Injured person detection device, injured person detection system, and injured person detection method

Abstract

To provide a casualty detection device, a casualty detection system, and a casualty detection method that reduce the false detection of casualty.SOLUTION: A casualty detection device of a casualty detection system 1 includes: a position extraction unit that extracts a person and a position of a human body part of a person in an image based on the image captured by at least one image capturing unit that captures a monitoring location to be monitored; a fall detection unit that detects a fall of a person based on the human body part extracted by the position extraction unit; a casualty determination unit that determines that the person in question is a casualty in a state of continuous fall when the fall of the person detected by the fall detection unit continues for a predetermined period of time after the fall is detected; and an alarm unit that outputs an alarm when the casualty determination unit determines that the person is a casualty.SELECTED DRAWING: Figure 1

Description

Technical Field

[0001] The present invention relates to an injured person detection device, an injured person detection system, and an injured person detection method.

Background Art

[0002] In normal times or during disasters, as the main means of discovering injured persons who have fallen due to injuries or the like in a large building (for example, a commercial facility, etc.) or outdoors, etc., there are facility patrols by security guards, visual inspection of surveillance camera images by monitors at disaster prevention centers, or reports from people who happened to be near the injured person. Also, since the response to injured persons may require urgency, there is also an injured person detection system that can quickly and surely detect injured persons.

[0003] In the prior art, as an example of an injured person detection system, a method of detecting a person who has fallen due to injuries or the like using shooting data of surveillance cameras installed in a building or outdoors has been proposed. For example, in Patent Document 1, a method of identifying a fallen person by detecting the movement of a person on the platform based on an image of a surveillance camera installed on the platform of a station is presented.

Prior Art Documents

Patent Documents

[0004]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0005] Incidentally, when a fallen person stands up and starts moving immediately after falling, a request for assistance is not necessary, so it is not preferable to detect such a person as an injured person. However, in the method described in Patent Document 1, since a person who starts falling at a speed and acceleration equal to or higher than a predetermined threshold value is detected as a fallen person, a person who stands up and starts moving after falling may be detected as a fallen person, or a person who curls up slowly and then does not move horizontally may not be detected as a fallen person. Thus, in the prior art as described above, there are cases where an injured person cannot be appropriately detected and false detection occurs.

[0006] The present invention has been made in view of such circumstances, and an object thereof is to provide an injured person detection system and an injured person detection method capable of reducing false detection of an injured person.

Means for Solving the Problems

[0007] In order to solve the above-described problems, one aspect of the present invention includes a position extraction unit that extracts the position of a person and a body part of the person in an image based on an image captured by at least one imaging unit that images a monitoring location that is a location to be monitored, a fall detection unit that detects a fall of the person based on the position of the body part extracted by the position extraction unit, an injured person determination unit that determines that the person is an injured person in a continuous fall state when the fall of the person detected by the fall detection unit continues for a predetermined period from the detection of the fall, and an alarm unit that outputs an alarm when the injured person determination unit determines the injured person, and is an injured person detection device.

[0008] Moreover, one aspect of the present invention is an injured person detection system including at least one imaging unit that images a monitoring location that is a location to be monitored and the above-described injured person detection device.

[0009] In addition, one aspect of the present invention includes a step in which a position extraction unit extracts the position of a person and the body parts of the person in the image based on an image captured by at least one imaging unit that captures an area to be monitored, which is the location of the object to be monitored; a step in which a fall detection unit detects a fall of the person based on the position of the body parts of the person; a step in which an injured person determination unit determines that the person is an injured person in a state of continuous fall when the fall of the person detected by the fall detection unit continues for a predetermined period from the detection of the fall; and a step in which an alarm unit outputs an alarm when the injured person determination unit determines the injured person. This is an injured person detection method.

Advantages of the Invention

[0010] According to the present invention, false detection of injured persons can be reduced.

Brief Description of the Drawings

[0011]

Figure 1

Figure 2

Figure 3

Figure 4

Figure 5

Figure 6

Figure 7

Figure 8

Figure 9

Embodiments for Carrying Out the Invention

[0012] Hereinafter, with reference to the drawings, a wounded person detection system according to an embodiment of the present invention will be described.

[0013] (Wounded person detection system) FIG. 1 is a diagram showing an outline of the configuration of a wounded person detection system 1 according to the present embodiment. The wounded person detection system 1 includes at least one imaging unit (10a, 10b) and a wounded person detection device 100.

