JP2017019351A - Fallen person detection device, fallen person detection system, fallen person detection method, and fallen person detection program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a technology to accurately detect a fallen person who has fallen from a structure on a peripheral region of the structure.SOLUTION: A fallen person detection device 1 detects a fallen person who has fallen from a structure whose floor surface is formed into a first height, on a peripheral region of the structure whose floor surface height is a second height. A moving image captured by a far-infrared camera 2 installed at an angle for capturing an image of a step between the structure and the peripheral region from a peripheral region side is input in an image input unit 12. An image processing unit 13 processes a frame image related to the moving image input in the image input unit 12, detects an object captured in the frame image, and determines whether or not the detected object has moved from a structure side to the peripheral region side. The image processing unit 13 determines whether or not the object is a fallen person using the result of determination on a movement direction of the object.SELECTED DRAWING: Figure 2

Description

  The present invention relates to a technique for detecting a fallen person who has fallen from a structure, and more particularly to a technique for detecting a fallen person who has fallen from a station platform to a train track.

  Conventionally, a technique for detecting a person located on a train track by image processing has been proposed (see, for example, Patent Document 1). This Patent Document 1 is configured to process a frame image obtained by capturing a train track with a near-infrared camera and detect an object located on the train track to determine whether the person is a person based on the size and shape of the object. is there.

JP 2007-186136 A

  However, Patent Document 1 is a configuration that detects whether or not there is a person on the train track, and is not a configuration that detects a fallen person who has fallen from the station platform to the train track. That is, in the configuration of Patent Document 1, the person who is detected to be on the train track is a fallen person who has fallen from the station platform to the train track, or entered the train track to perform track maintenance work or the like. It is impossible to detect whether it is a worker or not.

  For this reason, in the configuration described in Patent Literature 1, if a person on the train track falls down from the station platform to the train track, the person in charge of the station staff is notified, and for some reason, the train track It is impossible to realize the operation control that the worker who has entered does not notify the station staff or the like. Since the priority is given to detection and rescue of a fallen person who has fallen from the station platform to the train track, in the configuration described in Patent Document 1, even when the worker enters the train track for any reason, the station staff, etc. Inevitably, there is a need to increase the work burden of station staff and the like, and may impede train operation.

  An object of the present invention is to provide a technique capable of accurately detecting a fallen person who has fallen from a structure to a peripheral region of the structure.

  The fallen person detecting device of the present invention is configured as follows to achieve the above object.

  This fallen person detection device falls from a structure in which the floor surface is formed at the first height to a peripheral region of the structure in which the height of the floor surface is a second height lower than the first height. Detect fallen people. For example, the structure mentioned here is a station platform, and the surrounding area is a track of a train on which a track is laid.

  A moving image captured by the imaging device is input to the image input unit. The imaging device is installed at an angle that captures a step between the structure and the peripheral region from the peripheral region side. The imaging device may be a visible light camera that captures a visible light image, or may be a far infrared camera that captures a far infrared image.

  The object detection unit processes a frame image applied to the moving image input to the image input unit, and detects an object captured in the frame image. The object detection unit detects the imaged object by generating a background difference image or an inter-frame difference image.

The movement direction determination unit determines whether the object detected by the object detection unit has moved from the structure side to the peripheral region side. For example, the movement direction determination unit determines whether or not the object detected by the object detection unit has moved from the structure side to the peripheral region side.
(1) It is good also as a structure determined by whether it is the object which passed the boundary located between the 1st height defined on the frame image and the 2nd height from the structure side to the peripheral region side. And
(2) Whether or not a plurality of boundaries located between the first height and the second height defined on the frame image have passed from the structure side to the peripheral region side, and have passed the plurality of boundaries It is good also as a structure determined by order.

  The movement path of the object can be performed by tracking the position of the object (position on the frame image) detected by the object detection unit over a plurality of frame images.

  The fallen person determination unit determines whether the object detected by the object detection unit is a fallen person using the determination result of the movement direction determination unit. For example, even if an object is located in the peripheral area, an object that has not moved from the structure side to the peripheral area side is determined not to have fallen from the structure to the peripheral area.

  The object referred to here includes not only humans but also small animals such as birds and cats, and articles.

