KR100922784B1 - Image base fire sensing method and system of crime prevention and disaster prevention applying method thereof - Google Patents

Abstract

PURPOSE: An image based fire sensing method, and a crime prevention and disaster prevention system using the same are provided to accurately sense a fire by detecting an optimum candidate region from an image change of a monitoring object. CONSTITUTION: An image based fire sensing method comprises the following steps: a step for receiving an image photographed through an image photographing device(S310); a step for setting an interest region after measuring a pixel change rate of the image according to a time(S330); a step for extracting a property of pixels included in the interest region(S340); a step for comparing the property of the pixels included in the interest region with a fire detecting candidate region; a step for calculating a fire alarm level of the pixels set into the fire detecting candidate region after storing coordinate information of the pixels set into the fire detecting candidate region(S370); and a step for performing a predetermined alarm operation corresponding to the fire alarm level.

Description

IMAGE BASE FIRE SENSING METHOD AND SYSTEM OF CRIME PREVENTION AND DISASTER PREVENTION APPLYING METHOD THEREOF}

The present invention relates to an image-based fire detection method and a security and disaster prevention system using the same, and more particularly, to a method and system for detecting a fire by using an image change of a monitoring target photographed in real time.

In general, in case of a fire in a mountain area, a building, a factory, etc., the first fire evolution of the fire is performed by using a CCTV camera or a fire detection sensor. That is, the fire detection and monitoring system performs an early response to a fire by detecting a fire point early and generating a fire alarm when a fire is detected through a CCTV camera or a fire detection sensor.

In particular, since there is a limit to using a fire detection sensor in a large area such as a mountain area or a public place, video-based surveillance and monitoring systems such as surveillance cameras are commercially available.

In general, in the case of a forest fire monitoring system, a surveillance camera installed on a mountain top rotates 360 degrees and transmits the captured image to the situation room, and the situation room monitors the mountain area by monitoring the received captured image. However, since the surveillance camera that rotates 360 degrees takes a long time to make one rotation, it cannot be responded quickly to the time when the fire was ignited. .

In addition, a system was developed to automatically detect abnormal signs by comparing images taken during the first rotation of the surveillance camera with images taken during the second rotation, but misunderstood a moving object such as a car or a moving person as a fire Problems such as lighting, light reflections, and shadows as fire have occurred.

Therefore, the present invention is to provide an image-based fire detection method and a crime prevention and disaster prevention system using the same to remove the noise that can be mistaken as a fire and to accurately detect the fire.

In accordance with an aspect of the present invention, there is provided an image-based fire detection method comprising: receiving an image captured by an image capturing apparatus and statistically measuring a pixel change amount of the photographed image according to time of interest; Setting a value, extracting a property of pixels included in the ROI, determining whether a fire detection candidate area corresponds to the fire detection candidate area from the properties of the pixels included in the ROI, and setting the fire detection candidate area Storing coordinate information of the pixels, calculating a fire alarm level of pixels set as the fire detection candidate region, and performing a predetermined alarm operation corresponding to the fire alarm level.

The method may further include converting the captured image into an image on a YCbCr color space.

In the setting of the ROI, a Gaussian distribution may be obtained with respect to the pixel change amount of the captured image, and the ROI may be set as the ROI when the pixel change amount is greater than a threshold value.

The extracting the characteristic of the ROI may extract at least one of a color characteristic, a brightness characteristic, a frequency characteristic, an amplitude characteristic, a number characteristic, a color distribution characteristic, and a motion characteristic of a pixel included in the ROI. .

The determining of whether the area corresponds to the fire detection candidate region may be determined using the following equation.

M (x, t) = I (I, t) -I (f, t) -I (α, t) -I (r, t) -I (c, t) -I (m, t)

Here, M (x, t) is a function indicating whether a pixel included in a region of interest corresponds to a fire detection candidate region, I (I, t) denotes a brightness characteristic of the pixel, and I (f, t) denotes the pixel. Where I (α, t) is the amplitude characteristic of the pixel, I (r, t) is the size characteristic of the pixel, I (c, t) is the color distribution characteristic of the pixel, and I ( m, t) is a function representing the motion characteristic of the pixel.

The calculating of the fire alarm level may be performed by using the following equation on the pixel set as the fire detection candidate region.

