the equipment information is, for example, equipment installed to prevent accidents inside the moving body, and can include at least one of a strap, a handrail, and a seat. Other equipment whose position is indicated by the equipment information is, for example, equipment that may be involved in accidents inside the moving body, and examples thereof include doors and windows. Note that the examples here are merely examples, and the equipment whose position is indicated by the equipment information is not limited to these. The equipment information may be information that links information indicating the type of each piece of equipment (such as the name and identification information of the equipment) with the position of each piece of equipment.

[0021] The position of the equipment is indicated by the coordinates of a three-dimensional coordinate system (hereinafter sometimes referred to as the "world coordinate system") with a predetermined position of the moving body as the origin and the x-axis, y-axis, and z-axis in predetermined directions. The equipment information may indicate the position of each piece of equipment by the coordinates of a representative point (one or more points) of each piece of equipment. How to define the representative point is a design matter. Additionally, the equipment information may indicate the space that each piece of equipment occupies in the world coordinate system. The space that each piece of equipment occupies can be indicated using any known technology. For example, the space that each piece of equipment occupies may be indicated by the coordinates of multiple points on the outer surface of each piece of equipment, may be indicated by a function indicating the outer surface of each piece of equipment, or may be indicated by any other method.

[0022] Next, a description will be given of a method for acquiring facility information by the facility information acquisition unit 12. The facility information acquisition unit 12 can acquire facility information based on any one of the following first to third acquisition examples.

[0023] -First acquisition example- In this example, facility information indicating the position of each facility is generated in advance and stored in a storage device accessible from the detection device 10. Then, the facility information acquisition unit 12 acquires the facility information stored in the storage device.

[0024] -Second acquisition example- In this example, the equipment information acquisition unit 12 analyzes an image (an image generated using an image sensor) showing the inside of a moving object generated by a camera, detects each piece of equipment in the image, and identifies the position and type of each piece of detected equipment. The equipment information acquisition unit 12 acquires the image generated by the camera through real-time processing, analyzes it, and identifies the latest position of each piece of equipment. The detection device 10 and the camera are connected to be able to communicate with each other via wire and / or wirelessly.

[0025] The camera may have a fixed position and orientation (the direction of shooting (optical axis)) inside the moving body, or the position may be fixed but the orientation may be variable. Alternatively, the camera may be configured to be movable inside the moving body and be able to shoot while changing its position and orientation. The means for changing the orientation of a camera whose position is fixed and the manner in which the orientation changes (such as swinging periodically in a specified direction with a specified amplitude) are design matters, and any conventional technology may be adopted. Furthermore, the means for moving the camera inside the moving body (repeatedly moving to the same position along a lane, being equipped with an autonomous moving means and moving based on a specified computer control, etc.) is not particularly limited, and any conventional technology may be adopted.

[0026] The means for analyzing the image and detecting each piece of equipment in the image is not particularly limited, but may be realized by detecting the feature amount in the image, which is registered in advance, for example. Furthermore, the means for identifying the internal position of the moving object from the position in the image of each piece of equipment detected in the image (the means for converting from the image coordinate system to the world coordinate system) is not particularly limited, and any conventional technology may be adopted.

[0027] -Third Acquisition Example- In this example, the first acquisition example and the second acquisition example are combined. Specifically, facility information indicating the position of facility whose position inside the moving body is invariable is acquired in the first acquisition example, and facility information indicating the position of facility whose position inside the moving body is variable is acquired in the second acquisition example. Facilities whose position inside the moving body is invariable include seats and handrails. Facilities whose position inside the moving body is variable include straps and seats (auxiliary seats, etc.) whose shape, state, and position can be changed by being stored or folded.

[0028] The person information generating unit 11 analyzes an image (an image generated using an image sensor) showing the inside of a moving object generated by a camera, detects the skeletal structure of a person included in the image, and generates various information related to the detected person based on the detected skeletal structure. The person information generating unit 11 acquires the image generated by the camera through real-time processing, analyzes it, and generates various information related to the person. The detection device 10 and the camera are connected to be able to communicate with each other by wire and / or wirelessly. The camera has been described above, so a description thereof will be omitted here.

[0029] Detection of a person's skeletal structure is achieved using a skeletal detection technology such as OpenPose disclosed in Non-Patent Document 1. The detected skeletal structure is composed of "key points" which are characteristic points such as joints, and "bones (bone links)" which indicate the links between key points. "Key points" correspond to, for example, a person's "joints," and "bones" correspond to, for example, a person's "bones."

[0030] Fig. 3 shows an example of a skeletal structure detected from a two-dimensional image using a skeletal detection technique such as OpenPose, etc. Fig. 4 to Fig. 6 show examples of skeletal structure detection.

[0031] For example, the person information generating unit 11 extracts feature points that can be key points from an image, and detects each key point of the person by referring to information obtained by machine learning the image of the key points. In the example of Fig. 3, the head A1, neck A2, right shoulder A31, left shoulder A32, right elbow A41, left elbow A42, right hand A51, left hand A52, right hip A61, left hip A62, right knee A71, left knee A72, right foot A81, and left foot A82 are detected as the key points of the person. Furthermore, the bones of the person connected by these key points are detected as bones, including bone B1 connecting the head A1 and neck A2, bones B21 and B22 connecting the neck A2 to the right shoulder A31 and left shoulder A32 respectively, bones B31 and B32 connecting the right shoulder A31 and left shoulder A32 to the right elbow A41 and left elbow A42 respectively, bones B41 and B42 connecting the right elbow A41 and left elbow A42 to the right hand A51 and left hand A52 respectively, bones B51 and B52 connecting the neck A2 to the right hip A61 and left hip A62 respectively, bones B61 and B62 connecting the right hip A61 and left hip A62 to the right knee A71 and left knee A72 respectively, and bones B71 and B72 connecting the right knee A71 and left knee A72 to the right foot A81 and left foot A82 respectively.

[0032] Fig. 4 shows an example of detecting the skeletal structure of a person standing up. In Fig. 4, the person standing up is imaged from the front, and bones B1, B51 and B52, B61 and B62, and B71 and B72 are detected without overlapping when viewed from the front, and bones B61 and B71 of the right foot are slightly more bent than bones B62 and B72 of the left foot.