[0014] The imaging unit (10a, 10b) is a surveillance camera installed indoors or outdoors, images a surveillance location which is the location to be monitored, and adds additional information indicating conditions at the time of imaging (for example, the frame rate of the camera, etc.) to the captured image and transmits it to the wounded person detection device 100 (control unit 30 described later). In the present embodiment, when indicating an arbitrary imaging unit, or when not particularly distinguishing, it will be described as the imaging unit 10. Here, the additional information includes at least camera identification information for identifying the surveillance camera and the imaging time T of the image. In the present embodiment, as a rule for assigning camera identification information, those including "C" in the initial letter are used (for example, C01), and the imaging time T is expressed in the format of YYYYMMDDHHmmss (year / month / day / hour / minute / second). For example, when the imaging time T is 12:00:30 on January 1, 2020, it is expressed as 20200101120030.

[0015] Here, with reference to FIG. 2, the image captured by the imaging unit 10 will be described. FIG. 2 is a diagram showing an example of an image captured by the imaging unit 10 according to the present embodiment. In the image D, there is a person 50a standing upright with both hands and both feet spread on the ground G, and a person 50b lying horizontally on the ground G. In the present embodiment, when indicating an arbitrary person in the image, or when not particularly distinguishing, it will be described as the person 50.

[0016] In addition, in this embodiment, the control unit 30 (the position extraction unit 31 described later) assigns person identification information for identifying each person 50 in the image. In this embodiment, as the rule for assigning person identification information, it is assumed that the initials include "M". In the example shown in FIG. 2, the position extraction unit 31 assigns, for example, M001 to the person 50a and M002 to the person 50b as the person identification information.

[0017] The injured person detection device 100 is a personal computer or a server device that can communicate with the imaging unit 10. The injured person detection device 100 includes a storage unit 20, a control unit 30, and an alarm unit 40.

[0018] The storage unit 20 stores various data used by the control unit 30. The storage unit 20 is configured by, for example, a storage medium such as an HDD (Hard Disk Drive), a flash memory, an EEPROM (Electrically Erasable Programmable Read Only Memory), a RAM (Random Access read / write Memory), a ROM (Read Only Memory), or an arbitrary combination of these storage media. Further, as the storage unit 20, for example, a non-volatile memory can be used. The storage unit 20 includes an imaging unit information storage unit 21, a software storage unit 22, a position data storage unit 23, and a determination condition storage unit 24.

[0019] The imaging unit information storage unit 21 stores a monitoring camera - monitoring location correspondence table. Here, with reference to FIG. 3, the monitoring camera - monitoring location correspondence table will be described. FIG. 3 is a diagram showing an example of a monitoring camera - monitoring location correspondence table. The monitoring camera - monitoring location correspondence table includes camera identification information and the name of the monitoring location of the monitoring camera corresponding to the camera identification information. In the example shown in FIG. 3, it shows that the camera identification information and the monitoring location name are Area A for C01 and Area B for C02.

[0020] The software storage unit 22 stores a skeleton estimation program used by the control unit 30 (the position extraction unit 31, to be described later). The skeleton estimation program may be, for example, open source such as OpenPose or PoseNet, or a user-developed machine learning model that estimates the skeleton from a full-body image of a person.

[0021] The position data storage unit 23 updates and stores the position data created by the control unit 30 (the position extraction unit 31, to be described later). The details of the process in which the position extraction unit 31 extracts the person 50 and the positions of the body parts of the person 50 from the image will be described later. Here, with reference to FIG. 4, the details of the position data will be described. FIG. 4 is a diagram showing an example of the position data of the body parts extracted by the position extraction unit 31 from the image of FIG. 2. The position data includes person identification information, a monitoring location name, and the positions of the body parts in the image at the imaging time T. Here, the position in the image is shown as a two-dimensional position coordinate having, for example, the position coordinate in the horizontal direction (lateral direction) of the image and the position coordinate in the vertical direction (vertical direction) of the image as components. In the following description, the position coordinate in the horizontal direction of the image is referred to as position X, and the position coordinate in the vertical direction of the image is referred to as position Y. In the example shown in FIG. 4, for the person 50a (person identification information is M001), the monitoring location name and the positions of the body parts at the imaging time T are area A and R1(T), respectively. Also, for the person 50b (person identification information is M002), the monitoring location name and the positions of the body parts at the imaging time T are area A and R2(T), respectively. Here, R1(T) and R2(T) include the positions of all the body parts extracted from the image of FIG. 2.

[0022] The determination condition storage unit 24 updates and stores the human body frame generated by the control unit 30 (the fall detection unit 32, to be described later). The details of the process in which the fall detection unit 32 generates the human body frame will be described later. In addition, the determination condition storage unit 24 updates and stores the fall start time T0 and the movable area set by the control unit 30 (the injured person determination unit 33, to be described later). Note that, as an initial value, for example, 000000000000 (in the YYYYMMDDhhmmss format) is set for the fall start time T0. The details of the process in which the injured person determination unit 33 sets the fall start time T0 and the movable area will be described later.

