Fire detection system

Abstract

To detect a fire by identifying an actual fire occurrence position in a monitoring area from a photographed image even when the flame and smoke of the fire do not appear in the photographed image.SOLUTION: A fire detection system 1 includes: an event detection section 24 for detecting a fire on the basis of a photographed image obtained by photographing a monitoring area by a photographing device 10, and detecting an event which occurs on a predetermined face existing in the monitoring area due to the fire which has occurred in the monitoring area on the basis of the photographed image; and a fire detection section 26 for identifying a fire occurrence position on the basis of the detection result of the event by the event detection section 24, so as to detect the fire.SELECTED DRAWING: Figure 1

Description

Technical Field

[0001] The present invention relates to a fire detection system that detects a fire based on a captured image obtained by imaging a monitoring area with an imaging device.

Background Art

[0002] Conventionally, various fire detection devices and fire detection systems have been proposed that perform predetermined image processing on a captured image of a monitoring area captured by an imaging device and detect a fire by detecting a flame or smoke generated by the fire from the captured image (Patent Documents 1 and 2).

[0003] In addition, in a fire detection system, when detecting a fire, it has been studied to specify the occurrence position of the fire from the positions of the flame and smoke in the captured image.

Prior Art Documents

Patent Documents

[0004]

Patent Document 1

Patent Document 2

Summary of the Invention

Problems to be Solved by the Invention

[0005] However, even if the occurrence position of the fire (the position of the flame or smoke) in the captured image can be specified, since the captured image is a two-dimensional image obtained by imaging a monitoring area in a three-dimensional space by a perspective projection method, it is not possible to specify to what extent the coordinates of the occurrence position of the fire in the captured image include each of the actual height direction component in the monitoring area and the depth direction component perpendicular to the height direction, and it has been difficult to accurately specify the actual occurrence position and scale of the fire in the monitoring area.

[0006] In addition, in the detection of a fire using a conventional captured image, it is necessary for a flame or smoke to appear in the captured image. When the ignition source is blocked by an obstacle or the like and no flame or smoke appears in the captured image, there has been a problem that the flame or smoke cannot be detected from the captured image and the fire cannot be correctly detected. Further, since no flame or smoke appears in the captured image, the location where the fire has occurred cannot naturally be specified.

[0007] An object of the present invention is to provide a fire detection system that enables specifying the actual fire occurrence location in a monitoring area from a captured image, and further enables fire detection even when no flame or smoke due to the fire appears in the captured image.

Means for Solving the Problems

[0008] (Fire Detection System) A fire detection system that detects a fire based on a captured image obtained by imaging a monitoring area with an imaging device, an event detection unit that detects an event occurring on a predetermined surface existing in the monitoring area due to a fire occurring in the monitoring area based on the captured image; a fire detection unit that specifies the fire occurrence location based on the detection result of the event by the event detection unit and detects the fire; characterized by comprising the above.

[0009] (Specification of the Position in the Depth Direction in the Monitoring Area) The fire detection unit specifies the position in the depth direction of the fire occurrence location in a predetermined one direction in the monitoring area as the fire occurrence location.

[0010] (Event Detection Area in the Image) The event detection unit sets an in-image event detection area for a predetermined area of the surface for detecting an event in the captured image, detects an event occurring in the in-image event detection area, and the fire detection unit specifies the position in the depth direction of the fire occurrence location based on the detection result of the event in the in-image event detection area by the event detection unit.

[0011] (Projected Image) The event detection unit generates a projection image by performing predetermined image processing on the in-image event detection region in the captured image such that the axial direction of any coordinate axis in the captured image substantially coincides with the depth direction, and detects an event occurring in the in-image event detection region based on the projection image. The fire detection unit specifies the position in the depth direction of the occurrence position of the fire by specifying the position in the depth direction of the event detected by the event detection unit in the projection image.

[0012] (Projection division region) The event detection unit further divides the projection image into a plurality of projection division regions by dividing it with a predetermined width in the depth direction, and detects an event occurring in the in-image event detection region by detecting an event and a feature amount of the event for each projection division region. The fire detection unit specifies the position in the depth direction of the occurrence position of the fire as the position in the depth direction of the region in which the feature amount most indicative of an event is detected highly by the event detection unit among the plurality of projection division regions, or as the position in the depth direction at the center of the region in which the feature amount indicative of an event is detected.

[0013] (Determination of event due to image shake) The event detection unit further determines the correlation of image shake in mutually adjacent projection division regions in which an event is detected, determines that an event has occurred for the projection division region determined to have a high correlation of image shake, and determines that a non-fire factor other than an event has occurred for the projection division region determined to have a low correlation of image shake.

[0014] (Detection of event for a plurality of surfaces) The event detection unit performs processing to detect an event for each of a plurality of surfaces existing in the monitoring region. The fire detection unit specifies the occurrence position of the fire in a plurality of directions including the position in the depth direction of the occurrence position of the fire as the occurrence position of the fire based on the detection results of the events on the plurality of surfaces by the event detection unit.

[0015] (Detection of reflection due to fire) The event detection unit detects, as an event, a reflection occurring on a predetermined surface existing in the monitoring region due to the flame of the fire.

Advantages of the Invention

[0016] (Advantages of the Fire Detection System) A fire detection system that detects a fire based on a captured image of a monitoring area by an imaging device, comprising an event detection unit that detects an event occurring on a predetermined surface existing in the monitoring area due to a fire occurring in the monitoring area based on the captured image, and a fire detection unit that specifies the occurrence position of the fire based on the detection result of the event by the event detection unit and detects the fire. Therefore, by using the event (for example, the reflection occurring on the predetermined surface due to the flame of the fire) and its position information, etc. that occur on the predetermined surface of the monitoring area due to the fire, it is possible to more accurately specify the occurrence position of the fire.

[0017] In particular, in a tunnel extending in a predetermined one direction (depth direction) or a corridor of a building, etc., it is important to be able to specify the position in the extending direction as the occurrence position of the fire. Therefore, the fire detection system can specify the position in the depth direction of the occurrence position of the fire, which is a predetermined one direction in the monitoring area, as the occurrence position of the fire.

[0018] Also, the detection of the fire is not by detecting the flame or smoke of the fire, but by detecting an event occurring on a predetermined surface of the monitoring area due to the fire from the captured image. Therefore, it is possible to detect the fire even if the flame or smoke cannot be captured by the imaging device due to an obstacle or the like.

[0019] (Advantages of the Event Detection Area in the Image) In addition, the event detection unit sets an in-image event detection area for a predetermined area of the surface for detecting an event in the captured image, and detects an event occurring in the in-image event detection area. The fire detection unit specifies the position in the depth direction of the occurrence position of the fire based on the detection result of the event in the in-image event detection area by the event detection unit. Therefore, it is possible to set only the area where the event can be detected as the detection area without including the area where the detection of the event is difficult in the detection area, and it is possible to improve the detection accuracy of the event. Also, since a part of the surface for detecting the event can be set as the detection area, it is possible to reduce the load applied to the event detection process.

[0020] (Effect of the projected image) Further, the event detection unit generates a projected image by performing predetermined image processing on the in-image event detection region in the captured image so that the axial direction of any coordinate axis in the captured image substantially coincides with the depth direction, detects an event occurring in the in-image event detection region based on the projected image, and the fire detection unit specifies the depth direction position of the event detected by the event detection unit in the projected image to thereby specify the depth direction position of the fire occurrence position. Therefore, event detection processing is performed using a projected image in which the distance in the depth direction in the monitoring region is a constant distance in the image, making it possible to more accurately and easily specify the depth direction position of the fire occurrence position.