  For the object detected by the object detection unit, the stay time determination unit for determining whether the time staying in the peripheral area exceeds a predetermined threshold time, or the object detected by the object detection unit An object size determination unit that determines whether the size of the object exceeds a predetermined threshold size may be additionally provided. In the case of including these configurations, the faller determination unit uses the determination result of the stay time determination unit and the determination result of the object size determination unit to determine whether the object detected by the object detection unit is a fallen person. It may be configured to determine whether or not

  If comprised in this way, it can suppress determining that the object which moved from the structure side to the peripheral region side is a fallen person who fell small animals, such as a bird and a cat, and articles from the structure to the peripheral region. .

  According to the present invention, it is possible to accurately detect a fallen person who has fallen from a structure to a peripheral region of the structure.

It is a figure which shows the structure of a fallen person detection system. It is a block diagram which shows the structure of the principal part of a fallen person detection apparatus. It is the schematic which shows the imaging area of a far-infrared camera. It is a figure explaining the frame image imaged with a far-infrared camera. It is a figure which shows the function structure of an image process part. It is a flowchart which shows operation | movement of a fallen person detection apparatus. It is a figure explaining the movement of the person who falls on the track of a train from a station platform. It is a figure explaining the example of a movement of the object determined not to fall on the track of a train from a station platform.

  Hereinafter, a fallen person detection system which is an embodiment of the present invention will be described.

  FIG. 1 is a diagram illustrating a configuration of a fallen person detection system. This falling person detection system includes a falling person detection device 1, a far-infrared camera 2, a vehicle detector 3, and a notification device 4. The fallen person detection system according to this example detects a fallen person who has fallen from a station platform to a track of a train on which a track is laid. In addition, when a fallen person is detected, the notification device 4 notifies the fact that the faller has been detected to a station staff, a train driver, or the like by notifying the fact. In this example, the station platform corresponds to the structure referred to in the present invention, and the track of the train corresponds to the peripheral area referred to in the present invention.

  FIG. 2 is a block diagram illustrating a configuration of a main part of the fallen person detection device. The fallen person detection device 1 includes a control unit 11, an image input unit 12, an image processing unit 13, and an input / output unit 14.

  The control part 11 controls operation | movement of each part of the faller detection apparatus 1 main body.

  A moving image captured by the far-infrared camera 2 is input to the image input unit 12. The frame rate of the far-infrared camera 2 may be about 5 to 30 frames / sec. The frame image of the imaging area captured by the far infrared camera 2 is a far infrared image of this imaging area.

  FIG. 3 is a schematic diagram showing an imaging area of the far-infrared camera. FIG. 4 is a diagram for explaining a frame image captured by the far-infrared camera. FIG. 3 is a schematic plan view of the station platform and the train track as viewed from above. The frame image captured by the far-infrared camera 2 is not a visible light image as shown in FIG. 4 but a far-infrared image.

  As shown in FIG. 3, the far-infrared camera 2 has a station platform installed on the opposite side of the train track. As shown in FIG. 4, the far-infrared camera 2 images the floor surface of the station platform, the floor surface of the train track (laying surface of the track), and the level difference between the floor surface of the station platform and the floor surface of the train track. It is installed at an angle that fits in the area. The height direction of the frame image shown in FIG. 4 is a direction in which the imaging position in real space is far from the far-infrared camera 2. FIGS. 3 and 4 show an example in which the entire station platform is imaged by one far-infrared camera 2. However, the station platform is divided into a plurality of areas, and the far area that captures each area is divided. An infrared camera 2 may be provided.

  Note that the far-infrared camera 2 according to this example may be configured with a visible light camera that captures a visible light image.

  The image processing unit 13 processes a frame image (a frame image input to the image input unit 12) related to the moving image captured by the far-infrared camera 2, and detects a fallen person who has fallen from the station platform to the train track. Falling person detection processing is performed. The image processing unit 13 executes the fallen person detection method according to the present invention. The image processing unit 13 includes an image processing processor (computer) that executes each step of the fallen person detection program according to the present invention. Details of the image processing unit 13 will be described later.