[0023] The control unit 30 is, for example, a CPU (Central Processing Unit) that executes the control software of the injured person detection device 100, and comprehensively controls each part of the injured person detection device 100. The control unit 30 includes a position extraction unit 31, a fall detection unit 32, and an injured person determination unit 33.

[0024] Based on the image captured by at least one imaging unit 10 that images the monitoring location, the position extraction unit 31 extracts the person 50 in the image and the positions of the body parts of the person 50 in the image. In the present embodiment, the position extraction unit 31 uses the skeleton extraction program stored in the software storage unit 22 to extract the person 50 in the image and the positions of the body parts of the person 50. In addition, the position extraction unit 31 assigns person identification information for identifying each of the persons 50 extracted from the image.

[0025] Note that the body parts include, for example, the facial parts and the joint parts. Here, with reference to FIG. 5, an example of the extraction of body parts will be described. FIG. 5 is a diagram showing an example of the body parts extracted by the position extraction unit 31 according to the present embodiment. The facial parts are composed of five parts including both eyes (51a, 51b), both ears (52a, 52b), and the nose (53). The joint parts are composed of twelve parts including both shoulders (54a, 54b), both elbows (55a, 55b), both wrists (56a, 56b), both hips (57a, 57b), both knees (58a, 58b), and both ankles (59a, 59b). Here, the subscript a represents the left side, and the subscript b represents the right side. The position extraction unit 31 extracts the positions of the body parts shown in FIG. 5 like this. In addition, in the present embodiment, when the left and right sides of the eyes, ears, shoulders, elbows, wrists, hips, knees, and ankles are not particularly distinguished, they will be described as eyes 51, ears 52, shoulders 54, elbows 55, wrists 56, hips 57, knees 58, and ankles 59.

[0026] Also, the position extraction unit 31 acquires the camera identification information included in the additional information attached to the image and the imaging time T of the image. Further, the position extraction unit 31 refers to the monitoring camera - monitoring location correspondence table stored in the imaging unit information storage unit 21 based on the camera identification information, and determines the monitoring location of the imaging unit 10. For example, when the acquired camera identification information is C01, the position extraction unit 31 refers to the monitoring camera - monitoring location correspondence table shown in FIG. 3 and determines that the monitoring location of the imaging unit 10 is area A. Also, the position extraction unit 31 creates position data including person identification information, the name of the monitoring location, and the position of the human body part in the image at the imaging time T, and updates and stores it in the position data storage unit 23.

[0027] The fall detection unit 32 detects the fall of the person 50 based on the position data stored in the position data storage unit 23. Specifically, the fall detection unit 32 determines that the fall of the person 50 has been detected when any one of the following three fall determination conditions is satisfied, and determines that the fall of the person 50 has not been detected when none of the three fall determination conditions is satisfied.

[0028] Here, the details of the three fall determination conditions will be described. The first fall determination condition is that in a rectangular human body frame representing the contour range of the human body in the image, the ratio of the length in the vertical direction to the length in the horizontal direction of the human body frame is less than a predetermined threshold value (for example, 1.0). The fall detection unit 32 generates a human body frame based on the position data stored in the position data storage unit 23, determines whether the first fall determination condition is satisfied, and updates and stores the generated human body frame in the determination condition storage unit 24. The human body frame is a rectangle passing through the minimum horizontal position Xmin and the maximum horizontal position Xmax of the human body part in the image, and the minimum vertical position Ymin and the maximum vertical position Ymax of the image. The distance between the minimum horizontal position Xmin and the maximum horizontal position Xmax is defined as the horizontal length H, and the distance between the minimum vertical position Ymin and the maximum vertical position Ymax is defined as the vertical length V. FIG. 6 is a diagram showing an example of a human body frame. The human body frame FR is a rectangle passing through the left wrist 56a (the minimum horizontal position Xmin), the right ankle 59b (the maximum horizontal position Xmax), the left shoulder 54a (the maximum vertical position Ymax), and the left ankle 59a (the minimum vertical position Ymin). The distance between the minimum horizontal position Xmin and the maximum horizontal position Xmax is defined as the horizontal length H, and the distance between the minimum vertical position Ymin and the maximum vertical position Ymax is defined as the vertical length V.