[0033] Fig. 5 shows an example of detecting the skeletal structure of a person in a crouching state. In Fig. 5, the person is imaged from the right side, and bones B1, B51 and B52, B61 and B62, and B71 and B72 are detected as seen from the right side, and bones B61 and B71 of the right foot and bones B62 and B72 of the left foot are greatly bent and overlap each other.

[0034] Fig. 6 is an example of detecting the skeletal structure of a person lying down (sleeping, collapsed, etc.). In Fig. 6, the person lying down is imaged from the diagonal front left, and bones B1, B51 and B52, B61 and B62, and B71 and B72 as seen from the diagonal front left are detected, with bones B61 and B71 of the right foot and bones B62 and B72 of the left foot being bent and overlapping.

[0035] Next, an example of information (person information) about a person generated based on the detected skeletal structure will be described. The person information generating unit 11 can generate at least one of the following multiple pieces of person information for each skeletal structure (each person) detected in the image.

[0036] -First person information example- The person information generating unit 11 can extract features from the skeletal structure and use them as person information. The features of the skeletal structure indicate the characteristics of a person's skeleton, and based on the features, it is possible to estimate the posture of the person and to identify the same person in different images (tracking the person in the images).

[0037] Usually, this feature includes a plurality of parameters. The feature may be the feature of the entire skeletal structure, may be the feature of a part of the skeletal structure, or may include a plurality of feature such as each part of the skeletal structure. The method of calculating the feature may be any method such as machine learning or normalization, and the minimum value or maximum value may be obtained as normalization. As an example, the feature is a feature obtained by machine learning the skeletal structure, or the size of the skeletal structure from the head to the foot on the image. The size of the skeletal structure is the height in the vertical direction or the area of ​​the skeletal region including the skeletal structure on the image. The vertical direction (height direction or vertical direction) is the vertical direction (Y-axis direction) in the image, for example, a direction perpendicular to the ground (reference plane). The left-right direction (horizontal direction) is the left-right direction (X-axis direction) in the image, for example, a direction parallel to the ground.

[0038] For example, when the height or area of ​​a skeleton region is used as a feature, the person information generating unit 11 extracts a region including a skeleton structure and obtains the height (number of pixels) and area (pixel area) of the region. The height and area of ​​the skeleton region are obtained from the coordinates of the edge of the extracted skeleton region and the coordinates of the key points of the edge.

[0039] In the example of Figure 4, a skeletal region including all bones is extracted from the skeletal structure of a standing person. In this case, the top end of the skeletal region is key point A1 of the head, the bottom end of the skeletal region is key point A82 of the left foot, the left end of the skeletal region is key point A41 of the right elbow, and the right end of the skeletal region is key point A52 of the left hand. Therefore, the height of the skeletal region is calculated from the difference in the Y coordinates of key point A1 and key point A82. In addition, the width of the skeletal region is calculated from the difference in the X coordinates of key point A41 and key point A52, and the area is calculated from the height and width of the skeletal region.

[0040] In the example of Figure 5, a skeletal region including all bones is extracted from the skeletal structure of a crouching person. In this case, the top end of the skeletal region is key point A1 on the head, the bottom end of the skeletal region is key point A81 on the right foot, the left end of the skeletal region is key point A61 on the right hip, and the right end of the skeletal region is key point A51 on the right hand. Therefore, the height of the skeletal region is calculated from the difference in the Y coordinates of key points A1 and A81. The width of the skeletal region is calculated from the difference in the X coordinates of key points A61 and A51, and the area is calculated from the height and width of the skeletal region.

[0041] In the example of Figure 6, a skeletal region including all bones is extracted from the skeletal structure of a person lying down. In this case, the top end of the skeletal region is key point A32 on the left shoulder, the bottom end of the skeletal region is key point A52 on the left hand, the left end of the skeletal region is key point A51 on the right hand, and the right end of the skeletal region is key point A82 on the left foot. Therefore, the height of the skeletal region is calculated from the difference in the Y coordinates of key point A32 and key point A52. The width of the skeletal region is calculated from the difference in the X coordinates of key point A51 and key point A82, and the area is calculated from the height and width of the skeletal region.

[0042] As described above, various processes can be performed using the feature of the skeletal structure, but it is preferable to perform various processes using a feature suitable for each process among multiple types of feature. For example, in a process for identifying the same person included in different images, it is preferable to use a feature that is not affected by the person's orientation or posture.

[0043] -Example of second person information- The person information generating unit 11 can identify the position of a person present inside the moving body based on the detected skeletal structure of the person, and generate person information indicating the position of the person. The person information generating unit 11 can generate information indicating the position of a predetermined part of the person's body as the person information indicating the position of the person. The predetermined part can be at least one of a plurality of key points and a plurality of bones indicated by the skeletal structure. The predetermined part may be, for example, the right foot A81 or the left foot A82, or may be other parts. In addition, when the right foot A81 or the left foot A82 is hidden, the height of the person may be estimated using an existing method, and the position of the right foot A81 or the left foot A82 may be estimated from the estimated height and the position of the head A1. There is no particular restriction on the means for identifying the position inside the moving body from the position in the image of each person detected in the image (the means for converting from the image coordinate system to the world coordinate system), and any conventional technology may be adopted.

[0044] -Third person information example- The person information generating unit 11 can generate person information indicating a relative positional relationship between the position of the equipment and the position of the person, based on the equipment information acquired by the equipment information acquiring unit 12. The relative positional relationship is, for example, a distance in a world coordinate system. For example, the person information generating unit 11 can obtain a distance between the position of a predetermined part of the person's body and the position of each piece of equipment indicated by the equipment information. Note that, in the case where the space occupied by each piece of equipment in the world coordinate system is indicated in the equipment information, the distance between the position of a predetermined point in that space (such as a point that is the shortest distance from the position of a predetermined part of the person's body) and the position of a predetermined part of the person's body may be obtained.

[0045] As an example, the person information generating unit 11 may calculate the distance between the right hand A51 or the left hand A52 and a facility (a strap, a handrail, etc.) held by the hand.

[0046] Additionally, the person information generating unit 11 may calculate the distance between a predetermined part of the person's body (such as the right foot A81 or the left foot A82) and the door.