  The vehicle sensor 3 and the notification device 4 are connected to the input / output unit 14. The vehicle detector 3 senses train entry, stop, departure, passage, etc. at the station platform. That is, the vehicle detector 3 detects the presence or absence of a train around the station platform. The vehicle detector 3 inputs a detection signal indicating the presence or absence of a train around the station platform to the input / output unit 14. In addition, when there is an input from the fallen person detection apparatus 1 to the effect that the fallen person has been detected, the notification apparatus 4 notifies that fact. The notification device 4 performs, for example, lighting of a warning light installed in a station platform or station office, display control of a traffic light arranged on a track of a train, and the like. Notify the driver.

  FIG. 5 is a diagram illustrating a functional configuration of the image processing unit. The image processing unit 13 includes a processing target frame image selection function unit 21, a background difference image generation function unit 22, an object detection function unit 23, an object tracking function unit 24, a movement direction determination function unit 25, and a stay time determination. A function unit 26, an object size determination function unit 27, and a fallen person determination function unit 28 are provided.

  The processing target frame image selection function unit 21 performs a frame image (processing target) for performing image processing, which will be described later, from among the frame images input to the image input unit 12 (frame images related to moving images captured by the far-infrared camera 2). Select (Frame Image). The processing target frame image selection function unit 21 may be configured to select all the frame images related to the moving image input to the image input unit 12 as processing target frame images, or may be input to the image input unit 12. A configuration may be used in which a processing target frame image is selected at a predetermined frame interval or time interval from a frame image related to a moving image.

In addition, the relationship between the frame interval for selecting the processing target frame image and the time interval is as follows:
Frame interval for selecting the target frame image =
(Frame rate of far-infrared camera 2) × (time interval for selecting a frame image to be processed)
It is.

  The background difference image generation function unit 22 is a difference image (a so-called so-called “differential image”) between a selected processing target frame image and a background image (background image of the imaging area of the far-infrared camera 2) stored in an image memory (not shown). Background difference image). The background difference image generation function unit 22 has a function of updating the background image stored in the image memory using the processing target frame image. Since various techniques are already known for updating the background image, description thereof is omitted here.

  The object detection function unit 23 sets an object area based on the background difference image generated by the background difference image generation function unit 22. Specifically, a foreground pixel block (an area on the frame image) in the generated background difference image is estimated, and an area corresponding to the foreground pixel block is set as an object area. This object region is a region where an object imaged in the processing target frame image is detected.

  The object tracking function unit 24 tracks the movement of the object in the imaging area of the far-infrared camera 2 by detecting the position of the detected object (position on the frame image) on the processing target frame image.

  The movement direction determination function unit 25 determines whether or not the movement route based on the position of the object detected for each processing target frame image is the movement route of the fallen person who has fallen from the station platform to the train track.

  The stay time determination function unit 26 determines whether the time staying on the track side of the train is greater than or equal to a predetermined threshold time for an object located on the track side of the train.

  The object size determination function unit 27 determines whether or not the size of the object located on the track side of the train is greater than or equal to a predetermined threshold size.

  The faller determination function unit 28 determines the determination result of the movement direction determination function unit 25 regarding the movement direction of the object, the determination result of the stay time determination function unit 26 determines the stay time of the object, and the object size determination function unit 27 Using the determination result determined for the size of the object, it is determined whether this object is a fallen person who has fallen from the station platform to the train track.

  Hereinafter, the operation in which the fallen person detection apparatus 1 detects a fallen person is demonstrated in detail. FIG. 6 is a flowchart showing the operation of the fallen person detection device.

  The fallen person detection device 1 determines the presence or absence of a train around the station platform (s1). The fallen person detection device 1 makes a determination on s1 based on a detection signal from the vehicle sensor 3 input to the input / output unit 14. If there is a train around the station platform in s1, the fallen person detecting device 1 does not perform the processing after s2 (processing in the image processing unit 13) shown below.

  In this example, the far-infrared camera 2 is described as capturing a moving image of the imaging area regardless of whether there is a train around the station platform. However, if there is a train around the station platform, the imaging area The moving image may be temporarily stopped and the moving image in the imaging area may be captured only when there is no train around the station platform.

  If there is no train around the station platform, the faller detection device 1 has the image processing unit 13 input to the image input unit 12 and the frame image related to the moving image of the imaging area captured by the far-infrared camera 2. A processing target frame image is selected (s2). The processing target frame image selected in s2 is a frame image captured when there is no train around the station platform, and is a frame image that is later in time than the processing target frame image selected at the previous imaging timing. The processing related to s2 is performed by the processing target frame image selection function unit 21.