[0029] The second fall determination condition is that the vertical position Y of the face part (for example, the nose 53) is lower than the vertical position Y of any one of the shoulder 54, the hip 57, the knee 58, and the ankle 59 included in the joint part. Based on the position data stored in the position data storage unit 23, the fall detection unit 32 determines whether the second fall determination condition is satisfied. For example, as shown in FIG. 6, when the vertical position Y of the face part is lower than the vertical position Y of the left shoulder 54a or the left hip 57a, the fall detection unit 32 determines that the second fall determination condition is satisfied.

[0030] The third fall determination condition is that the horizontal position X of the center of gravity of the human body in the image is not between the horizontal position X of the left ankle 59a and the horizontal position X of the right ankle 59b. Based on the position data stored in the position data storage unit 23, the fall detection unit 32 calculates the position of the center of gravity of the human body and determines whether the third fall determination condition is satisfied. In this embodiment, the center of gravity of the human body is assumed to be at a distance of L / 3 from the hip side when the length of the straight line connecting the midpoint P1 of both shoulders (54a, 54b) and the midpoint P2 of both hips (57a, 57b) is L. FIG. 7 is a diagram showing the position of the center of gravity of the human body calculated based on the image of FIG. 6. The center of gravity P0 of the human body is a point on a straight line connecting the midpoint P1 of both shoulders (54a, 54b) and the midpoint P2 of both hips (57a, 57b), and is a point at a distance of L / 3 from the midpoint P2. In this case, since the lateral position X of the center of gravity P0 of the human body is not between the lateral position X of the left ankle 59a and the lateral position X of the right ankle 59b, the fall detection unit 32 determines that the third fall determination condition is satisfied.

[0031] When the fall detection unit 32 detects a fall of the person 50, the injured person determination unit 33 determines whether the detected fall is a new fall that has newly occurred. The details of this determination process will be described later.

[0032] In addition, when the injured person determination unit 33 determines that the detected fall is a new fall, it sets the fall start time T0 to the imaging time T and updates and stores it in the determination condition storage unit 24. Here, the fall start time T0 is the imaging time T of the image in which a new fall is detected. In addition, as will be described in detail later, when the detected fall is not continuous or when the detected fall has continued for a predetermined period or more, the injured person determination unit 33 sets the start time T0 to the initial value and updates and stores it in the determination condition storage unit 24.

[0033] Furthermore, when the injured person determination unit 33 determines that the detected fall is a new fall, it sets a movable area (an example of a predetermined area) used for determining the continuation of the fall and updates and stores it in the determination condition storage unit 24. Specifically, the injured person determination unit 33 sets a movable area for each human body part based on the position data stored in the position data storage unit 23 and the human body frame stored in the determination condition storage unit 24. Here, the movable area is a square centered on the position of the human body part at the fall start time, and the side length of the square is the length H of the long side of the human body frame reduced by a predetermined magnification (for example, one tenth). Here, the movable area will be described with reference to FIG. 8. FIG. 8 is a diagram showing an example of a movable area of a human body part extracted from the image of FIG. 6. For each of the 13 human body parts, a square movable area S is set with a length of one side being 1 / 10 of the length H of the long side of the human body frame FR (see FIG. 6).

[0034] When the fall of the person 50 detected by the fall detection unit 32 continues for a predetermined period from the detection of a new fall (fall start time T0), the injured person determination unit 33 determines that the person 50 is an injured person in a state of continued fall. At this time, the injured person determination unit 33 transmits an injured person detection signal indicating the detection of an injured person to the alarm unit 40. Specifically, when the movement of a predetermined part (for example, 12 joint parts) among the 17 human body parts is within a predetermined movable area within a predetermined period (for example, within 10 seconds) from the fall start time T0, the injured person determination unit 33 determines that the person 50 is an injured person in a state of continued fall. Note that the predetermined period may be any time, and for example, it may be determined based on the time during which the posture after falling continues to the extent that it is considered that the person cannot stand up immediately even if they fall and requires assistance. On the other hand, when the fall of the person 50 detected by the fall detection unit 32 has not continued for a predetermined period from the detection of a new fall, the injured person determination unit 33 determines that the person 50 is not in a state of continued fall. In this embodiment, an injured person refers to a person who maintains a substantially the same falling posture as the fall start time T0 for a predetermined period, and for example, a person who is injured and unable to move when falling is assumed.

[0035] The alarm unit 40 has a function of outputting an alarm when it is determined in the determination result of the injured person determination unit 33 that the person 50 in the image is an injured person, and is, for example, an alarm having a function of emitting an alarm sound. The alarm unit 40. When receiving an injured person detection signal from the injured person determination unit 33, it emits an alarm sound notifying that an injured person has occurred in the monitoring area. Further, the alarm unit 40 may be provided with a communication function such as a wireless communication function, and when receiving an injured person detection signal, it transmits a signal notifying the occurrence of an injured person to a receiver carried by a security guard.