[0047] -Fourth example of personal information- The person information generating unit 11 can identify the posture of a person based on the detected skeletal structure of the person and generate person information indicating the posture of the person. The postures to be identified include standing, sitting, lying down (sleeping, collapsed, etc.). The person information generating unit 11 can estimate the posture of each person based on the above-mentioned feature amount of the person's skeletal structure. For example, the posture of a person of each skeletal structure may be estimated using an estimation model generated by machine learning based on learning data in which the feature amount of the skeletal structure is associated with identification information (such as name) of the posture.

[0048] - Fifth example of personal information - The person information generating unit 11 can identify a reference point based on a plurality of predetermined points in the detected skeletal structure of the person, and generate person information further indicating a change in the position of the reference point (coordinates in the world coordinate system) over time. The change in the position of the reference point over time is used to determine, for example, whether the person is moving or simply swaying. For this reason, the plurality of points used to identify the reference point are suitable for the detection. For example, the plurality of points used to identify the reference point may be key points included in the lower body (right hip A61, left hip A62, right knee A71, left knee A72, right foot A81, and left foot A82 in FIG. 3), or may be all key points of the skeletal structure, or may be others.

[0049] The reference point can be determined by any calculation formula based on the coordinates of multiple points used to specify the reference point. For example, x coordinate, y coordinate, and z For each coordinate, an average value of the coordinate values ​​at a plurality of points used to specify the reference point may be calculated and used as the coordinate of the reference point.

[0050] The detection unit 13 detects a dangerous state occurring inside the moving body based on the facility information and the person information. The dangerous state is at least one of a state in which an accident (such as a fall) is likely to occur inside the moving body, and an accident (such as a fall) that has occurred inside the moving body.

[0051] The detection unit 13 may obtain mobile object state information indicating the state of the mobile object, and detect the dangerous state by a detection method according to the state of the mobile object. The detected mobile object state includes at least one of moving, stopped, wipers in use, turn signals on, door open, door closed, window open, window closed, and speed.

[0052] The detection unit 13 may detect these states by analyzing an image captured so as to include these facilities, or may obtain information indicating the state of each facility from a mobile system (such as an Electronic Control Unit) that controls the operation of these facilities. The following are some examples: For example, it may be possible to detect whether a moving object is moving or stopped based on whether the scenery outside the moving object seen through a window or door is changing. Detection of whether the wipers are in use, the turn signal is on, the door is open, the door is closed, the window is open, or the window is closed can be realized by detecting the external characteristics of each piece of equipment specific to each state from an image.

[0053] Next, an example of a method for detecting a dangerous state will be described. The detection unit 13 can detect a dangerous state by using at least one of a plurality of detection methods exemplified below.

[0054] -First detection method- When the state of the moving object is in a predetermined state, the detection unit 13 can detect the presence of a person standing without holding onto a strap or handrail as a dangerous state.

[0055] The predetermined state of the moving body is, for example, “moving.” Even when the moving body is simply moving, a sudden steering operation or deceleration operation may occur at any time for safety reasons, and the inertial force caused by such an operation may be applied to a person in the moving body, causing an accident such as a fall.

[0056] The predetermined state of the moving body may also be "moving" and "with blinker on." The blinker on indicates that the moving body is about to turn right or left, to slow down to stop, to accelerate to overtake, or that such an action is currently occurring, and the inertial force caused by this may be applied to a person in the moving body, causing an accident such as a fall.

[0057] The predetermined state of the moving object may also be "moving" and "wiper in use." When the wipers are in use, it may be raining, and the inside of the moving object may be wet and slippery. This may cause an accident.

[0058] The predetermined state of the moving object may also be "moving," "the turn signal is on," and "the wipers are in use."

[0059] A standing person can be detected based on the person's posture information. Whether or not the person is holding a strap or a handrail can be determined based on the distance between the person's right hand A51 or left hand A52 and the facility (the nearest strap, handrail, etc.). If the distance is equal to or less than a threshold, the detection unit 13 determines that the person is holding the facility, and if the distance is greater than the threshold, the detection unit 13 determines that the person is not holding the facility.

[0060] The threshold value may be changed based on the state of the moving object. For example, when the moving object is in a state where an accident is relatively likely to occur, such as "the blinker is on," "the wipers are in use," or "the moving object is traveling at a speed of V or more," the threshold value may be set smaller than in other states. The distance between the person's right hand A51 or left hand A52 and the equipment may be different in a state where the strap or handrail is firmly held and a state where, for example, the fingertips are hooked on the strap or handrail. Specifically, the distance is smaller in a state where the strap or handrail is firmly held. Therefore, when the moving object is in a state where an accident is relatively likely to occur, the threshold value may be set smaller than in other states, and the "state where the strap or handrail is firmly held" may be detected as the "state where the strap or handrail is being held."

[0061] -Second detection method- When the moving object is in a predetermined state, the detection unit 13 can detect the presence of a person moving without holding onto a strap or a handrail as a dangerous state. The predetermined state of the moving object is as described in the first detection method.

[0062] Detection of a moving person can be realized based on whether or not the time change of the reference point described in the fifth example of person information satisfies a predetermined condition. The predetermined condition is that there is a feature (design item) that appears in the time change when moving. Whether or not the person is holding onto a strap or handrail can be identified in the same manner as in the first detection method.

[0063] -Third detection method- The detection unit 13 can detect, as a dangerous state, a state in which there is a person standing still near the door for a predetermined time (design matter) or more when the state of the moving object is "door open state." When the door is open, it is considered that passengers will get on and off through the door. On the other hand, a person standing still near the door for a predetermined time or more in this state is considered to have no intention of getting on or off and is simply standing there. If such a person is near the door, an accident such as a collision may occur between that person and people getting on or off.

[0064] Detection of a person standing still near a door for a predetermined time (design matter) or longer can be realized based on the change in the distance between the door and the person over time.

[0065] -The fourth detection method- The detection unit 13 can detect the presence of a person sitting or lying down anywhere other than a seat as a dangerous state. A person sitting or lying down anywhere other than a seat may be a person who has fallen. Also, a person sitting or lying down anywhere other than a seat may be in poor health and may easily fall or collide with equipment due to the shaking of the moving body. Also, a person sitting or lying down anywhere other than a seat may be an obstacle to other people, and an accident such as a collision may occur between that person and other people.