[0021] (Effect of the projected division region) Further, the event detection unit divides the projected image into a plurality of projected division regions by further dividing it with a predetermined width in the depth direction, detects an event and the feature amount of the event for each projected division region to detect an event occurring in the in-image event detection region, and the fire detection unit specifies the depth direction position of the region in which the feature amount most characteristic of an event is detected highest by the event detection unit among the plurality of projected division regions, or the depth direction position at the center of the region in which the feature amount characteristic of an event is detected as the depth direction position of the fire occurrence position. Therefore, event detection processing is performed using a projected division region obtained by further subdividing the projected image, making it possible to more accurately and easily specify the depth direction position of the fire occurrence position.

[0022] (Effect of event detection due to image fluctuation) In addition, the event detection unit further determines the correlation of image fluctuations in mutually adjacent projection division regions where an event has been detected. For projection division regions determined to have a high correlation of image fluctuations, it is determined that an event has occurred, and for projection division regions determined to have a low correlation of image fluctuations, it is determined as a non-fire factor other than an event. Therefore, for example, an event having the same fluctuation frequency as a flame, such as a reflection occurring on a predetermined surface due to the flame of a fire, can be detected more distinctly from non-fire factors such as a surface with a color close to that of the flame or lighting, making it possible to suppress and prevent false detection of a fire.

[0023] (Effect by detecting events for a plurality of surfaces) In addition, the event detection unit performs a process of detecting an event for each of a plurality of surfaces existing in the monitoring region, and the fire detection unit specifies the occurrence position of a fire in a plurality of directions including the position in the depth direction of the occurrence position of the fire, based on the detection results of the events on the plurality of surfaces by the event detection unit. Therefore, in addition to the depth direction of the occurrence position of the fire, for example, the position in the left-right direction perpendicular to the depth direction can be specified, making it possible to more accurately specify the position of the occurrence position of the fire.

Brief Description of the Drawings

[0024]

Figure 1

Figure 2

Figure 3

Figure 4

Figure 5

Figure 6

Figure 7

Figure 8

Embodiments for Carrying Out the Invention

[0025] Hereinafter, embodiments of the fire detection system according to the present invention will be described in detail with reference to the drawings. Note that the present invention is not limited by the following embodiments.

[0026] [Basic Concept of the Embodiment] First, the basic concept of the embodiment will be described. The embodiment relates to a fire detection system that detects a fire based on a captured image obtained by imaging a monitoring area with an imaging device. Note that the "fire detection system" includes the concepts of "fire detection equipment" and "fire detection device".

[0027] Here, the "monitoring area" is an area to be monitored, which is an outdoor or indoor space with a certain extent. Since the fire detection system detects events occurring on a predetermined surface existing in the monitoring area, there are structures having a predetermined surface such as walls, floors, and ceilings in the "monitoring area". For example, it is a concept including areas such as rooms, corridors, and stairways in a building, and particularly targets spaces extending in one direction such as tunnels and corridors as the monitoring area.

[0028] Also, the "imaging device" is installed in the monitoring area and images the monitoring area, and includes concepts such as a television camera or a monitoring camera that images a moving image such as an ITV camera.

[0029] And the fire detection system of the embodiment is characterized by including an event detection unit and a fire detection unit in order to detect a fire based on a captured image obtained by imaging a monitoring area with an imaging device.

[0030] The "event detection unit" detects, based on captured images, an event that occurs on a predetermined surface existing in the monitoring area due to a fire that has occurred in the monitoring area. The "fire detection unit" detects a fire by identifying the location where the fire has occurred based on the detection result of the event by the event detection unit.

[0031] Here, the "predetermined surface" on which the event detection unit detects an event is not limited to one surface. If there are a plurality of surfaces in the monitoring area, events may be detected for the plurality of surfaces. For example, in a space surrounded by each direction except the traveling direction, such as a tunnel or a corridor, when the traveling direction is taken as the depth direction, an arbitrary surface is selected from the ceiling surface and the floor surface existing in the vertical direction and the wall surfaces existing in the left - right direction.

[0032] Also, the "event" detected by the event detection unit is arbitrary as long as it occurs on a predetermined surface existing in the monitoring area due to a fire that has occurred in the monitoring area. For example, it includes "reflection" etc. that occurs on a predetermined surface existing in the monitoring area due to the flames of a fire.

[0033] Also, the "location where the fire has occurred" identified by the fire detection unit is not limited to precisely identifying the location where the fire has occurred. For example, it may be to identify the position in the depth direction of the location where the fire has occurred in a predetermined one - direction in the monitoring area. In particular, when the monitoring area is a space extending in one direction such as a tunnel or a corridor, when identifying the location where the fire has occurred, it is important to identify the position in the extending direction. Even if the positions in the other two directions perpendicular to the extending direction are not identified, since the location where the fire has occurred is sufficiently identified, a fire detection unit that identifies the position in the depth direction of the location where the fire has occurred in a predetermined one - direction also has a sufficient function to identify the location where the fire has occurred. Note that the "depth direction" is the axial direction on one axis in a three - dimensional space where three axes are mutually perpendicular. Which direction to set the "depth direction" in the monitoring area is arbitrary, and the other two axial directions other than the "depth direction" shall be referred to as the "height direction" and the "left - right direction".

[0034] In addition, the "event detection result by the event detection unit" used by the fire detection unit for specifying the occurrence position of a fire and detecting a fire includes a determination result as to whether an event has occurred, predetermined information calculated to reach the determination result, and any information necessary for the fire detection unit to specify the occurrence position of a fire and detect a fire.

[0035] In addition, the event detection unit sets an in-image event detection area for a predetermined area of the surface for detecting an event in the captured image, detects an event occurring in the in-image event detection area, and the fire detection unit specifies the position in the depth direction of the occurrence position of a fire based on the detection result of the event in the in-image event detection area by the event detection unit.

[0036] Here, the number and size of the "in-image event detection area" are arbitrary. The "predetermined area where the in-image event detection area is set" may be the entire area of the surface for detecting an event or a partial area of the surface, and may also be set for a plurality of surfaces. For example, in a monitoring area where vehicles pass through like a tunnel, areas obstructed by passing vehicles or areas where lighting is installed may be set to be excluded from the in-image event detection area.

[0037] In addition, the event detection unit generates a projection image by performing predetermined image processing that substantially aligns the axial direction of any coordinate axis in the captured image with the depth direction for the in-image event detection area in the captured image, detects an event occurring in the in-image event detection area based on the projection image, and the fire detection unit specifies the position in the depth direction of the occurrence position of a fire by specifying the position in the depth direction of the event detected by the event detection unit in the projection image.

[0038] Here, the "predetermined image processing that substantially aligns the axial direction of any coordinate axis in the captured image with the depth direction" is a process of correcting the image so that the axial direction of either of the two coordinate axes (X-axis, Y-axis) in the captured image, which is a two-dimensional plane, substantially coincides with the depth direction in the monitoring area, and a predetermined distance in the actual depth direction becomes a certain distance in the axial direction of either of the coordinate axes adjusted to the depth direction in the image, and includes, for example, known projective transformation, etc.

[0039] Further, the event detection unit divides the projection image into a plurality of projection divided regions by further dividing in a predetermined width in the depth direction, and detects an event and a feature amount of the event for each projection divided region to detect an event occurring in the in-image event detection region. The fire detection unit specifies, as the position in the depth direction of the fire occurrence position, the position in the depth direction of the region in which the feature amount most indicative of an event is detected by the event detection unit among the plurality of projection divided regions, or the position in the depth direction that is the center of the region in which the feature amount indicative of an event is detected.