[0036] (Injury and illness detection process) Next, with reference to FIG. 9, the injury and illness detection process according to this embodiment will be described. FIG. 9 is a flowchart showing the injury and illness detection process procedure according to this embodiment.

[0037] When the position extraction unit 31 receives an image captured by at least one imaging unit 10 that captures the monitoring location, it acquires the camera identification information included in the additional information attached to the image and the imaging time T of the image (step S201). Furthermore, based on the acquired camera identification information, the position extraction unit 31 refers to the monitoring camera - monitoring location correspondence table stored in the imaging unit information storage unit to determine the name of the monitoring location of the imaging unit 10.

[0038] Based on the image captured by the imaging unit 10, the position extraction unit 31 uses the skeleton extraction program stored in the software storage unit 22 to extract the position of the person 50 in the image and the body parts of the person 50 (step S202). At this time, the position extraction unit 31 assigns person identification information to the person 50 extracted from the image.

[0039] The position extraction unit 31 creates position data including the person identification information, the name of the monitoring location, and the position of the body parts at the imaging time T, and updates and stores it in the position data storage unit 23 (step S203).

[0040] Based on the position data stored in the position data storage unit 23, the fall detection unit 32 generates a rectangular human body frame representing the contour range of the human body in the image, and determines whether the first fall determination condition that the ratio of the vertical length to the horizontal length of the human body frame is less than a predetermined threshold is satisfied (step S204). Also, the fall detection unit 32 updates and stores the generated human body frame in the determination condition storage unit 24.

[0041] When the first fall determination condition is satisfied (step S204 - YES), the fall detection unit 32 determines that a person's fall has been detected and advances the process to step S207. On the other hand, when the first fall determination condition is not satisfied (step S204 - NO), the fall detection unit 32 determines that a person's fall has not been detected and advances the process to step S205.

[0042] Based on the position data stored in the position data storage unit 23, the fall detection unit 32 determines whether the second fall determination condition that the vertical position Y of the face part is lower than the vertical position Y of any one of the shoulder 54, hip 57, knee 58, and ankle 59 included in the joint part is satisfied (step S205). When the second fall determination condition is satisfied (step S205 - YES), the fall detection unit 32 determines that a person's fall has been detected and advances the process to step S207. On the other hand, when the second fall determination condition is not satisfied (step S205 - NO), the fall detection unit 32 determines that a person's fall has not been detected and advances the process to step S206.

[0043] Based on the position data stored in the position data storage unit 23, the fall detection unit 32 calculates the position of the center of gravity of the human body and determines whether the third fall determination condition that the horizontal position X of the center of gravity of the human body in the image is not between the horizontal position X of the left ankle 59a and the horizontal position X of the right ankle 59b is satisfied (step S206). At this time, the fall detection unit 32 calculates the position of the center of gravity of the human body based on the position data of the shoulder 54 and hip 57 included in the joint part. When the third fall determination condition is satisfied (step S206 - YES), the fall detection unit 32 determines that a person's fall has been detected and advances the process to step S207. On the other hand, when the third fall determination condition is not satisfied (step S206 - NO), the fall detection unit 32 determines that a person's fall has not been detected, returns the process to step S201, and repeats the subsequent process.

[0044] In step S207, the injured person determination unit 33 determines whether the fall detected by the fall detection unit 32 is a new fall that newly started from the imaging time T based on the fall start time T0 stored in the determination condition storage unit 24. More specifically, when the fall start time T0 is the initial value (for example, 000000000000), the injured person determination unit 33 determines that it is a new fall (step S207 - YES) and proceeds with the process to step S208. On the other hand, when the fall start time T0 is not the initial value, the injured person determination unit 33 determines that it is not a new fall (step S207 - NO) and proceeds with the process to step S209.

[0045] In step S208, the injured person determination unit 33 sets the fall start time T0 and the movable area, updates and stores them in the determination condition storage unit 24. Specifically, the injured person determination unit 33 sets the fall start time T0 to the imaging time T at which a new fall is detected. Further, based on the position data stored in the position data storage unit 23 and the human body frame stored in the determination condition storage unit 24, the injured person determination unit 33 sets a movable area for each human body part.

[0046] In step S209, the injured person determination unit 33 determines whether the fall of the person 50 in the image continues according to whether the position of the human body part at the imaging time T falls within the movable range. Specifically, when the position of the human body part falls within the movable area (step S209 - YES), the injured person determination unit 33 determines that the fall continues and proceeds with the process to step S210. On the other hand, when the position of the human body part does not fall within the movable area (step S209 - NO), the injured person determination unit 33 determines that the fall does not continue, sets the fall start time T0 to the initial value, updates and stores it in the determination condition storage unit 24. Then, the injured person determination unit 33 returns the process to step S201 and repeats the subsequent processes.