[0066] Detection of people sitting and lying down can be achieved based on the person's posture information, and sitting or lying down outside the seat can be identified based on the position of the person relative to the seat.

[0067] When the detection unit 13 detects a dangerous state occurring inside the moving object, the output unit 14 outputs information indicating that fact. The output unit 14 may further output information indicating the content of the detected dangerous state (such as the detection of a person standing without holding onto a strap or handrail).

[0068] For example, the above information may be notified to a manager of the moving object (such as a driver or conductor), or may be notified to other people (such as passengers) in the moving object. The notification means is not particularly limited, and can be realized via any output device such as a display, a speaker, a warning lamp, a projection device, etc.

[0069] In addition, if the moving object has an automatic driving function, the information may be input to a system that controls the automatic driving. The system controls the operation of the moving object based on the input information.

[0070] Next, an example of the process flow of the detection device 10 will be described with reference to FIGS.

[0071] The process shown in the flowchart of FIG. 7 shows a process flow in the case where generation of facility information based on an image generated by a camera capturing an image of the inside of a moving object is not executed (the first acquisition example described above).

[0072] When the detection device 10 acquires an image generated by a camera capturing an image of the inside of a moving object (S10), the person information generation unit 11 analyzes the image and generates person information (S11). Then, the person information generation unit 11 inputs the generated person information to the detection unit 13 (S12). Thereafter, the same process is repeated. Note that the details of these processes have been described above, so a description thereof will be omitted here.

[0073] The processing shown in the flowchart of Figure 8 shows the processing flow when facility information is generated based on an image generated by a camera that captures the inside of a moving body (the second acquisition example or the third acquisition example described above).

[0074] When the detection device 10 acquires an image generated by a camera capturing an image of the inside of a moving object (S20), the person information generation unit 11 analyzes the image to generate person information, and the equipment information acquisition unit 12 analyzes the image to generate equipment information (S21). Then, the person information generation unit 11 inputs the generated person information to the detection unit 13, and the equipment information acquisition unit 12 inputs the generated equipment information to the detection unit 13 (S22). Thereafter, the same process is repeated. Note that the details of these processes have been described above, so a description thereof will be omitted here.

[0075] The flowchart of FIG. 9 is executed by the detection unit 13 and the output unit 14. The detection unit 13 executes a process of detecting a dangerous state occurring inside the moving object based on the person information generated by the person information generation unit 11 and the facility information acquired by the facility information acquisition unit 12 (S30). Then, when a dangerous state is detected by the detection unit 13 (Yes in S31), the output unit 14 outputs information indicating that (S32). The output unit 14 can notify the manager of the moving object (driver, conductor, etc.) and / or other persons (passengers, etc.) present in the moving object of the above through any output device such as a display, speaker, warning lamp, projector, etc. After that, the same process is repeated.

[0076] "Action and effect" The detection device 10 detects a dangerous state occurring inside the moving body based on a relative positional relationship between the position of the equipment and the position of the person calculated based on equipment information indicating the position of the equipment present inside the moving body and person information indicating the position of the person present inside the moving body. By making a judgment based on the relative positional relationship between the position of the equipment and the position of the person, it becomes possible to accurately detect a dangerous state occurring inside the moving body.

[0077] In addition, the detection device 10 can detect a dangerous state occurring inside a moving object based on the relative positional relationship between a predetermined part of a person's body and the equipment (the relative positional relationship between a hand and a strap or a handrail). Such a detection device 10 can accurately detect a dangerous state occurring inside a moving object.

[0078] In addition, the detection device 10 detects the skeletal structure of a person included in an image (image generated using an image sensor) captured inside a moving object by a camera or the like, and identifies matters related to the person, such as the person's position and posture, based on the detected skeletal structure of the person. As a method for detecting the position, posture, etc. of a person based on the skeletal structure of a person included in an image generated by a camera or the like, several methods that are robust against changes in lighting conditions are known. By detecting the skeletal structure of a person included in an image generated by a camera or the like and detecting the position, posture, etc. of the person based on the skeletal structure of the person, as in the detection device 10, an analysis that is robust against changes in lighting conditions is realized. It is then possible to accurately detect a dangerous situation occurring inside a moving object regardless of the lighting conditions.

[0079] Furthermore, the detection device 10 can detect a dangerous state occurring inside a moving body using a detection method according to the state of the moving body. Some states that can occur inside a moving body become dangerous or not dangerous depending on the state of the moving body. According to the detection device 10 that can detect a dangerous state occurring inside a moving body using a detection method according to the state of the moving body, a truly dangerous state can be accurately detected as a dangerous state occurring inside the moving body.

[0080] <Second embodiment> Hereinafter, the second embodiment will be described with reference to the drawings. In the second embodiment, a specific example of calculating the feature amount of the skeletal structure in the first embodiment will be described. In the second embodiment, the feature amount is obtained by normalizing using the person's height. The rest is the same as in the first embodiment.

[0081] The person information generating unit 11 calculates (estimates) the standing height (called height pixel count) of the person in the image based on the detected skeletal structure. The height pixel count can also be said to be the height of the person in the image (the length of the person's entire body in the two-dimensional image space). The person information generating unit 11 obtains the height pixel count (number of pixels) from the length of each bone of the detected skeletal structure (length in the two-dimensional image space).

[0082] In the following examples, concrete examples 1 to 3 are used as a method for calculating the height pixel number. Any of the concrete examples 1 to 3 may be used, or a combination of a plurality of arbitrarily selected methods may be used. In concrete example 1, the height pixel number is calculated by adding up the lengths of the bones from the head to the feet among the bones of the skeletal structure. If the top of the head and the feet are not detected, correction can be performed by multiplying a constant as necessary. In concrete example 2, the height pixel number is calculated using a human body model that indicates the relationship between the length of each bone and the length of the whole body (height in a two-dimensional image space). In concrete example 3, the height pixel number is calculated by fitting (applying) a three-dimensional human body model to a two-dimensional skeletal structure.