[0040] Here, the number and size of the “projection divided regions” to be divided are arbitrary. The smaller the size (width in the depth direction) and the larger the number of projection divided regions, the more accurately the position in the depth direction of the fire occurrence position can be specified.

[0041] Also, the “feature amount of the event” is one of the predetermined information included in the detection result of the event, which is calculated to determine whether or not an event is occurring. The closer the feature amount is to the value of the feature amount of the event to be detected, the higher the possibility that an event is occurring. For example, when the event to be detected is “reflection due to fire”, since the “reflection” generated by the flame has characteristics similar to those of the flame, the event detection unit calculates feature amounts related to “flame-like color” and “flame-like fluctuation” and detects whether or not reflection is occurring.

[0042] Further, the event detection unit determines the correlation of image fluctuations in adjacent projection divided regions where an event has been detected. For the projection divided region determined to have a high correlation of image fluctuations, it is determined that an event is occurring, and for the projection divided region determined to have a low correlation of image fluctuations, it is determined that a non-fire factor other than the event is occurring.

[0043] For example, when the event to be detected, such as a backdraft due to a fire, has characteristics similar to a flame, the event will have the characteristic of "flame-like fluctuation", and the characteristic of "flame-like fluctuation" will appear as an image fluctuation in the captured image or the projection image obtained by processing the captured image. Therefore, in the projection segmentation region where the event is detected, if a real backdraft is occurring, an image fluctuation corresponding to "flame-like fluctuation" will occur, while if a non-fire factor such as lighting is detected as an event, no image fluctuation corresponding to "flame-like fluctuation" will occur. By determining this as the correlation of image fluctuations in mutually adjacent projection segmentation regions, it is distinguished, and the detection of non-fire factors as events is suppressed and prevented.

[0044] Further, the event detection unit may perform a process of detecting an event for each of a plurality of surfaces existing in the monitoring region, and the fire detection unit may specify the occurrence position of the fire in a plurality of directions including the position in the depth direction of the occurrence position of the fire as the occurrence position of the fire based on the detection results of the events on the plurality of surfaces by the event detection unit.

[0045] Hereinafter, specific embodiments will be described. In the specific embodiments shown below, the case where the target monitoring region is a "tunnel extending in the depth direction" and the event detected by the event detection unit is a "backdraft due to a fire" will be described.

[0046] [Specific Content of the Embodiment] Embodiments of the fire detection system will be described separately as follows. a. Fire detection system b. Event detection unit b1. Image event detection region setting unit b2. Projection image generation unit b3. Projection image segmentation unit b4. Event determination unit c. Fire detection unit c1. Fire occurrence position specifying unit c2. Fire determination unit d. Fire detection operation e. Detection of events for a plurality of surfaces e1. Event detection unit e2. Fire detection unit f. Modification example of the present invention

[0047] [a. Fire detection system] First, the configuration of the fire detection system will be described. In this description, refer to FIG. 1 showing an example of the configuration of the fire detection system.

[0048] The fire detection system 1 is a system that detects a fire occurring in a tunnel by specifying its occurrence position based on an imaging image obtained by imaging the tunnel serving as a monitoring area. As shown in FIG. 1, it includes an imaging device 10 and a fire detection device 20. Note that the imaging device 10 and the fire detection device 20 may be integrated into one device.

[0049] The imaging device 10 is, for example, a monitoring camera installed at a predetermined position in the tunnel and is communicably connected to the fire detection device 20. The communication between the imaging device 10 and the fire detection device 20 may be wired or wireless, and any communication method such as communication via a USB cable or communication via a wireless LAN is adopted.

[0050] In addition, the imaging device 10 captures a moving image consisting of a continuous sequence of color frame images by sequentially imaging the monitoring area at a frame rate of, for example, 30 frames per second or 60 frames per second, and transmits the image data of the RGB frame image (imaging image) of the monitoring area at 30 frames per second or 60 frames per second from the captured moving image to the fire detection device 20. Here, the size of the image data transmitted from the imaging device 10 to the fire detection device 20 is arbitrary, but for example, it is 1920×1080 pixels. Also, since the color pixels of the color frame image are composed of R pixels, G pixels, and B pixels, each of which is 8-bit data, the pixel value of one color pixel is 24-bit data.

[0051] The fire detection device 20 is composed of a computer circuit including a CPU, a memory, various input / output ports, etc. as hardware, and includes an image acquisition unit 22, an event detection unit 24, a fire detection unit 26, a display operation unit 28, and a storage unit 30.

[0052] The image acquisition unit 22 receives the image data of the captured image transmitted from the imaging device 10 (hereinafter referred to as the captured image), and transmits the received captured image to, for example, the event detection unit 24 and the display operation unit 28. In addition, for example, the captured image may be transmitted to the storage unit 30 so that the storage unit 30 stores the captured image. Note that, since the accuracy of the captured image is improved in the processes described later, such as projective transformation, Hough transformation, and vanishing point detection, it is preferable to perform correction of lens distortion.

[0053] The event detection unit 24 detects a reflection (event) caused by a fire on a predetermined surface existing in the tunnel due to the flame of a fire occurring in the tunnel based on the captured image, and includes an in-image event detection area setting unit 2410, a projective image generation unit 2420, a projective image division unit 2430, and an event determination unit 2440 to realize its function.

[0054] The fire detection unit 26 detects a fire and specifies the occurrence position of the fire based on the detection result of the reflection by the event detection unit 24, and includes a fire occurrence position specifying unit 2610 and a fire determination unit 2620 to realize its function. Note that the details of the event detection unit 24 and the fire detection unit 26 will be described separately later.

[0055] The display operation unit 28 includes a display unit that displays the captured image received from the image acquisition unit 22 and also displays predetermined information necessary for fire detection and its settings, and an operation unit that performs settings and other necessary operations in the preparation stage of the fire detection system. The storage unit 30 stores determination conditions and the like necessary for the detection of events by the event detection unit 24, the detection of fires by the fire detection unit 26, and the specification of the occurrence positions of fires, as well as the detection results of events and fires, and also stores predetermined information necessary for fire detection and its settings.

[0056] [b. Event Detection Unit] Next, the event detection unit will be described. As described above, the event detection unit 24 detects a specular reflection (event) occurring on a predetermined surface existing in the tunnel due to the flames of a fire occurring in the tunnel based on the captured image.

[0057] Here, as shown in FIG. 2 which shows an example of the captured image 40 near the tunnel entrance 55 captured by the imaging device 10, the tunnel is surrounded by the ceiling surface 51, the floor surface 52, the left wall surface 53, and the right wall surface 54. For the predetermined surface on which the event detection unit 24 detects the specular reflection, one or a plurality of surfaces will be selected from these four surfaces. In the description of the event detection unit 24, the case where the specular reflection is detected for the right wall surface 54 will be described.

[0058] Also, regarding the directions in the three-dimensional space in the tunnel, as shown in FIG. 2, the traveling direction of the vehicle is defined as the depth direction, and the three mutually orthogonal directions are the left-right direction, the depth direction, and the height direction (up-down direction). Also, up, down, right, left, forward (front), and backward (rear) are defined as shown in FIG. 2 in the same way.

[0059] (b1. Image event detection area setting unit) First, the image event detection area setting unit of the event detection unit will be described. For this description, refer to FIG. 3 which shows the captured image with the image event detection area set in the captured image of FIG. 2. Note that (A) of FIG. 3 shows the captured image with the image event detection area set, and (B) shows the setting stage of the detection limit coordinates on the back side in the image event detection area.