[0047] In step S210, the injured person determination unit 33 determines whether the fall of the person 50 in the image has continued for a predetermined period based on the fall start time T0 stored in the determination condition storage unit 24, and determines whether the person 50 is an injured person according to the determination result. Specifically, when the time difference between the imaging time T and the fall start time T0 is equal to or longer than a predetermined period (step S210 - YES), the injured person determination unit 33 determines that an injured person has been detected, sets the fall start time T0 as an initial value, updates and stores it in the determination condition storage unit 24. Further, the injured person determination unit 33 transmits an injured person detection signal indicating that an injured person has been detected to the alarm unit 40, and advances the process to step S211. On the other hand, when the time difference between the imaging time T and the fall start time T0 is less than a predetermined period (step S210 - NO), the injured person determination unit 33 determines that no injured person has been detected, returns the process to step S201, and repeats the subsequent processes.

[0048] When receiving the injured person detection signal from the injured person determination unit 33, the alarm unit 40 outputs an alarm notifying the occurrence of an injured person (step S211).

[0049] As described above, the injured person detection device 100 according to the embodiment of the present invention includes a position extraction unit 31, a fall detection unit 32, an injured person determination unit 33, and an alarm unit 40. The position extraction unit 31 extracts the position of the person 50 in the image and the body part of the person 50 based on an image captured by at least one imaging unit 10 that captures an area to be monitored, which is a monitoring location. The fall detection unit 32 detects the fall of the person 50 based on the position of the body part extracted by the position extraction unit 31. The injured person determination unit 33 determines that the person 50 is an injured person in a continued fall state when the fall of the person 50 detected by the fall detection unit 32 has continued for a predetermined period since the detection of the fall. The alarm unit 40 outputs an alarm when the injured person determination unit 33 determines an injured person. Thereby, the injured person detection device 100 according to the embodiment of the present invention can reduce false detection of injured persons because it detects a fall based on the position of the body part extracted from the image and determines the continued state of the detected fall. The injured person detection device 100 determines that a person 50 is an injured person when the person's fall continues for a predetermined period. For example, if the person 50 stands up and starts moving immediately after the fall (within a few seconds after the fall), the device does not determine that the person is an injured person. Also, when the person 50 crouches slowly and then lies on their side without moving, the injured person detection device 100 determines that the person is an injured person. In this way, the injured person detection device 100 can appropriately detect the person 50 as an injured person.

[0050] Also, in the embodiment of the present invention, the fall detection unit 32 generates a rectangle (human body frame) that passes through the minimum and maximum positions in the horizontal direction of the image and the minimum and maximum positions in the vertical direction of the image among the positions of the human body parts, where the distance between the minimum and maximum positions in the horizontal direction is the horizontal length and the distance between the minimum and maximum positions in the vertical direction is the vertical length. When the ratio of the vertical length to the horizontal length of the rectangle is less than a predetermined threshold (when the first fall determination condition is satisfied), the fall detection unit 32 detects that the person 50 has fallen. Thereby, the injured person detection device 100 according to the embodiment of the present invention can detect a person in a fallen posture that satisfies the first fall determination condition and reduce the omission of detecting a fallen person, so that false detection of an injured person can be reduced. For example, the injured person detection device 100 can appropriately detect, based on the first fall determination condition, that a person 50 with their head and feet facing horizontally in the horizontal direction of the image has fallen.

[0051] Also, in the embodiment of the present invention, the human body parts include the face part and the joint parts. When the vertical position Y of the face part in the image is lower than the vertical position Y of any one of the shoulder 54, hip 57, knee 58, and ankle 59 included in the joint parts (when the second fall determination condition is satisfied), the fall detection unit 32 detects that the person 50 has fallen. As a result, the injured person detection device 100 according to the embodiment of the present invention can detect a person in a falling posture that satisfies the second falling determination condition, and can reduce the omission of detection of the fallen person, so that the false detection of the injured person can be reduced. For example, the injured person detection device 100 can appropriately detect that a person 50 who turns his head downward in the vertical direction of the image (lower side of the image) and turns his feet upward in the vertical direction of the image (upper side of the image) is falling according to the second falling determination condition.