[0083] The person information generating unit 11 normalizes the skeletal structure (skeletal information) of the person based on the calculated number of pixels of the height of the person. The person information generating unit 11 normalizes the height of each key point (feature point) included in the skeletal structure on the image by the number of pixels of the height. For example, the height direction is the up-down direction (Y axis direction) in the two-dimensional coordinate (XY coordinate) space of the image. In this case, the height of the key point can be obtained from the value (number of pixels) of the Y coordinate of the key point. Alternatively, the height direction may be the direction of the vertical projection axis (vertical projection direction) obtained by projecting the direction of the vertical axis perpendicular to the ground (reference plane) in the three-dimensional coordinate space of the real world onto the two-dimensional coordinate space. In this case, the height of the key point can be obtained from the value (number of pixels) along the vertical projection axis obtained by projecting the axis perpendicular to the ground in the real world onto the two-dimensional coordinate space based on the camera parameters. Note that the camera parameters are imaging parameters of the image, and for example, the camera parameters are the attitude, position, imaging angle, focal length, etc. of the camera. A camera captures an object whose length and position are known in advance, and the camera parameters can be obtained from the image. Distortion occurs at both ends of the captured image, and the vertical direction in the real world may not match the up-down direction of the image. In response to this, the parameters of the camera that captured the image can be used to determine the degree to which the vertical direction in the real world is tilted in the image. Therefore, by normalizing the value of the keypoint along the vertical projection axis projected into the image based on the camera parameters by height, the keypoint can be characterized by taking into account the deviation between the real world and the image. The left-right direction (horizontal direction) is the left-right direction (X-axis direction) in the two-dimensional coordinate (XY coordinate) space of the image, or the direction parallel to the ground in the three-dimensional coordinate space of the real world projected into the two-dimensional coordinate space.

[0084] Here, specific examples 1 to 3 of the height pixel number calculation process will be described.

[0085] -Example 1- In specific example 1, the height pixel number is obtained using the lengths of the bones from the head to the feet. In specific example 1, the person information generating unit 11 obtains the length of each bone and sums up the obtained lengths of each bone.

[0086] Then, the person information generating unit 11 obtains the length of the bones on the two-dimensional image from the head to the feet of the person, and obtains the height pixel number. That is, from the image in which the skeletal structure is detected, the length (number of pixels) of each of the bones in FIG. 10, bone B1 (length L1), bone B51 (length L21), bone B61 (length L31), and bone B71 (length L41), or bone B1 (length L1), bone B52 (length L22), bone B62 (length L32), and bone B72 (length L42), is obtained. The length of each bone can be obtained from the coordinates of each key point in the two-dimensional image. The total of these, L1+L21+L31+L41, or L1+L22+L32+L42, is multiplied by a correction constant to calculate the height pixel number (h). When both values ​​can be calculated, for example, the longer value is set as the height pixel number. That is, each bone will be the longest in the image when it is photographed from the front, and will appear shorter when tilted in the depth direction relative to the camera. Therefore, it is more likely that a longer bone was photographed from the front, and is closer to the true value. For this reason, it is preferable to select the longer value.

[0087] In the example of Fig. 11, bone B1, bones B51 and B52, bones B61 and B62, and bones B71 and B72 are detected without overlapping. The totals of these bones, L1+L21+L31+L41 and L1+L22+L32+L42, are calculated, and the height pixel number is determined by multiplying the detected bone L1+L22+L32+L42 on the left foot side, which has the longer length, by a correction constant.

[0088] 12, bone B1, bone B51 and bone B52, bone B61 and bone B62, and bone B71 and bone B72 are detected, and bones B61 and B71 of the right foot overlap with bones B62 and B72 of the left foot. The totals of these bones, L1+L21+L31+L41 and L1+L22+L32+L42, are calculated, and the height pixel number is determined by multiplying the detected bone length L1+L21+L31+L41 on the right foot side, which is longer, by a correction constant.

[0089] 13, bone B1, bone B51 and bone B52, bone B61 and bone B62, and bone B71 and bone B72 are detected, and bones B61 and B71 of the right foot overlap with bones B62 and B72 of the left foot. The totals of these bones, L1+L21+L31+L41 and L1+L22+L32+L42, are calculated, and the height pixel number is determined by multiplying the detected bone length L1+L22+L32+L42 on the left foot side, which is longer, by a correction constant.

[0090] In Example 1, height can be calculated by adding up the lengths of the bones from head to toe, so the height pixel count can be calculated in a simple manner. In addition, because it is sufficient to detect at least the skeleton from head to toe using skeleton detection technology using machine learning, the height pixel count can be estimated with high accuracy even when the entire person is not necessarily captured in the image, such as when the person is crouching.

[0091] -Example 2- In specific example 2, the height pixel number is calculated using a two-dimensional skeleton model that indicates the relationship between the length of bones included in a two-dimensional skeleton structure and the length of the entire body of a person in a two-dimensional image space.

[0092] FIG. 14 shows a human body model (two-dimensional skeletal model) 301 used in the concrete example 2, which shows the relationship between the length of each bone in the two-dimensional image space and the length of the whole body in the two-dimensional image space. As shown in FIG. 14, the relationship between the length of each bone of an average person and the length of the whole body (the ratio of the length of each bone to the length of the whole body) is associated with each bone of the human body model 301. For example, the length of the head bone B1 is the length of the whole body×0.2 (20%), the length of the right hand bone B41 is the length of the whole body×0.15 (15%), and the length of the right foot bone B71 is the length of the whole body×0.25 (25%). By storing such information of the human body model 301 in the database 110, the average whole body length can be obtained from the length of each bone. In addition to the human body model of an average person, a human body model may be prepared for each person's attributes such as age, sex, and nationality. This allows the length of the whole body (height) to be obtained appropriately according to the person's attributes.

[0093] In specific example 2, the person information generation unit 11 acquires the length of each bone. The person information generation unit 11 acquires the lengths of all bones (lengths in two-dimensional image space) in the detected skeletal structure. FIG. 15 shows an example in which a person in a crouching state is imaged from diagonally behind to the right and a skeletal structure is detected. In this example, the person's face and left side are not captured, so the head bone and the left arm and left hand bones cannot be detected. For this reason, the lengths of the detected bones B21, B22, B31, B41, B51, B52, B61, B62, B71, and B72 are acquired.