[0060] In order for the event detection unit 24 to be able to detect the specular reflection occurring on the right wall surface 54, in the preparation stage before operating the fire detection system 1, the image event detection area setting unit 2410 sets in advance the image event detection area 60 for detecting the specular reflection in the area of the right wall surface 54 shown in the captured image 40, and stores it in the storage unit 30. The image event detection area 60 does not need to cover all the areas of the right wall surface 54 shown in the captured image 40, and an arbitrary area can be set.

[0061] The method for setting the in-image event detection area 60 by the in-image event detection area setting unit 2410 is arbitrary, and an example of the setting method is shown below.

[0062] In the setting method shown as an example, the in-image event detection area 60 set by the in-image event detection area setting unit 2410 is the area indicated by the diagonal lines surrounded by four points of detection limit coordinates 6010 to 6040 as shown in Fig. 3(A).

[0063] The in-image event detection area 60 is set, for example, as an area that covers the area from the front side to the back side of the upper part of the right wall surface 54. It is easily affected by passing vehicles and the like, and the lower part of the right wall surface 54 where the detection of the reflection is hindered by the event detection unit 24 is excluded from the reflection detection area. Also, when specifying the occurrence position of a fire occurring in the tunnel, what is important is the traveling direction of the vehicle and the position in the depth direction which is also the direction in which the tunnel extends. Therefore, the in-image event detection area 60 is set so as to cover from the front side to the back side so that the position in the depth direction can be specified.

[0064] When setting the in-image event detection area 60 shown in Fig. 3(A), first, the front-side detection limit coordinates 6010 and 6020 are set. As a method for setting the front-side detection limit coordinates 6010 and 6020, for example, the imaging image 40 is displayed on the display operation unit 28, and the coordinates to be the detection limit coordinates 6010 and 6020 are set by manual operation.

[0065] In addition, other methods include installing a marker detectable in the imaging image 40 in the monitoring area at the site, detecting the marker on the imaging image 40 by the in-image event detection area setting unit 2410 or artificially, and setting the coordinates of the detection position as the detection limit coordinates 6010 and 6020, or setting a predetermined area near the edge in the imaging image 40 or an area where the detection of the reflection may be hindered due to passing or the like as a non-detection area, and then setting the detection limit coordinates 6010 and 6020 so that the detection limit coordinates cannot be set in the non-detection area. The front-side detection limit coordinates 6010 and 6020 may be set by an appropriate method.

[0066] After setting the front - side detection - limited coordinates 6010 and 6020, the back - side detection - limited coordinates 6030 and 6040 are set. Regarding the back - side detection - limited coordinates 6030 and 6040, they may be set manually in the same way as the front - side detection - limited coordinates 6010 and 6020. However, for example, the in - image event detection area setting unit 2410 obtains the vanishing point in the captured image 40, and there is also a method of setting using the vanishing point and the front - side detection - limited coordinates 6010 and 6020.

[0067] Here, the captured image 40 is a two - dimensional image captured by perspective projection within a tunnel that is a three - dimensional space. The infinity of the straight lines existing in the captured image 40 converges to a predetermined vanishing point in the two - dimensional plane. And the vanishing point necessary for setting the back - side detection - limited coordinates 6030 and 6040 is the vanishing point to which the straight lines substantially parallel to the depth direction converge.

[0068] In extracting the straight lines substantially parallel to the depth direction necessary for obtaining this vanishing point, as a method of detecting the straight lines existing in the captured image 40, for example, there is straight - line detection using a known Hough transform. The straight - line detection using the Hough transform performs grayscale conversion to convert the captured image 40, which is a color image, into a grayscale image, performs edge processing to detect edges on the captured image 40 converted into a grayscale image, and is performed on the edge image having the detected edge information.

[0069] Also, among the straight lines detected from the edge image by straight - line detection using the Hough transform, there are various straight lines other than the straight lines substantially parallel to the depth direction. Therefore, the straight lines substantially parallel to the depth direction necessary for obtaining the vanishing point according to a predetermined condition are extracted.

[0070] For example, the straight lines detected by straight line detection using the Hough transform include straight lines such as straight lines 61 to 63 shown in FIG. 3(B). Straight lines 61 and 62 are straight lines that are substantially parallel to the depth direction and are the straight lines necessary for obtaining the vanishing point. However, for a straight line such as straight line 63 that is substantially parallel to the left-right direction, that is, substantially orthogonal to the depth direction, it is necessary to exclude it. The extraction conditions are set so as to exclude straight lines extending in the left-right direction or height direction such as straight line 63.

[0071] Then, as shown in FIG. 3(B), for the vanishing point 64 obtained from straight lines such as straight lines 61 and 62 that are substantially parallel to the depth direction, auxiliary lines 65 are drawn from each of the front-side detection limit coordinates 6010 and 6020. For example, by setting the coordinates of two intersection points with the boundary line 66 drawn around the boundary between the outside and inside of the tunnel near the entrance / exit 55 of the tunnel as the back-side detection limit coordinates 6030 and 6040, all of the detection limit coordinates 6010 to 6040 are set, and the region that becomes the in-image event detection region 60 is determined.

[0072] In addition, when the fire detection system 1 performs fire detection, the in-image event detection region setting unit 2410 adds the in-image event detection region 60 set in the captured image 40 received from the image acquisition unit 22.

[0073] (b2. Projection image generation unit) Next, the projection image generation unit of the event detection unit will be described. In this description, refer to FIG. 4 showing the projection image and the projection division region. Note that FIG. 4(A) shows a captured image in which an in-image event detection region including illumination is set in the region, and (B) shows the projection image and the projection division region generated from the in-image event detection region of (A). Also, in the captured image and the projection image shown in FIG. 4, an X-axis (X1 axis or X2 axis) and a Y-axis (Y1 axis or Y2 axis) for indicating the coordinate position information on the image are shown, and the coordinates of the upper left pixel of the image are set to (0, 0).

[0074] In the fire detection device 20, it is necessary to detect the specular reflection and identify the position where the specular reflection occurs, specifically the position where the specular reflection occurs in the depth direction, so that the occurrence position of the fire can be identified along with the detection of the fire. Therefore, in order to generate an image that facilitates the identification of the position where the specular reflection occurs in the depth direction, the projection image generation unit 2420 performs a process of generating a projection image by performing a predetermined image process, such as a known projective transformation, on the image event detection region 60 in the captured image 40. Note that the processing method is not limited as long as it is an image process that facilitates the identification of the position where the specular reflection occurs in the depth direction.

[0075] The projective transformation by the projection image generation unit 2420 is a process of converting the captured image 40 into an image in which the axial direction of the X1 coordinate axis of the captured image 40 where the image event detection region 60 exists substantially coincides with the X2 coordinate axis that is the depth direction in the tunnel, and the axial direction of the Y1 coordinate axis substantially coincides with the Y2 coordinate axis that is the height direction in the tunnel. Specifically, using a predetermined projective transformation matrix consisting of 3 rows and 3 columns, each coordinate of the detection limit coordinates 6010 to 6040 in the captured image 40 represented by the X1 coordinate and the Y1 coordinate is converted into the coordinates of the projection image represented by the X2 coordinate (coordinate in the depth direction) and the Y2 coordinate (coordinate in the height direction), thereby converting the image event detection region 60 in the captured image 40 into an image represented by the coordinates in the depth direction and the coordinates in the height direction.