[0052] Further, in the embodiment of the present invention, the fall detection unit 32 calculates the position of the center of gravity of the human body based on the position of the joint part, and in the image, when the horizontal position X of the center of gravity is not between the horizontal position X of the left ankle 59a and the horizontal position X of the right ankle 59b (when the third fall determination condition is satisfied), it is detected that the person 50 is falling. As a result, the injured person detection device 100 according to the embodiment of the present invention can detect a person in a falling posture that satisfies the third falling determination condition, and can reduce the omission of detection of the fallen person, so that the false detection of the injured person can be reduced. For example, the injured person detection device 100 can appropriately detect that a person 50 who turns his head and feet horizontally in the horizontal direction of the image or a person 50 who is inclined obliquely with respect to the vertical direction of the image is falling according to the third falling determination condition.

[0053] Further, in the embodiment of the present invention, when the movement of the human body part in the image is within a predetermined area (within the movable area) within a predetermined period, the injured person determination unit 33 determines that the person 50 is an injured person. As a result, the injured person detection device 100 according to the embodiment of the present invention can detect a person who maintains a falling posture substantially the same as the start of the fall for a predetermined period as an injured person, so that the false detection of the injured person can be reduced. For example, within a predetermined period, if a part of a human body part (e.g., the knee, etc.) trembles slightly due to convulsions, most of the movement of the human body part is within the movable area. Therefore, the injured person determination unit 33 can determine that the person 50 is an injured person. Also, even if the position of the elbow 55 or the wrist 56 does not fall within the movable area because the person 50 is lying on the ground asking for help and waving one arm greatly, since the positions of the remaining human body parts are within the movable area, the injured person determination unit 33 can determine that the person 50 is an injured person.

[0054] In addition, the injured person detection system 1 according to the embodiment of the present invention includes at least one imaging unit 10 that images a monitoring location which is a location to be monitored, and the injured person detection device 100 described above. Thereby, the injured person detection system 1 according to the embodiment of the present invention has the same effect as the above-described injured person detection device 100, and can reduce the false detection of injured persons.

[0055] In addition, the injured person detection method according to the embodiment of the present invention includes a step in which the position extraction unit 31 extracts the position of the person 50 in the image and the position of the human body part of the person 50 based on an image captured by at least one imaging unit 10 that images a monitoring location which is a location to be monitored; a step in which the fall detection unit 32 detects the fall of the person 50 based on the position of the human body part of the person 50; a step in which the injured person determination unit 33 determines that the person 50 is an injured person in a state where the fall continues when the fall of the person 50 detected by the fall detection unit 32 continues for a predetermined period from the detection of the fall; and a step in which the alarm unit 40 outputs an alarm when the injured person determination unit 33 determines an injured person. Thereby, the injured person detection method according to the embodiment of the present invention has the same effect as the above-described injured person detection device 100, and can reduce the false detection of injured persons.

[0056] In the embodiment of the present invention, an example where the imaging unit 10 is a monitoring camera has been described, but the present invention is not limited thereto. For example, it may be a monocular camera. Thereby, the cost related to the introduction or operation of the monitoring camera can be suppressed.

[0057] Also, in the embodiment of the present invention, an example has been described in which the position extraction unit 31 uses the skeleton estimation program stored in the software storage unit 22 to extract the position of a person in the image and the body parts of the person 50. However, the present invention is not limited to this. For example, the position of the body parts may be extracted by connecting to a website that can execute the skeleton estimation program online via a network. Thereby, even when it is difficult to introduce the skeleton estimation program or construct the execution environment of the skeleton estimation program, the position of a person and body parts in the image can be extracted.

[0058] Also, in the embodiment of the present invention, an example has been described in which the fall detection unit 32 detects a fall when any one of the three fall determination conditions is satisfied based on the positions of 17 body parts. However, the present invention is not limited to this. The selection method of the body parts, the number of fall determination conditions, and the content of the fall determination conditions may be reset according to, for example, the hit rate of the injured person detection. Thereby, the accuracy of the injured person detection can be improved.

[0059] The above-described wounded person detection process in the embodiment may be implemented by a computer. In that case, a program for implementing this function may be recorded on a computer-readable recording medium, and the program recorded on this recording medium may be read into a computer system and executed to implement it. Here, the "computer system" is assumed to include hardware such as an OS and peripheral devices. Also, the "computer-readable recording medium" refers to a portable medium such as a flexible disk, magneto-optical disk, ROM, CD-ROM, etc., and a storage device such as a hard disk built into a computer system. Furthermore, the "computer-readable recording medium" refers to something that 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, and may also include something that holds a program for a certain period of time, like the volatile memory inside a computer system that serves as a server or client in that case. Also, the above program may be for implementing a part of the aforementioned functions, and furthermore, it may be something that can be implemented in combination with a program already recorded in a computer system for the aforementioned functions, or it may be implemented using a programmable logic device such as an FPGA (Field Programmable Gate Array).