Q (t) = Q (t-1) + v (2M (x, t) -1),

Where Q (t) represents the fire alarm level at t time, and v represents the speed constant.

In the performing of the alarm operation in response to the fire alarm level, the fire alarm signal may be transmitted when the fire alarm level Q (t) maintains a value greater than or equal to a threshold value for a predetermined time.

An image-based fire detection apparatus according to another embodiment of the present invention, an image acquisition unit for receiving an image captured by the image capturing apparatus, interested in setting the region of interest by measuring the amount of pixel change of the captured image statistically over time A region determiner, a feature extractor which extracts characteristics of pixels included in the ROI, and a property of a pixel included in the ROI are determined to determine whether they correspond to a fire detection candidate region, and are set as the fire detection candidate region. And a pattern matching unit configured to calculate a fire alarm level of the pixels and to perform a predetermined alarm operation corresponding to the fire alarm level.

The image capturing apparatus includes a camera, and when the camera is a rotatable camera, the image capturing apparatus further comprises a central control unit for fixing the camera to capture the fire detection candidate region.

The image-based fire detection system according to another embodiment of the present invention, an image capturing apparatus for capturing an image through a camera, sets a region of interest by measuring the amount of pixel change of the captured image statistically over time, the region of interest A fire detection device extracting a property of pixels included in the device and determining whether the fire detection candidate area corresponds to a fire detection candidate area from the property of pixels included in the ROI, and calculating a fire alarm level of the pixels set as the fire detection candidate area; And an alarm generating device that performs an alarm operation in response to the fire alarm level.

As described above, according to the present invention, the optimum candidate region is detected from the image change of the monitoring target photographed in real time, and the characteristics of the candidate region are continuously extracted and analyzed to remove noise that may be mistaken as a fire, The malfunction rate can be reduced, so the fire can be detected accurately.

DETAILED DESCRIPTION Hereinafter, exemplary embodiments of the present invention will be described in detail with reference to the accompanying drawings so that those skilled in the art may easily implement the present invention.

1 is a view showing the configuration of a fire detection system according to a first embodiment of the present invention. As shown in FIG. 1, a fire detection system according to an exemplary embodiment of the present invention may include an image photographing apparatus 100, a fire detecting apparatus 200, and an alarm generating apparatus 300.

The image capturing apparatus 100 is individually provided at a position capable of acquiring a wide range of images or a main element of the surveillance region for monitoring a fire in a surveillance region such as a mountainous area or a public place. The image capturing apparatus 100 includes a device capable of capturing a moving image or a still image of one or more cameras, DVRs, CCTVs, and the like, and may also use an existing image capturing apparatus for security purposes.

In particular, the camera may include a pan / tilt rotating camera, a dome camera, an infrared camera, a fixed camera, a UTP camera, or the like. That is, the camera can be applied to both the rotating type and the fixed type, and it is more preferable to include a zoom-in and zoom-out function, a day / night conversion function, and the like.

The fire detection apparatus 200 analyzes the captured image received from the image capturing apparatus 100 and divides the image into a part that becomes a background and a part that has a motion. Then, the ROI is extracted from the moving parts, and the fire, flame, and smoke region of interest are extracted, and the fire is detected by analyzing the change of the ROI image over time.

When the fire detection signal is input from the fire detection device 200, the alarm generating device 300 notifies the fire information to the person in charge through a display device, an alarm alarm device, an SMS text notification system, an automatic pop-up system, and the like.

2 is a view showing the configuration of a fire detection apparatus according to an embodiment of the present invention. According to FIG. 2, the fire detection apparatus 200 includes an image acquirer 210, an ROI determiner 220, a feature extractor 230, and a pattern matcher 240.

The image acquisition unit 210 receives an image photographed by the image capturing apparatus 100 and converts the captured image of the RGB form into an image of the YCbCr form.

The region of interest determiner 220 detects a region of interest (ROI) that may be estimated as a fire by statistically measuring a color component of the photographed image and a change amount of movement of the target object over time. Here, fire includes smoke and flame, and different algorithms are applied to smoke and flame according to their characteristics.

The feature extractor 230 removes the noise included in the ROI by extracting the change characteristics of the brightness, the size, the color, the movement, the shape, and the like of the ROI with respect to the time change, and more accurately detects the smoke and the flame.