[0076] Since the projective transformation matrix can be obtained from the coordinates of four points in the captured image 40 (image represented by the X1 coordinate and the Y1 coordinate) and the corresponding four points in the projection image (image represented by the X2 coordinate and the Y2 coordinate), for example, markers are installed at the site to specify the positions of the four points in the monitoring area, and the coordinates of the four points in the captured image 40 and the coordinates in the projection image are identified, and the projective transformation matrix is obtained in advance from the coordinates of the four points in each image and stored in the storage unit 30.

[0077] In addition, the projection image generated by the projection image generation unit 2420 may perform a projection transformation other than the above. In the embodiment, for example, as shown in FIG. 4(B), after performing the projection transformation, the front side of the in-image event detection region 60 is located on the left side and the back side is located on the right side, and the coordinates of the detection limit coordinates 6010 located at the upper left of the in-image event detection region 60 are corrected to be (0, 0). A projection transformation process including a process of generating a projection image 70 of a color image with a size of 800x100 is performed.

[0078] In addition, since the captured image 40 is a two-dimensional image captured by perspective projection in a tunnel that is a three-dimensional space, a predetermined distance in the depth direction in the tunnel appears longer on the front side and shorter on the back side of the captured image 40. As shown in FIG. 4(A), the same-sized illuminations 56 provided at equal intervals in the in-image event detection region 60 appear larger in size and wider in the interval between the illuminations 56 on the front side, while they appear smaller in size and narrower in the interval between the illuminations 56 on the back side.

[0079] On the other hand, in the projection image 70, since a predetermined distance in the depth direction is represented by the same distance regardless of the position on the projection image 70, as shown in FIG. 4(B), the more the image is located on the back side of the in-image event detection region 60, the more it is an image that extends in the axial direction (depth direction) of the X2 coordinate axis compared to the captured image 40. The width of the illumination 56 and the interval between the illuminations 56 in the axial direction of the X2 coordinate axis are converted substantially uniformly regardless of the position.

[0080] In the projected image 70 of the embodiment, the distance corresponding to 100 pixels in the axial direction (depth direction) of the X2 coordinate axis is equivalent to the distance of 5 m in the depth direction inside the tunnel. For example, by storing in advance in the storage unit 30 the positional relationship between the position of X2 coordinate 0 in the projected image 70 and the position inside the tunnel, and the correspondence relationship between the distance in the axial direction of the X2 coordinate axis of the projected image 70 and the actual distance in the depth direction inside the tunnel (information such as 100 pixels = 5 m), it is possible to specify the occurrence position of the reflection in the actual depth direction inside the tunnel from the position of the reflection detected in the projected image 70.

[0081] (b3. Projected Image Division Unit) Next, the projected image division unit of the event detection unit will be described. In this description, continue to refer to FIG. 4.

[0082] The projected image division unit 2430 divides the projected image 70 generated by the above-described projected image generation unit 2420 from the in-image event detection region 60 in the captured image 40 into a plurality of projected division regions by dividing it with a predetermined width in the axial direction of the X2 coordinate axis that is the depth direction.

[0083] Here, the number of projected division regions to be divided and the size of each projected division region are arbitrary. In the embodiment, as shown in FIG. 4(B), the projected image division unit 2430 divides the projected image 70 into eight projected division regions 7010 to 7080 so that the width in the axial direction of the X2 coordinate axis is 100 pixels (corresponding to 5 m).

[0084] (b4. Event Judgment Unit) Next, the event judgment unit of the event detection unit will be described.

[0085] The event judgment unit 2440 determines whether or not a reflection has occurred in each of the eight projected division regions 7010 to 7080 divided by the projected image division unit 2430. Further, since the reflection is an event caused by the flame of a fire, similar to the judgment of the flame used in the conventional fire detection system, a feature amount of a flame-like feature is calculated, and based on that feature amount, it is determined whether or not a reflection has occurred.

[0086] As an example, in the embodiment, a case will be described in which the event determination unit 2440 determines whether or not there is a glare based on the feature amounts of "color" and "fluctuation" as the flame-like features.

[0087] First, the event determination unit 2440 selects one projected image division area, for example, the projected division area 7010, from the eight projected division areas 7010 to 7080, and extracts a glare candidate area that is an area showing a flame-like color based on a predetermined extraction condition from the projected division area 7010.

[0088] The extraction of the glare candidate area by the event determination unit 2440 is performed based on the color information in the projected division area 7010. Since the color information of the projected division area 7010 is expressed in RGB, for example, the event determination unit 2440 extracts, as a glare candidate area, an area having color information in which R, which is a red component, exceeds a predetermined threshold value as a predetermined extraction condition. Further, the extraction of the glare candidate area is not limited to the extraction based on R. The projected division area 7010 expressed in RGB may be converted into an HSV image that expresses the color information as a combination of the strengths of three elements of "hue" (H), "saturation" (S), and "value" (V), and the glare candidate area may be extracted based on the hue H or the saturation S.

[0089] In addition, when the event determination unit 2440 extracts the glare candidate area, it also calculates the ratio (= glare candidate pixel number / area pixel number) between the total number of pixels (area pixel number) of the projected division area and the number of pixels (glare candidate pixel number) of the glare candidate area.

[0090] Subsequently, when a specular reflection candidate region is extracted, the event determination unit 2440 calculates the fluctuation frequency of the image fluctuation in the specular reflection candidate region as a characteristic amount of a flame-like fluctuation after the extraction of the specular reflection candidate region. As a method for calculating the fluctuation frequency, for example, there is a method of calculating by performing a Fourier transform on the time-series change of the total luminance exceeding a predetermined threshold value in the specular reflection candidate region of the projection division regions 7010 that are serially continuous in time series.

[0091] When the calculated fluctuation frequency satisfies a predetermined frequency condition that can be regarded as the fluctuation frequency of a flame, the event determination unit 2440 determines that specular reflection has occurred in the projection division region 7010. In this way, by determining whether specular reflection has occurred based on the characteristics related to the flame fluctuation in accordance with the characteristics related to the flame color, it is possible to prevent misjudging non-fire factors such as painting or lighting that approximate the flame color as specular reflection.

[0092] Similar to the projection division region 7010, the event determination unit 2440 also determines whether specular reflection has occurred based on the characteristics related to the flame color and the flame fluctuation for the projection division regions 7020 to 7080. When the determination of all the projection division regions is completed, the specular reflection detection result including the ratio of the calculated region pixel number to the specular reflection candidate pixel number, the fluctuation frequency (characteristic amount), and the determination result of each projection division region is transmitted to the fire detection unit 26.

[0093] Also, the event determination unit 2440 does not determine whether specular reflection has occurred at the stage when the extraction of the specular reflection candidate region and the determination of the fluctuation frequency are completed in one projection division region. After the extraction of the specular reflection candidate region and the determination of the fluctuation frequency are completed in all the projection division regions, for the projection division regions where the specular reflection candidate regions are extracted, the correlation of the fluctuations of the adjacent projection division regions is determined. Based on the determination result of the fluctuation correlation, it may be determined whether specular reflection has occurred in each of the projection division regions.

[0094] For example, assuming that the reflection candidate area is extracted from three projection division areas 7040 to 7060 and the fluctuation frequency is calculated, the event determination unit 2440 determines the correlation between the fluctuations of the projection division area 7040 and the projection division area 7050 and the correlation between the fluctuations of the projection division area 7050 and the projection division area 7060 based on the fluctuation frequency. If the correlation of the fluctuations exceeds a predetermined threshold and is high, it is determined that reflection has occurred. If the correlation of the fluctuations is low and below the predetermined threshold, it is determined as a non-fire factor. Regarding the projection division area 7050, since there are two adjacent projection division areas and there are two correlations of fluctuations, that is, the correlation of fluctuations with the projection division area 7040 and the correlation of fluctuations with the projection division area 7060, for example, it is determined based on the average value of the correlations of the fluctuations.