[0060] As described above, the embodiments of the present invention have been described in detail with reference to the drawings. However, the specific configuration is not limited to this embodiment, and designs and the like within the scope that do not deviate from the gist of this invention are also included.

Explanation of Reference Numerals

[0061] 1... wounded person detection system, 10, 10a, 10b... imaging unit, 20... storage unit, 21... imaging unit information storage unit, 22... software storage unit, 23... position data storage unit, 24... determination condition storage unit, 30... control unit, 31... position extraction unit, 32... fall detection unit, 33... wounded person determination unit, 40... alarm unit, 50, 50a, 50b... person, 51... eye, 52... ear, 53... nose, 54... shoulder, 55... elbow, 56... wrist, 57... hip, 58... knee, 59... ankle

Claims

1. A position extraction unit that extracts the position of a person and the position of the human body part of the person in the image based on an image captured by at least one imaging unit that images a monitoring location that is a location to be monitored; A fall detection unit that detects the fall of the person based on the position of the human body part extracted by the position extraction unit; An injured person determination unit that determines that the person is an injured person in a state of continuous fall when the fall of the person detected by the fall detection unit continues for a predetermined period from the detection of the fall; An alarm unit that outputs an alarm when the injured person determination unit determines that the person is the injured person; Comprising; The fall detection unit, Among the positions of the human body part, a rectangle passing through the minimum position and the maximum position in the horizontal direction of the image and the minimum position and the maximum position in the vertical direction of the image, where the distance between the minimum position and the maximum position in the horizontal direction is the horizontal length, and the distance between the minimum position and the maximum position in the vertical direction is the vertical length. When the ratio of the vertical length to the horizontal length of the rectangle is less than a predetermined threshold, a first means for detecting that the person has fallen; The human body part includes a face part and a joint part. In the image, when the vertical position of the face part is lower than the vertical position of any one of the shoulder, hip, knee, and ankle included in the joint part, a second means for detecting that the person has fallen; Based on the position of the joint part, calculate the position of the center of gravity of the human body. In the image, when the horizontal position of the center of gravity is not between the horizontal position of the left ankle and the horizontal position of the right ankle, a third means for detecting that the person has fallen; Comprising; The fall detection unit proceeds to the second means when the person is not detected as having fallen in the first means, and proceeds to the third means when the person is not detected as having fallen in the second means, and is configured to detect whether the person has fallen. An injured person detection device.

2. The injured person determination unit, During the predetermined period, when the movement of the human body part in the image is within a predetermined area, determining that the person is the injured person; The injured person detection device according to claim 1, characterized in that.

3. At least one imaging unit that images a monitoring location that is a location to be monitored; The injured person detection device according to claim 1 or claim 2; An injured person detection system comprising.

4. The position extraction unit, Based on an image captured by at least one imaging unit that images a monitoring location which is the location to be monitored, extracting the position of a person in the image and the position of the body parts of the person. A fall detection unit detecting the fall of the person based on the position of the body parts of the person. An injured person determination unit determining that the person is an injured person in a continuous fall state when the fall of the person detected by the fall detection unit continues for a predetermined period from the detection of the fall. An alarm unit outputting an alarm when the injured person determination unit determines the injured person. including The step of detecting the fall of the person is the fall detection unit generating a rectangle that passes through the minimum and maximum positions in the horizontal direction and the minimum and maximum positions in the vertical direction of the image among the positions of the body parts, with the distance between the minimum and maximum positions in the horizontal direction being the horizontal length and the distance between the minimum and maximum positions in the vertical direction being the vertical length, and detecting that the person has fallen when the ratio of the vertical length to the horizontal length of the rectangle is less than a predetermined threshold, which is the first fall determination condition. The body parts include a face part and joint parts. In the image, when the vertical position of the face part is lower than the vertical position of any one of the shoulder, hip, knee, and ankle included in the joint parts, detecting that the person has fallen, which is the second fall determination condition. calculating the position of the center of gravity of the human body based on the position of the joint parts, and detecting that the person has fallen when the horizontal position of the center of gravity is not between the horizontal positions of the left ankle and the right ankle in the image, which is the third fall determination condition. determining whether the first fall determination condition, the second fall determination condition, and the third fall determination condition are satisfied. The step of detecting the fall of the person is performing a first determination to detect the fall of the person according to the first fall determination condition. In the first determination, when the fall of the person is not detected, performing a second determination to detect the fall of the person according to the second fall determination condition. An injured person detection method that performs a third determination to detect the fall of the person according to the third fall determination condition when the fall of the person is not detected in the second determination.

Images