The pattern matching unit 240 determines whether it corresponds to a fire detection candidate region through a feature function of the ROI extracted from the feature extractor 230. In addition, when the fire alarm level of the pixels set as the fire detection candidate region is calculated and determined to be fire, the alarm generating device 300 is notified. Here, the pattern matching unit 240 may differently set the type of fire alarm to an initial alarm, a caution alarm, an emergency alarm, or the like according to a period during which the captured image is determined to be a fire by the characteristic function.

Hereinafter, a fire detection method of a fire detection system according to an embodiment of the present invention will be described. 3 is a flowchart illustrating a fire detection method of a fire detection system according to an exemplary embodiment of the present invention.

First, when the image capturing apparatus 100 captures a surveillance target region such as a mountainous region or a public place, the captured image is input to the fire detection apparatus 200 (S310). Here, the image captured by the image capturing apparatus 100 is an image on an RGB color space, and the image captured by the camera is a pixel composed of red (R), green (G), and blue (B). Expressed as a combination, each color pixel is represented again by the human eye acting on a different wavelength. Each color pixel is quantized to discrete levels and generally has 256 (8 bit per color plane) levels. For example, in the case of white (R, G, B) is represented by (255, 255, 255), in the case of black (R, G, B) is represented by (0, 0, 0). Each pixel constituting the image plane has a grid shape and can be represented by spatial coordinates (x, y), and color information of pixels located at spatial positions (x, y) is (R (x, y), G (x, y), B (x, y)).

However, although the RGB color space is easy to represent a color signal, there is a problem that it is not applied to the classification rule of smoke or flame pixels when the brightness of the image is changed because it is sensitive to the change in brightness.

Therefore, the image acquisition unit 210 converts the captured image on the RGB color space into the YCbCr color space where brightness and color are well distinguished (S320). According to an embodiment of the present invention, the smoke and flame color models are set using the YCbCr color space in which the RGB color space is linearly converted.

Equation 1 below shows the linear conversion of the RGB color space to the YCbCr color space.

In Equation 1, Y is luminance, Cb is blue color difference signal component (Chrominance Blue), and Cr is red color difference signal component (Chrominance Red). The range of Y value is [16,235], and the range of Cb and Cr is [16,240].

The ROI determiner 220 sets the ROI by statistically measuring the amount of pixel change of the captured image converted into the YCbCr color space over time (S330). That is, the ROI determiner 220 measures the amount of pixel change in the photographed image over time to distinguish between the background and the moving part, thereby moving the moving part into a region of interest (ROI). Set it. Here, the ROI determiner 220 repeatedly calculates a plurality of Gaussian distributions for pixels included in the captured image for a predetermined time, so that active pixels having a relatively large variation in pixel values are selected as the ROI. Can be set.

In general, the flame has a flickering property as the intensity of the emitted light increases and decreases, and this property appears as the increase and decrease of brightness in the captured image. Typically, the flame swing cycle is about 1-10 Hz. In the case of the smoke, the smoke generated from the ignition point has a property of spreading in a funnel shape by convection and the area is increased or decreased by the amount of smoke, and typically the fluctuation period of the smoke is about 2-5 Hz.

By using the motion characteristics of the flame and the smoke, the ROI-determining unit 220 detects an object whose movement period Hz corresponds to the movement period Hz of the flame and the smoke and sets it as the ROI. The region of interest determiner 220 calculates a Gaussian distribution of the pixels included in the captured image for statistical motion estimation, and applies a Gaussian mixture model (EMM) of an iterative technique to calculate the Gaussian distribution. Gaussian Mixture Model can be used. The Gaussian mixture model (GMM) may be expressed as Equation 2 below, and Equation 2 illustrates a Gaussian mixture model (GMM) for a region of interest estimated to be smoke.

,

,

Where p is the expected function, M is the number of Gaussian mixture models, and π _m represents the mixing coefficient.

On the other hand, the ROI may include not only smoke and flames, but also noises such as automobiles having a color similar to smoke or flame, people dressed in similar colors, and animals having similar hairs. Accordingly, the feature extractor 230 removes noise included in the ROI and more accurately detects smoke and flames, and compares the brightness and size of the pixel included in the ROI for a predetermined time period with respect to the captured image of the previous frame. In operation S340, the color, movement, shape, and the like are changed.