[0094] Next, the person information generating unit 11 calculates the height pixel number from the length of each bone based on the human body model. The person information generating unit 11 refers to a human body model 301 showing the relationship between each bone and the length of the whole body as shown in Fig. 14, and calculates the height pixel number from the length of each bone. For example, since the length of the right hand bone B41 is the whole body length x 0.15, the height pixel number based on the bone B41 is calculated by dividing the length of the bone B41 by 0.15. Also, since the length of the right foot bone B71 is the whole body length x 0.25, the height pixel number based on the bone B71 is calculated by dividing the length of the bone B71 by 0.25.

[0095] The human body model to be referenced at this time is, for example, an average human body model, but a human body model may be selected according to the attributes of the person, such as age, sex, and nationality. For example, when a person's face is captured in a captured image, the attributes of the person are identified based on the face, and a human body model corresponding to the identified attributes is referenced. The attributes of the person can be recognized from the facial features of the image by referring to information obtained by machine learning of the face for each attribute. Also, when the attributes of the person cannot be identified from the image, an average human body model may be used.

[0096] Furthermore, the height pixel count calculated from the bone lengths may be corrected by camera parameters. For example, when a person is photographed looking down from a high position with a camera, the horizontal length of the shoulder width bones in the two-dimensional skeletal structure is not affected by the camera's inclination angle, but the vertical length of the neck-waist bones decreases as the camera's inclination angle increases. As a result, the height pixel count calculated from the horizontal length of the shoulder width bones tends to be larger than the actual height pixel count. Therefore, by utilizing camera parameters, the angle at which the camera is looking down on the person can be known, and this inclination angle information can be used to correct the two-dimensional skeletal structure to one as if photographed from the front. This allows the height pixel count to be calculated more accurately.

[0097] Next, the person information generating unit 11 calculates the optimum value of the height pixel number. The person information generating unit 11 calculates the optimum value of the height pixel number from the height pixel number obtained for each bone. For example, as shown in FIG. 16, a histogram of the height pixel number obtained for each bone is generated, and the largest height pixel number is selected from among the plurality of height pixel numbers obtained based on the plurality of bones. In other words, a height pixel number longer than the others is selected from among the plurality of height pixel numbers obtained based on the plurality of bones. For example, the top 30% is set as a valid value, and the height pixel numbers of bones B71, B61, and B51 are selected in FIG. 16. The average of the selected height pixel numbers may be set as the optimum value, or the largest height pixel number may be set as the optimum value. Since the height is calculated from the length of the bone in the two-dimensional image, if the bone cannot be imaged from the front, that is, if the bone is imaged while tilted in the depth direction as viewed from the camera, the length of the bone will be shorter than when it is imaged from the front. In this case, a value with a large height pixel number is more likely to be imaged from the front than a value with a small height pixel number, and is therefore a more likely value, so a larger value is set as the optimum value.

[0098] In Example 2, the height pixel number is calculated based on the detected bones of the skeletal structure using a human body model showing the relationship between the bones in the two-dimensional image space and the length of the whole body, so that even if all the bones from the head to the feet cannot be obtained, the height pixel number can be calculated from some of the bones. In particular, the height pixel number can be estimated with high accuracy by adopting the larger value among the values ​​calculated from multiple bones.

[0099] -Example 3- In a third specific example, the detected two-dimensional skeletal structure is fitted to a three-dimensional human body model (three-dimensional skeletal model), and a whole-body skeletal vector is obtained using the height pixel count of the fitted three-dimensional human body model.

[0100] In the specific example 3, the person information generating unit 11 first calculates camera parameters based on images captured by the camera. The person information generating unit 11 extracts an object whose length is known in advance from a plurality of images captured by the camera, and obtains the camera parameters from the size (number of pixels) of the extracted object. Note that the camera parameters may be obtained in advance and acquired as necessary.

[0101] Next, the person information generating unit 11 adjusts the position and height of the three-dimensional human body model. The person information generating unit 11 prepares a three-dimensional human body model for calculating the height pixel number for the detected two-dimensional skeletal structure, and places it in the same two-dimensional image based on the camera parameters. Specifically, the "relative positional relationship between the camera and the person in the real world" is specified from the camera parameters and the two-dimensional skeletal structure. For example, assuming that the camera position is at coordinates (0,0,0), the coordinates (x,y,z) of the position where the person is standing (or sitting) are specified. Then, the two-dimensional skeletal structure and the three-dimensional human body model are superimposed by assuming an image when the three-dimensional human body model is placed at the same position (x,y,z) as the specified person and captured.

[0102] FIG. 17 shows an example in which a person who is crouching is imaged from the left diagonal front and a two-dimensional skeletal structure 401 is detected. The two-dimensional skeletal structure 401 has two-dimensional coordinate information. It is preferable that all bones are detected, but some bones may not be detected. For this two-dimensional skeletal structure 401, a three-dimensional human body model 402 as shown in FIG. 18 is prepared. The three-dimensional human body model (three-dimensional skeletal model) 402 has three-dimensional coordinate information and is a skeletal model having the same shape as the two-dimensional skeletal structure 401. Then, as shown in FIG. 19, the prepared three-dimensional human body model 402 is arranged and superimposed on the detected two-dimensional skeletal structure 401. In addition to the superimposition, the height of the three-dimensional human body model 402 is adjusted to match the two-dimensional skeletal structure 401.

[0103] The three-dimensional human body model 402 prepared at this time may be a model in a state close to the posture of the two-dimensional skeletal structure 401, as shown in Fig. 19, or may be a model in an upright state. For example, a three-dimensional human body model 402 in an estimated posture may be generated using a technology that uses machine learning to estimate a posture in three-dimensional space from a two-dimensional image. By learning information on the joints in the two-dimensional image and the joints in three-dimensional space, it is possible to estimate a three-dimensional posture from a two-dimensional image.