[0095] In addition, if the fluctuation frequencies calculated in each of the projection division areas 7040 to 7060 satisfy a predetermined frequency condition that can be regarded as the fluctuation frequency of the flame, even if the correlation of the fluctuations is low and below the predetermined threshold, instead of determining it as a non-fire factor, it is determined as "there may be reflection", indicating that the accuracy of the reflection is lower than the determination of "reflection has occurred", so that a determination that distinguishes the accuracy of the reflection may be made.

[0096] In this way, by adding the correlation of the fluctuations as a determination condition, it becomes possible to more accurately determine whether reflection has occurred.

[0097] [c. Fire detection unit] Subsequently, the fire detection unit will be described. As described above, the fire detection unit 26 identifies the fire occurrence location and detects a fire based on the reflection detection result transmitted from the event detection unit 24.

[0098] (c1. Fire occurrence location identification unit) First, the fire occurrence position specifying unit of the fire detection unit will be described. In this description, refer to FIG. 5 showing a conceptual diagram of the projection division region when specular reflection occurs. Note that in the projection division region shown in FIG. 5, it is assumed that specular reflection occurs on the right side of the projection division region 7040, the entire area of the projection division region 7050, and the left side of the projection division region 7060.

[0099] The fire occurrence position specifying unit 2610 specifies the fire occurrence position based on the ratio of the number of area pixels to the number of specular reflection candidate pixels included in the specular reflection detection result transmitted from, for example, the event detection unit 24, from the projection division region determined to have specular reflection occurred.

[0100] When specifying the fire occurrence position from the specular reflection detection result for the projection division region shown in FIG. 5, in the determination process by the event determination unit 2440 of the event detection unit 24, for the projection division regions 7040 to 7060 where specular reflection has occurred, since the specular reflection candidate region 7090 has been extracted, the ratio of the number of area pixels to the number of specular reflection candidate pixels is calculated in accordance with the extraction of the specular reflection candidate region 7090, and the ratio of the number of area pixels to the number of specular reflection candidate pixels of the projection division region 7050 where specular reflection has occurred throughout the entire area is calculated to be the highest. Therefore, the fire occurrence position specifying unit 2610 specifies the position in the depth direction of the projection division region 7050 with the highest ratio of the number of area pixels to the number of specular reflection candidate pixels as the fire occurrence position in the depth direction.

[0101] In addition, alternatively, the fire occurrence position may be specified based on the amount of change in the pixel value of the projection division region when a fire occurs from the pixel value of the projection division region in the normal state where no fire has occurred.

[0102] For example, the in-image event detection area setting unit 2410 of the event detection unit 24 sets the in-image event detection area at the preparation stage before operating the fire detection system 1. At this time, the projected image generation unit 2420 generates a projected image from the set in-image event detection area, the projected image generation unit 2430 divides the projected image into projected division areas, and the event determination unit 2440 calculates the average pixel value of the entire area for each of the projected division areas and stores it in the storage unit 30 as a reference pixel value.

[0103] Then, when the event determination unit 2440 of the event detection unit 24 extracts a specular reflection candidate area during the fire detection of the fire detection system 1, the average pixel value of the entire projected division area is calculated accordingly, and the average pixel value of the entire projected division area is also included in the specular reflection detection result transmitted from the event detection unit 24.

[0104] The fire occurrence position specifying unit 2610 compares, for example, the image value of R with the average pixel value of the entire projected division area included in the specular reflection detection result and the reference pixel value, and specifies the position in the depth direction of the projected division area with the largest change amount as the fire occurrence position in the depth direction. Note that instead of processing as a color image, the image may be processed after being grayscale-converted.

[0105] (c2. Fire determination unit) Next, the fire determination unit of the fire detection unit will be described.

[0106] The fire determination unit 2620 determines that there is a fire, for example, when the detection result of specular reflection transmitted from the event detection unit 24 includes a projected division area in which it is determined that "specular reflection has occurred" in the determination result of each projected division area, and determines that there is no fire when the determination result of "specular reflection has occurred" is not included in the determination result of any projected division area.

[0107] In addition, as a condition for determining whether or not there is a fire, the fire determination unit 2620 may use elements other than the detection result of the specular reflection transmitted from the event detection unit 24. For example, similar to a conventional fire detection system, the fire determination unit 2620 may be configured to further detect the flames and smoke of a fire from the captured image, and determine whether or not there is a fire based on a condition combining the detection results of the flames and smoke of the fire and the detection result of the specular reflection.

[0108] For example, if flames or smoke of a fire can be detected from the captured image, both the determination based on the detection result of the flames or smoke of the fire and the determination based on the detection result of the specular reflection are performed. However, the determination result based on the detection result of the flames or smoke of the fire is weighted, and the determination results of both are comprehensively judged to determine whether or not there is a fire. If flames or smoke of a fire cannot be detected from the captured image, it may be configured to determine whether or not there is a fire based on the detection result of the specular reflection.

[0109] [d. Fire Detection Operation] Subsequently, the fire detection operation by the fire detection device will be described. In this description, reference is made to FIG. 6 showing an example of the fire detection operation in a flowchart, and FIG. 7 showing an example of the event detection operation in the fire detection operation shown in FIG. 6 in a flowchart.

[0110] As shown in FIG. 6, before starting fire detection, the fire detection device 24 sets an in-image event detection area 60 for a predetermined surface existing in the tunnel by the in-image event detection area setting unit 2410 and stores it in the storage unit 30 (step S1).

[0111] After the setting of the in-image event detection area 60 is completed, the fire detection device 24 starts fire detection. At a predetermined detection timing, the projection image generation unit 2420 performs predetermined image processing on the in-image event detection area 60 in the captured image 40 to generate a projection image, and the projection image division unit 2430 divides the generated projection image into projection division areas in the depth direction (steps S2 to S5).

[0112] Here, the predetermined detection timing is arbitrary timing. For example, when the average pixel value of the entire event detection region 60 in the image changes beyond a predetermined threshold, at a predetermined cycle, when an artificial operation is detected, etc., any condition can be set as the detection timing.

[0113] Subsequently, the event determination unit 2440 performs a detection operation for specular reflection on each of the divided projection divided regions (step S6). As shown in FIG. 7, the detection operation for specular reflection by the event determination unit 2440 first selects one unprocessed projection divided region from the projection divided regions divided by the projection image division unit 2430, and determines whether the selected projection divided region satisfies a predetermined extraction condition (steps S11 to S12).

[0114] When the event determination unit 2440 satisfies the predetermined extraction condition, it extracts a specular reflection candidate region, calculates the ratio of the number of region pixels to the number of specular reflection candidate pixels, and subsequently calculates the fluctuation frequency from the specular reflection candidate regions of the temporally consecutive projection divided regions, and determines whether the calculated fluctuation frequency satisfies a predetermined frequency condition that can be regarded as the fluctuation frequency of a flame. (Steps S13 to S16). On the other hand, when the event determination unit 2440 does not satisfy the predetermined extraction condition and does not extract a specular reflection candidate region, it determines that there is no event (step S19).

[0115] When the calculated fluctuation frequency satisfies the predetermined frequency condition, the event determination unit 2440 determines it as a specular reflection, and when the calculated fluctuation frequency does not satisfy the predetermined frequency condition, it determines it as a non-fire factor (steps S17, S18).