  The image processing unit 13 generates a background difference image for the processing target frame image selected in s2 (s3), and detects an object captured in the processing target frame image (s4). The processing relating to s3 and s4 is object detection processing by a known background difference method. The process related to s3 is performed by the background difference image generation function unit 22, and the process related to s4 is performed by the object detection function unit 23. In s4, not only people but also small animals such as birds and cats and articles are detected as objects.

  In s4, the radiant heat from the train or passengers reflected on the track or the like may be erroneously detected as an object.

  For each object detected in s4, the image processing unit 13 creates an object map that associates the ID for identifying the object, the position on the current processing target frame image, and the imaging timing of the current processing target frame image. Create (s5). In s5, an ID is assigned to an object that has not been detected in the previous processing target frame image, but a new ID is not assigned to an object that has already been detected in the previous processing target frame image. The object map created in s5 is data indicating a change in the position of the object on the frame image, that is, the movement path of the object.

  The image processing unit 13 determines whether there is an object located on the track side of the train among the objects detected in s4 (s6). In s6, it is determined whether or not the position of the object detected in s4 is on the track side of the train shown in FIG. The position of each object detected in s4 is detected when the object map is created in s5. The processing related to s5 and s6 is performed by the object tracking function unit 24.

  If the image processing unit 13 determines that there is no object located on the train track in s6, the image processing unit 13 determines that there is no person who has fallen from the station platform to the train track (s11), and returns to s1.

  If the image processing unit 13 determines that there is an object located on the train track in s6, the image processing unit 13 determines whether or not the moving path of the object is in the direction of falling from the station platform to the train track (s7). .

  FIG. 7 is a diagram illustrating the movement of a person who falls from a station platform to a train track. The first boundary line and the second boundary line shown in FIG. 7 are lines arranged in the height direction of the step between the floor surface of the station platform and the floor surface of the train track. The first boundary line is set at a position lower than the floor of the station platform and higher than the second boundary line. Further, the second boundary line is set at a position higher than the floor surface of the train track.

  Note that the first boundary line and the second boundary line are only shown for convenience, and the image processing unit 13 converts the first boundary line and the second boundary line into the processing target frame image. I'm not drawing. Further, the second boundary line may be defined in a region on the track side of the train on the frame image.

  As shown in FIG. 7, the person who falls from the station platform to the train track passes through the first boundary line and the second boundary line in this order from the station platform side, and reaches the train track side. The direction of the fall from the station platform to the track of the train is a direction that passes from the station platform side through the first boundary line and the second boundary line in this order and reaches the train track side. Therefore, it is determined that the bird (object) that has moved as shown in FIG. 8 is not in the direction of falling from the station platform to the train track in s7.

  The moving direction of the object located on the track of the train is obtained from the object map created for the object. The determination relating to s7 is performed for each object determined to be located on the train track in s6. The determination relating to s7 is not performed for objects that have not been determined to be located on the train track in s6. The process related to s7 is performed by the movement direction determination function unit 25.

  The image processing unit 13 determines that the object located on the train track detected in s6 is not a fallen person if there is no object determined in s7 to be in the direction of falling from the station platform to the train track. (S11). The process related to s11 is performed by the fallen person determination function unit 28.

  If there is an object determined to be in the direction of falling from the station platform to the train track in s7, the image processing unit 13 sets a predetermined threshold time for the time that the object stays on the train track. It is determined whether or not it exceeds (s8). The time spent on the train track is obtained from the object map created for that object. The determination relating to s8 is performed for each object that has been determined in s7 to be in the direction of falling from the station platform to the train track. The determination relating to s8 is not performed for an object that has not been determined in s7 to be in the direction of falling from the station platform to the train track. The process relating to s8 is performed by the stay time determination function unit 26.

  If there is no object determined that the time staying on the train track exceeds the threshold time, the image processing unit 13 determines in s7 that the direction is to fall from the station platform to the train track. It is determined that the object is not a fallen person (s11).