[0104] Next, the person information generating unit 11 fits the three-dimensional human body model to the two-dimensional skeletal structure. As shown in FIG. 20, the person information generating unit 11 deforms the three-dimensional human body model 402 so that the postures of the three-dimensional human body model 402 and the two-dimensional skeletal structure 401 match in a state where the three-dimensional human body model 402 is superimposed on the two-dimensional skeletal structure 401. That is, the height, body orientation, and joint angles of the three-dimensional human body model 402 are adjusted and optimized so that there is no difference with the two-dimensional skeletal structure 401. For example, operations such as rotating the joints of the three-dimensional human body model 402 within the range of human movement, rotating the entire three-dimensional human body model 402, and adjusting the size of the entire three-dimensional human body model 402 are performed. The fitting of the three-dimensional human body model to the two-dimensional skeletal structure is performed in a two-dimensional space (two-dimensional coordinates). In other words, a 3D human body model is mapped onto a 2D space, and the 3D human body model is optimized to a 2D skeletal structure, taking into consideration how the deformed 3D human body model changes in the 2D space (image).

[0105] Next, the person information generation unit 11 calculates the height pixel number of the fitted three-dimensional human body model. When the difference between the three-dimensional human body model 402 and the two-dimensional skeletal structure 401 disappears and the postures match as shown in FIG. 21, the person information generation unit 11 calculates the height pixel number of the three-dimensional human body model 402 in that state. With the optimized three-dimensional human body model 402 in an upright state, the length of the entire body in two-dimensional space is calculated based on the camera parameters. For example, the height pixel number is calculated based on the bone length (number of pixels) from the head to the feet when the three-dimensional human body model 402 is upright. As in the specific example 1, the bone lengths from the head to the feet of the three-dimensional human body model 402 may be summed up.

[0106] In specific example 3, a 3D human body model is fitted to a 2D skeletal structure based on the camera parameters, and height pixel count is calculated based on the 3D human body model. This makes it possible to estimate height pixel count with high accuracy even when not all bones are viewed from the front, i.e., even when all bones are viewed at an angle, resulting in a large error.

[0107] Next, the normalization process will be described. The person information generating unit 11 performs the normalization process following the height pixel number calculation process. In the normalization process, the person information generating unit 11 calculates the key point height. The person information generating unit 11 calculates the key point height (number of pixels) of all key points included in the detected skeletal structure. The key point height is the length (number of pixels) in the height direction from the lowest end of the skeletal structure (the key point of any foot) to the key point. Here, as an example, the key point height is obtained from the Y coordinate of the key point in the image. As described above, the key point height may be obtained from the length in the direction along the vertical projection axis based on the camera parameters. For example, in the example of FIG. 10, the height (yi) of the neck key point A2 is the value obtained by subtracting the Y coordinate of the right foot key point A81 or the left foot key point A82 from the Y coordinate of the key point A2.

[0108] Next, the person information generating unit 11 specifies a reference point for normalization. The reference point for normalization is a point that serves as a reference for expressing the relative height of the key point. The reference point for normalization may be set in advance, or may be selected by the user. The reference point for normalization is preferably the center of the skeletal structure or higher than the center (upper in the vertical direction of the image), and may be, for example, the coordinates of the key point of the neck. Note that the reference point for normalization is not limited to the neck, and the coordinates of the head or other key points may be used. The reference point for normalization is not limited to the key point, and any coordinates (such as the center coordinates of the skeletal structure) may be used.

[0109] Next, the person information generating unit 11 normalizes the keypoint height (yi) by the height pixel number. The person information generating unit 11 normalizes each keypoint using the keypoint height, the reference point for normalization, and the height pixel number of each keypoint. Specifically, the person information generating unit 11 normalizes the relative height of the keypoint with respect to the reference point for normalization by the height pixel number. Here, as an example focusing on only the height direction, only the Y coordinate is extracted, and normalization is performed with the reference point for normalization being the neck keypoint. Specifically, the Y coordinate of the reference point for normalization (neck keypoint) is set as (yc), and the feature amount (normalized value) is obtained using the following formula (1). Note that when a vertical projection axis based on the camera parameters is used, (yi) and (yc) are converted into values ​​in the direction along the vertical projection axis.

[0110]

number

[0111] For example, if there are 18 keypoints, the coordinates of the 18 keypoints (x0, y0), (x1, y1), ... (x17, y17) are converted into 18-dimensional features using the above formula (1) as follows:

[0112]

number

[0113] FIG. 22 shows an example of the feature amount of each key point obtained by the person information generating unit 11. In this example, the neck key point A2 is used as the reference point for normalization, so the feature amount of the key point A2 is 0.0, and the feature amounts of the right shoulder key point A31 and the left shoulder key point A32, which are at the same height as the neck, are also 0.0. The feature amount of the head key point A1, which is higher than the neck, is -0.2. The feature amount of the right hand key point A51 and the left hand key point A52, which are lower than the neck, is 0.4, and the feature amount of the right foot key point A81 and the left foot key point A82 are 0.9. If the person raises his / her left hand from this state, the left hand becomes higher than the reference point for normalization as shown in FIG. 23, so the feature amount of the left hand key point A52 is -0.4. On the other hand, since normalization is performed using only the coordinate of the Y axis, the feature amount does not change even if the width of the skeletal structure changes as compared to FIG. 22, as shown in FIG. 24. That is, the feature amount (normalized value) in this embodiment indicates the feature in the height direction (Y direction) of the skeletal structure (key point), and is not affected by changes in the lateral direction (X direction) of the skeletal structure.

[0114] As described above, the detection device 10 of this embodiment achieves the same effects as those of the first embodiment. In addition, the detection device 10 of this embodiment detects a person's skeletal structure from a two-dimensional image, and normalizes each key point of the skeletal structure using the height pixel count (height when standing upright in two-dimensional image space) calculated from the detected skeletal structure. By using this normalized feature amount, it is possible to improve robustness during various processes using the feature amount. In other words, the feature amount of this embodiment is not affected by lateral changes in the person as described above, and is therefore highly robust to changes in the person's orientation and body shape.

[0115] Furthermore, since the detection device 10 of this embodiment can be realized by detecting a person's skeletal structure using a skeletal detection technology such as OpenPose, there is no need to prepare learning data for learning a person's posture, etc. Also, since clear and easy-to-understand features can be obtained by normalizing key points of the skeletal structure, users can be more convinced of the processing results, unlike black-box algorithms such as machine learning.

[0116] <Modification> Modifications that can be applied to the first and second embodiments will be described below.