[0116] As described above, the determination process for one projection division area is completed, and the event determination unit 2440 determines whether there is an unprocessed projection division area. If there is an unprocessed projection division area, the processes of steps S11 to S19 are repeated until there is no unprocessed projection division area. If there is no unprocessed projection division area, a reflection detection result including the ratio of the calculated area pixel number to the reflection candidate pixel number, the frequency of fluctuation (feature amount), and the determination result of each projection division area is generated and transmitted to the fire detection unit 26 (steps S20, S21).

[0117] Returning to FIG. 6 again, the fire determination unit 26 that has received the reflection detection result from the event determination unit 2440 specifies the fire occurrence position based on, for example, the ratio of the area pixel number to the reflection candidate pixel number included in the reflection detection result transmitted from the event detection unit 24 by the fire occurrence position specifying unit 2610 (step S7).

[0118] Subsequently, the fire determination unit 26 determines whether it is a fire based on the determination result of each projection division area included in the reflection detection result transmitted from the event detection unit 24. If the determination result of the projection division area determined to be a reflection is included, it is determined to be a fire. If the determination result of a reflection is not included in the determination result of any projection division area, it is determined not to be a fire. The fire detection result including the fire occurrence position specified by the fire occurrence position specifying unit 2610 and the determination result determined by the fire determination unit 2620 is notified, for example, by display by the display operation unit 28 (steps S8, S9).

[0119] Thereafter, the fire detection device 24 detects a predetermined detection end timing, ends a series of detection operations, and returns to the process before step S3 again (step S10). Here, the predetermined detection end timing is arbitrary, and for example, when the completion of fire recovery is detected, when an artificial operation is detected, etc., any condition can be set as the detection end timing.

[0120] Also, after step S8 or step S9, the detection result of a fire or the detection result of an event, etc. may be transmitted to a predetermined external device, and the external device may be enabled to confirm the detection result of the fire or the event, display it, or provide an audible notification.

[0121] [e. Detection of events for multiple surfaces] Subsequently, the detection of events for multiple surfaces will be described. In this description, reference will be made to FIG. 8 showing an example of a captured image in which an in-image event detection area is set for the right wall surface and the left wall surface.

[0122] In the description of the fire detection system 1 up to this point, the surface for detecting the reflection (event) due to a fire by the event detection unit 24 has been the right wall surface 54 which is one surface, and the position of the fire occurrence specified by the fire determination unit 26 has been the position in the depth direction. However, the surface for detecting the reflection by the event detection unit 24 is not limited to one surface. By detecting the reflections occurring on multiple surfaces by the event detection unit 24, the fire determination unit 26 can specify not only the position in the depth direction but also the positions in other directions.

[0123] (e1. Event detection unit) The event detection unit 24 targets, for example, the left wall surface 53 and the right wall surface 54 as the surfaces for detecting the reflection.

[0124] The in-image event detection area setting unit 2410 sets a first in-image event detection area 80 for the left wall surface 53 and a second in-image event detection area 82 for the right wall surface 54 as shown in FIG. 8 at the preparation stage before operating the fire detection system 1. Also, the first in-image event detection area 80 and the second in-image event detection area 82 are set such that the areas in the depth direction are substantially the same.

[0125] Furthermore, the method for setting the in-image event detection area by the in-image event detection area setting unit 2410 is the same as the content described in "b1. In-image event detection area setting unit" except for setting two in-image event detection areas, and thus the description thereof is omitted. Also, the first in-image event detection area 80 and the second in-image event detection area 82 do not necessarily have the same size, and can be set to arbitrary sizes respectively.

[0126] The projection image generation unit 2420 performs a process of generating a first projection image and a second projection image by performing, as predetermined image processing, for example, a known projective transformation on each of the first in-image event detection area 80 and the second in-image event detection area 82 in the captured image 40. Since the generated first projection image and second projection image are basically the same as the projection image 70 shown in FIG. 4, the description thereof is omitted.

[0127] The projection image division unit 2430 divides each of the first projection image and the second projection image in the X2-axis direction, which is the depth direction, to divide the first projection image into a plurality of first projection divided areas and the second projection image into a plurality of second projection divided areas.

[0128] The event determination unit 2440 determines whether or not there is a specular reflection in each of the first projection divided areas and the second projection divided areas divided by the projection image division unit 2430. After the determination process is completed for all the projection divided areas, the ratio of the area pixel number calculated in the first projection divided areas to the specular reflection candidate pixel number, the frequency of fluctuation (feature amount), and the first detection result of specular reflection including the determination result of each first projection divided area, and the ratio of the area pixel number calculated in the second projection divided areas to the specular reflection candidate pixel number, the frequency of fluctuation (feature amount), and the second detection result of specular reflection including the determination result of each second projection divided area are transmitted to the fire detection unit 26. Furthermore, the method for determining an event by the event determination unit 2440 is the same as the content described in "b4. Event determination unit" except for performing the determination process on two in-image event detection areas, and thus the description thereof is omitted.

[0129] Currently, one event detection unit 24 detects events for the event detection areas within two images. However, the fire detection device 12 may be configured to include one event detection unit for each event detection area within one image. By corresponding the event detection areas within the image and the event detection units on a one-to-one basis, it becomes possible to perform the detection process for the specular reflection for the event detection areas within the image in parallel.

[0130] (e2. Fire detection unit) Based on the first detection result and the second detection result of the specular reflection transmitted from the event detection unit 24, the fire detection unit 26 will identify the occurrence position of the fire and detect the fire. As the occurrence position of the fire, in addition to the position in the depth direction, the position in the left-right direction is also identified.

[0131] The fire occurrence position identification unit 2610 identifies, for example, among the ratios of the number of area pixels in the first projection division area to the number of specular reflection candidate pixels included in the first detection result of the specular reflection transmitted from the event detection unit 24, the position of the first projection division area with the highest ratio as the first occurrence position of the specular reflection in the depth direction. Similarly, among the ratios of the number of area pixels in the second projection division area to the number of specular reflection candidate pixels included in the second detection result of the specular reflection, the position of the second projection division area with the highest ratio is identified as the second occurrence position of the specular reflection in the depth direction.

[0132] Then, the fire occurrence position identification unit 2610 identifies the occurrence position of the fire in the depth direction, for example, as the intermediate position between the first occurrence position of the specular reflection and the second occurrence position of the specular reflection in the depth direction.

[0133] Subsequently, the fire occurrence position identification unit 2610 identifies the occurrence position of the fire in the left-right direction based on, for example, a comparison between the total number of pixels (first total number of pixels) of the specular reflection candidate areas extracted in each first projection division area and the total number of pixels (second total number of pixels) of the specular reflection candidate areas extracted in each second projection division area.

[0134] Here, regarding the identification of the fire occurrence position in the left - right direction, the resolution at which it is performed is arbitrary. For example, it is assumed that the identification is made by three categories: "left side", "center", and "right side". For the identification of the three categories of "left side", "center", and "right side", for example, when the ratio of the first total number of pixels to the second total number of pixels (= first total number of pixels / second total number of pixels) exceeds a predetermined first threshold greater than 1, it is assumed that there is a flame on the first projection division region side, that is, on the left wall surface 53 side, and the fire occurrence position in the left - right direction is identified as "left side". When the ratio of the first total number of pixels to the second total number of pixels is less than a predetermined second threshold less than 1, it is assumed that there is a flame on the second projection division region side, that is, on the right wall surface 54 side, and the fire occurrence position in the left - right direction is identified as "right side". When the ratio of the first total number of pixels to the second total number of pixels falls between the first threshold and the second threshold, it is identified as "center".