  If there is an object for which the time staying on the track of the train exceeds the threshold time, the image processing unit 13 determines whether the size of the object is equal to or larger than the threshold size (s9). In s9, the size of the object on the processing target frame image is corrected by the distance from the far-infrared camera 2 to the object, and it is determined whether or not the corrected size of the object is a predetermined size or more. Specifically, a correction coefficient that varies depending on the position in the height direction of the frame image shown in FIGS. 4 and 7 is determined. This correction coefficient is a coefficient that increases as the position of the frame image in the height direction increases. The image processing unit 13 sets a value obtained by multiplying the correction coefficient corresponding to the imaging position in the height direction on the processing target frame image and the size of the object on the processing target frame image to the corrected object size, Compare with threshold size. The size of the object on the processing target frame image may be the number of pixels related to the object, or may be the area of a rectangular region surrounding the object on the processing target frame image. The determination relating to s9 is performed for each object that has been determined in s8 that the time staying on the track of the train exceeds the threshold time. The determination relating to s9 is not performed for an object that has not been determined in s8 that the time staying on the track of the train exceeds the threshold time.

  If there is no object exceeding the threshold size, the image processing unit 13 determines that the object determined to have exceeded the threshold time in s8 is not a fallen person (s11).

  In addition, if there is an object determined to exceed the threshold size in s9, the image processing unit 13 determines that the object is a fallen person (s10).

  The fallen person detection device 1 outputs the determination result of s10 or s11 to the notification device 4 (s12), and returns to s1.

  When the determination result indicating that there is a fallen person is input from the faller detection apparatus 1, the notifying apparatus 4 notifies that fact.

  As described above, the fallen person detection device 1 according to this example is configured to determine whether or not an object located on a train track is a fallen person, depending on the moving direction of the object entering the train path. Therefore, it is possible to distinguish and detect a fallen person who falls from the station platform to the train track and a worker who enters the train track to perform track maintenance work. Workers who perform track maintenance work generally enter the track from the opposite side of the station platform, and do not get off the track from the station platform.

  In addition, because it is configured to determine whether or not you are a fallen person based on the stay time on the train track and the size of the object, small animals such as birds and cats that entered the train track, stations A small article or the like dropped from the home is not mistakenly determined to be a fallen person.

  Further, as described above, in this example, for all the objects captured in the processing target frame image, the falling direction is not determined in s7, but the objects whose falling direction is determined in s7 are narrowed down in s6. . Further, the objects for determining whether or not the threshold time is exceeded in s8 are narrowed down in s6 and s7. Further, the objects for determining whether or not the threshold size is exceeded at s9 are narrowed down by the processes of s6 to s8. Therefore, the processing load on the image processing unit 13 can be suppressed.

  Further, the order of the processes related to s7, s8, and s9 described above is not limited to the above example, and may be any order.

  In the above example, two boundary lines (first boundary line and second boundary line) are set, and whether or not the moving direction of the object is the falling direction is determined by the order of passing through the boundary line. Although it is determined, the number of the boundary lines may be one or may be three or more.

  Further, since the far-infrared camera 2 is less prone to halation caused by train headlights or poor imaging due to insufficient illumination, it can detect objects with less influence from the surrounding environment.

  Further, in the above example, the fall detection device 1 detects an object that has fallen from the station platform side to the train track side, and performs tracking processing on the object captured in the processing target frame image (creation of an object map according to s5). However, it may be performed by the following process.

  The fall detection device 1 uses, as a reference, a processing target frame image in which an object located on the first boundary line is detected, and a predetermined number of frames from the reference processing target frame image (or the imaging timing has elapsed for a predetermined time). When an object located on the second boundary line is detected in the processing target frame image processed until the time of (), it is determined that there is an object that has fallen from the station platform side to the train track side. In this configuration, since the tracking process for the object captured in the processing target frame image is not necessary, the processing load of the image processing unit 13 can be suppressed.