[0117] -First modified example- In the first and second embodiments, a person's skeletal structure is detected from an image generated by a camera, and various personal information such as the person's position, posture, etc. is generated based on the detected skeletal structure of the person. As a modified example, at least a part (or all) of the personal information may be generated by directly analyzing the image generated by a camera or the like (based on the external characteristics of the person shown in the image) without using the detection result of the person's skeletal structure. In this modified example, although an analysis that is robust to changes in lighting conditions is not realized among the operational effects described in the first and second embodiments, the other operational effects are realized.

[0118] -Second modified example- In the first and second embodiments, the detection device 10 is a device mounted on a moving object. As a modified example, the detection device 10 may be a server installed at a location remote from the moving object. The detection device 10 may communicate with a system mounted on the moving object, acquire an image generated by a camera that captures the inside of the moving object, and transmit a detection result by the detection unit 13 to the system mounted on the moving object via the output unit 14.

[0119] Although the embodiments of the present invention have been described above with reference to the drawings, these are merely examples of the present invention, and various configurations other than those described above can also be adopted.

[0120] In addition, in the flowcharts used in the above description, multiple steps (processes) are described in order, but the order of execution of the steps performed in each embodiment is not limited to the order described. In each embodiment, the order of the steps shown in the figures can be changed to the extent that the content is not affected. In addition, the above-mentioned embodiments can be combined to the extent that the content is not contradictory.

[0121] A part or all of the above-described embodiments can be described as follows, but is not limited to the following. 1. A facility information acquisition means for acquiring facility information indicating the location of a facility present inside a moving object; a person information generating means for identifying a position of a person present inside the moving body based on an image showing the inside of the moving body and generating person information showing the position of the person; a detection means for detecting a dangerous state occurring inside the moving object based on the facility information and the person information; A detection device having the following: 2. A detection device according to claim 1, wherein the person information generation means generates information indicating the position of a predetermined part of the person's body as the person information indicating the position of the person. 3. A detection device according to 1 or 2, wherein the person information generation means generates the person information indicating a relative positional relationship between the position of the equipment and the position of the person based on the equipment information. 4. A detection device described in any one of 1 to 3, wherein the person information generation means detects a skeletal structure of a person included in the image and identifies a position of the person based on the detected skeletal structure of the person. 5. A detection device described in any one of 1 to 4, wherein the person information generation means detects a skeletal structure of a person included in the image, identifies the person's posture based on the detected skeletal structure of the person, and generates the person information further indicating the person's posture. 6. A detection device described in any one of 1 to 5, wherein the person information generation means detects a skeletal structure of a person included in the image, identifies reference points based on a predetermined number of locations in the skeletal structure of the person, and generates the person information further indicating changes in the positions of the reference points over time. 7. A detection device according to any one of 1 to 6, wherein the detection means acquires moving body status information indicating the status of the moving body, and detects the hazardous state using a detection method according to the status of the moving body. 8. A detection device according to any one of 1 to 7, wherein the equipment present inside the moving body includes at least one of a strap, a handrail, and a seat. 9. The computer: Acquire facility information indicating the location of the facility inside the mobile object; Identifying a position of a person present inside the moving body based on an image showing the inside of the moving body, and generating person information showing the position of the person; A detection method for detecting a dangerous state occurring inside the moving body based on the facility information and the person information. 10. Computer A facility information acquisition means for acquiring facility information indicating a location of a facility present inside the mobile object; a person information generating means for identifying a position of a person present inside the moving body based on an image showing the inside of the moving body, and generating person information showing the position of the person; a detection means for detecting a dangerous state occurring inside the moving object based on the facility information and the person information; A program that functions as a [Explanation of symbols]

[0122] 10. Detection Device 11 Person information generation section 12 Equipment information acquisition department 13 Detection section 1A Processor 2A Memory 3A input / output I / F 4A Peripheral circuit 5A Bus 401 2D Skeleton Structure 402 3D Human Body Model

Claims

1. A facility information acquisition means for acquiring facility information indicating a location of a facility present inside the mobile body; a person information generating means for identifying a position of a person present inside the moving body based on an image showing the inside of the moving body and generating person information showing the position of the person; a detection means for detecting a dangerous state occurring inside the moving object based on the facility information and the person information; A detection device having the following:

2. The detection device according to claim 1 , wherein the person information generating means generates information indicating a position of a predetermined part of the person's body as the person information indicating the position of the person.

3. The detection device according to claim 1 , wherein the person information generating means generates the person information indicating a relative positional relationship between a position of the facility and a position of the person, based on the facility information.

4. The detection device according to claim 1 , wherein the person information generating means detects a skeletal structure of a person included in the image, and identifies a position of the person based on the detected skeletal structure of the person.

5. The detection device according to any one of claims 1 to 4, wherein the person information generation means detects a skeletal structure of a person included in the image, identifies a posture of the person based on the detected skeletal structure of the person, and generates the person information further indicating the posture of the person.

6. The detection device described in any one of claims 1 to 5, wherein the person information generation means detects a skeletal structure of a person included in the image, identifies reference points based on a predetermined number of locations in the skeletal structure of the person, and generates the person information further indicating changes in the position of the reference points over time.

7. The detection device according to claim 1 , wherein the detection means acquires moving body state information indicating a state of the moving body, and detects the hazardous state using a detection method according to the state of the moving body.

8. The detection device according to claim 1 , wherein the facilities present inside the moving body include at least one of a strap, a handrail, and a seat.

9. The detection means A detection device as described in any one of claims 1 to 8, which estimates the usage status of the equipment by the person based on the position of the equipment inside the moving body and the position of the person inside the moving body, and detects a dangerous state occurring inside the moving body based on the result of the estimation.

10. The computer Acquire facility information indicating the location of the facility inside the mobile object; Identifying a position of a person present inside the moving body based on an image showing the inside of the moving body, and generating person information showing the position of the person; A detection method for detecting a dangerous state occurring inside the moving body based on the facility information and the person information.

11. Computer, A facility information acquisition means for acquiring facility information indicating a location of a facility present inside the mobile object; a person information generating means for identifying a position of a person present inside the moving body based on an image showing the inside of the moving body, and generating person information showing the position of the person; a detection means for detecting a dangerous state occurring inside the moving object based on the facility information and the person information; A program that functions as a

Images