[0135] In addition, it may also be possible to identify the fire occurrence position in the left - right direction based on the ratio of the first total number of pixels to the second total number of pixels (= first total number of pixels: second total number of pixels). For example, when the ratio of the first total number of pixels to the second total number of pixels is 7:3, the fire occurrence position in the left - right direction is identified as the position where the ratio of the distance from the fire occurrence position to the left wall surface 53 to the distance from the fire occurrence position to the right wall surface 54 is 3:7.

[0136] In addition, it may also be possible to identify the fire occurrence position in the left - right direction based on the total change amount of pixel values when a fire occurs from the normal state, instead of the total number of pixels.

[0137] The fire determination unit 2620 determines that there is a fire when, for example, the determination result of the projection division region determined as a reflection in the determination result of each first projection division region or second projection division region included in the first detection result or the second detection result of the reflection transmitted from the event detection unit 24 is included. When the determination result of the reflection is not included in the determination result of any projection division region, it determines that there is no fire.

[0138] As described above, by detecting the reflections occurring on the left wall surface 53 and the right wall surface 54 by the event detection unit 24 and specifying the occurrence position of the fire in the depth direction and the left-right direction by the fire determination unit 26, it becomes possible to specify the occurrence position of the fire with higher accuracy.

[0139] [f. Modification example of the present invention] A modification example of the fire detection system according to the present invention will be described. The fire detection system of the present invention includes the following modifications in addition to the above-described embodiment.

[0140] (Monitoring area) In the embodiment, the monitoring area is a tunnel, but the monitoring area is not limited to a tunnel, and the fire detection system of the present invention can be applied to any monitoring area where there is a surface on which a predetermined event occurs due to a fire.

[0141] (Fire detection system) In the embodiment, the configuration includes one imaging device and one fire detection device, but the target monitoring area may be divided into a plurality of sections, and an imaging device may be installed in each section. Further, when there are a plurality of imaging devices, the fire detection device may be one, or may include a plurality of fire detection devices corresponding to the number of imaging devices. Further, when a plurality of fire detection devices are provided, an upper device for managing the operations of all the fire detection devices may be provided.

[0142] (Fire detection device) In the embodiment, the fire detection device includes a display operation unit and a storage unit, but instead of the fire detection device including a display operation unit and a storage unit, a display operation device and a storage device may be provided externally, and the display operation device and the storage device may be connectable to the fire detection device. Further, although the detection result of the fire is notified by display by the display operation unit, a voice output unit may be provided so that it can be notified by voice.

[0143] (Setting of event detection area in image) In the embodiment, in order to basically enable the identification of the position in the depth direction that is the advancing direction of the tunnel, it has been described as an in-image event detection region extending in the depth direction. However, an in-image event detection region extending in the height direction may be set on a predetermined plane existing in the monitoring region to identify the position in the height direction, or an image event detection region extending in the left-right direction may be set on a predetermined plane existing in the monitoring region to identify the position in the left-right direction. Further, by setting a plurality of in-image event detection regions such as an in-image event detection region extending in the depth direction and an in-image event detection region extending in the left-right direction, it may be possible to identify the positions in a plurality of directions.

[0144] (Event to be detected) In the embodiment, the event occurring on the predetermined plane existing in the monitoring region detected by the event detection unit was the reflection due to a fire. However, the event occurring on the predetermined plane existing in the monitoring region to be detected is not limited to the reflection due to a fire. For example, the captured image may be a thermal image capable of detecting temperature, and the change in the surface temperature of the predetermined plane due to a fire may be detected as an event, or the change in the coating or painting applied to the predetermined plane due to a fire may be detected as an event.

[0145] (Others) Further, the present invention includes appropriate modifications that do not impair its object and advantages, and is not limited by the numerical values shown in the above embodiments.

Description of Reference Numerals

[0146] 1: Fire detection system 10: Imaging device 20: Fire detection device 22: Image acquisition unit 24: Event detection unit 2410: In-image event detection region setting unit 2420: Projection image generation unit 2430: Projection image division unit 2440: Event determination unit 26: Fire detection unit 2610: Fire occurrence position identification unit 2620: Fire determination unit 28: Display operation unit 30: Memory unit 40: Captured image 51: Ceiling surface 52: Floor surface 53: Left wall surface 54: Right wall surface 55: Tunnel entrance / exit 56: Lighting 60: Event detection area within image 6010~6040: Detection limit coordinates 61~63: Straight lines 64: Vanishing point 65: Auxiliary line 66: Boundary line 70: Projected image 7010~7080: Projective segmentation area 7090: Glare candidate area 80: First event detection area within image 82: Second event detection area within image

Claims

1. A fire detection system that detects a fire based on a captured image of a monitoring area captured by an imaging device, comprising: an event detection unit that detects, based on the captured image, an event that occurs on a predetermined surface existing in the monitoring area due to a fire occurring in the monitoring area; a fire detection unit that detects a fire by specifying the occurrence position of the fire based on the detection result of the event by the event detection unit; A fire detection system characterized by comprising the above.

2. The fire detection system according to Claim 1, wherein: the fire detection unit specifies, as the occurrence position of the fire, the position in the depth direction of the occurrence position of the fire in a predetermined one direction in the monitoring area.

3. The fire detection system according to Claim 2, wherein: the event detection unit sets an in-image event detection area for a predetermined area of the surface for detecting an event in the captured image, and detects an event occurring in the in-image event detection area; the fire detection unit specifies the position in the depth direction of the occurrence position of the fire based on the detection result of the event in the in-image event detection area by the event detection unit.

4. The fire detection system according to Claim 3, wherein: the event detection unit generates a projection image by performing predetermined image processing that substantially aligns the axial direction of any coordinate axis in the captured image with the depth direction with respect to the in-image event detection area in the captured image, and detects an event occurring in the in-image event detection area based on the projection image; the fire detection unit specifies the position in the depth direction of the occurrence position of the fire by specifying the position in the depth direction of the event detected by the event detection unit in the projection image.

5. The fire detection system according to Claim 4, wherein: the event detection unit further divides the projection image into a plurality of projection divided areas by dividing it with a predetermined width in the depth direction, and detects an event and a feature amount of the event for each projection divided area to detect an event occurring in the in-image event detection area; the fire detection unit specifies, as the position in the depth direction of the occurrence position of the fire, the position in the depth direction of the area in which the most event-like feature amount is detected by the event detection unit among the plurality of projection divided areas, or the position in the depth direction at the center of the area in which the event-like feature amount is detected.

6. The fire detection system according to Claim 5, wherein: The event detection unit further determines the correlation of image fluctuations in mutually adjacent projection division regions where an event has been detected, determines that an event has occurred for the projection division region in which it is determined that the correlation of the image fluctuations is high, and determines that a non-fire factor other than an event has occurred for the projection division region in which it is determined that the correlation of the image fluctuations is low. A fire detection system characterized by this.

7. A fire detection system according to claim 2, wherein the event detection unit performs a process of detecting an event for each of a plurality of surfaces existing in the monitoring region, and the fire detection unit specifies the occurrence position of the fire in a plurality of directions including the position in the depth direction of the occurrence position of the fire as the occurrence position of the fire based on the detection results of the events on the plurality of surfaces by the event detection unit. A fire detection system characterized by this.

8. A fire detection system according to claim 1, wherein the event detection unit detects, as the event, the specular reflection generated on a predetermined surface existing in the monitoring region by the flame of the fire. A fire detection system characterized by this.

Images