DESCRIPTION OF SYMBOLS 1 ... Falling person detection apparatus 2 ... Far-infrared camera 3 ... Vehicle detector 4 ... Notification apparatus 11 ... Control part 12 ... Image input part 13 ... Image processing part 14 ... Input / output part 21 ... Process target frame image selection function part 22 ... Background difference image generation function part 23 ... Object detection function part 24 ... Object tracking function part 25 ... Movement direction determination function part 26 ... Stay time determination function part 27 ... Object size determination function part 28 ... Falling person determination function part

Claims (10)

Detects a fallen person who has fallen from a structure in which the floor surface is formed at a first height to a peripheral region of the structure, in which the height of the floor surface is a second height lower than the first height. A falling person detecting device,
An image input unit for inputting a moving image captured by an imaging device installed at an angle for capturing a step between the structure and the peripheral region from the peripheral region side;
An object detection unit that processes a frame image applied to the moving image input to the image input unit and detects an object captured in the frame image;
A moving direction determination unit that determines whether or not the object detected by the object detection unit has moved from the structure side to the peripheral region side;
A fallen person detection device comprising: a fallen person determination part that determines whether or not the object detected by the object detection part is a fallen person using the determination result of the movement direction determination part.   The moving direction determination unit determines whether the object detected by the object detection unit has moved from the structure side to the peripheral region side on the frame image, the first height, and the first The fallen person detecting device according to claim 1, wherein the fallen person detecting device determines whether or not the boundary is located between a height of 2 and an object that has passed from the structure side to the peripheral region side.   The moving direction determination unit determines whether the object detected by the object detection unit has moved from the structure side to the peripheral region side on the frame image, the first height, and the first The fallen person detection according to claim 1, wherein a plurality of boundaries located between two heights is determined from whether or not a plurality of boundaries have passed from the structure side to the peripheral region side and a plurality of the boundaries are passed through. apparatus. About the object detected by the object detection unit, comprising a stay time determination unit for determining whether the time staying in the peripheral area exceeds a predetermined threshold time,
The fall of the faller according to any one of claims 1 to 3, wherein the faller determination unit determines whether the object detected by the object detection unit is a fallen person, using the determination result of the stay time determination unit. Person detection device. For an object detected by the object detection unit, an object size determination unit for determining whether the size of the object exceeds a predetermined threshold size,
The fallen person determination unit according to any one of claims 1 to 4, wherein the fallen person determination unit determines whether or not the object detected by the object detection unit is a fallen person using a determination result of the object size determination unit. Person detection device.   The fallen person detection device according to claim 1, wherein the imaging device is a far-infrared camera that captures a far-infrared image. The structure is a station platform,
The fallen person detection device according to any one of claims 1 to 6, wherein the peripheral region is a track of a train on which a track is laid. Detects a fallen person who has fallen from a structure in which the floor surface is formed at a first height to a peripheral region of the structure, in which the height of the floor surface is a second height lower than the first height. A falling person detection device to
A fallen person detection system comprising an imaging device installed at an angle for imaging a step between the structure and the peripheral area from the peripheral area side,
The fallen person detecting device is
An image input unit for inputting a moving image captured by the imaging device;
An object detection unit that processes a frame image applied to the moving image input to the image input unit and detects an object captured in the frame image;
A moving direction determination unit that determines whether or not the object detected by the object detection unit has moved from the structure side to the peripheral region side;
A faller determination unit that determines whether the object detected by the object detection unit is a fallen person using the determination result of the moving direction determination unit;
Falling person detection system. Detects a fallen person who has fallen from a structure in which the floor surface is formed at a first height to a peripheral region of the structure, in which the height of the floor surface is a second height lower than the first height. A fallen person detection method that
An image input device, which is input to the image input unit, processes a frame image related to a moving image captured at an angle that captures a step between the structure and the peripheral region from the peripheral region side, and is captured by the frame image. An object detection step for detecting a moving object;
A moving direction determining step for determining whether or not the object detected by the object detecting step has moved from the structure side to the peripheral region side;
A fallen person detection method comprising: a fallen person determination step that determines whether or not the object detected by the object detection step is a fallen person using the determination result of the moving direction determination step. Detects a fallen person who has fallen from a structure in which the floor surface is formed at a first height to a peripheral region of the structure, in which the height of the floor surface is a second height lower than the first height. A fallen person detection program that causes a computer to execute steps for
An image input device, which is input to the image input unit, processes a frame image related to a moving image captured at an angle that captures a step between the structure and the peripheral region from the peripheral region side, and is captured by the frame image. An object detection step for detecting a moving object;
A moving direction determining step for determining whether or not the object detected by the object detecting step has moved from the structure side to the peripheral region side;
A fallen person detection program for causing a computer to execute a fallen person determination step of determining whether or not the object detected by the object detection step is a fallen person using the determination result of the moving direction determination step.