The fire includes smoke and flame, and the color of smoke and flame is the most distinguishable among the various characteristics of smoke and flame, so the color characteristics of smoke and flame will be described below.

First, the color characteristics of the smoke will be described. In the gray color, which is the color that most smoke includes, the Cb and Cr components are converted to 128 in the YCbCr color space, respectively. Using the following Equation 3 generated using the color characteristics of the smoke, the feature extractor 230 may detect whether the corresponding pixel can be estimated as smoke.

Here, S (x, y) is a smoke color characteristic function of the corresponding pixel (x, y), ε is a comparison constant, and the range of ε is typically 20 or less corresponding to a gray area. In Equation 3, | Cb (x, y) -128 | Value and | Cr (x, y) -128 | If all the values are smaller than ε, S (x, y) is 1. In Equation 3, if S (x, y) is 1, the corresponding pixel (x, y) is estimated to be similar to the color of smoke. If S (x, y) is 0, the corresponding pixel (x, y) is the color of smoke. It is assumed that it is not similar to.

Next, the color characteristics of the flame will be described. First, the average value (Y _mean , Cb _mean , Cr _mean ) for each YCbCr component in the captured image may be defined as in Equation 4 below.

Here, (x _i , y _i ) is the pixel coordinate of the captured image, and K is the total number of pixels of the captured image.

In general, the flame has a characteristic of being mainly red and yellow in the RGB color space, and the R value of the flame-generating region is larger than the R _mean value of the entire screen region in which the flame is generated. Such a relationship may be expressed as in Equation 5 below.

When the characteristic in the RGB color space as shown in Equation 5 is converted into the YCbCr color space, it can be expressed as shown in Equation 6 below.

As shown in Equation 6, the brightness component Y of the flame has a larger value than the Cb component, and the Cr component has a larger value than the Cb component. In addition, since the flame region generally includes the brightest region in the image, the Y component has a value larger than the average value of the Y component (Y _mean ), while the Cb component has a value smaller than the average value of the Cb component (Cb _mean ). The component represents a value larger than the _mean value Cr average of the Cr component. The following equation (7) can be obtained by considering the empirical color distribution of the flame region.

Here, F (x, y) is a flame color characteristic function of the corresponding pixel (x, y). If F (x, y) is 1 in Equation 7, the corresponding pixel (x, y) is the color of the flame. If F (x, y) is 0, then the pixel (x, y) is assumed to be not similar to the color of the flame.

On the other hand, in the case of the pixel corresponding to the flame, the color difference between the Cb component and the Cr component is very large.

Τ may be defined by the empirical value as a comparison constant.

As such, the feature extractor 230 may detect candidate regions estimated to be smoke and flame from the ROI by using Equations 7 and 8.

The feature extractor 230 may remove noise from pixels included in the region of interest by using not only the color characteristics of the fire but also brightness, size, movement, and shape characteristics. Hereinafter, other characteristics of the fire will be described. do.

First, the pixels contained in the region of interest _(ROI) (Ω ROI) the smoke or to the fire to determine if consistent with the brightness properties, the brightness value for the pixel (Ω _ROI) included in the region of interest _(ROI) I ROI (t ) The brightness value I _ROI (t) for the pixel Ω _ROI included in the _ROI may be expressed by Equation 9 below.

_ROI amplitude α (t) of the _ROI I (t) the frequency f _ROI (t) and, _ROI I (t) is to estimate the amount of change in the brightness value I by analyzing _ROI (t) over the time t.

Next, a pixel (Ω _ROI) the smoke or to the fire to determine if consistent with the size characteristics, number of pixels r _ROI (t) of (Ω _ROI) included in the region of interest (ROI) included in the region of interest (ROI) Obtain The number r _ROI (t) of the pixels Ω _ROI included in the _ROI may be defined as a size of Ω _ROI as shown in Equation 10.

Next, the pixel (Ω _ROI) color distribution characteristics of the pixel (Ω _ROI) included in the, region of interest (ROI) in order to determine if consistent with the color distribution characteristics of the smoke or fire contained in the region of interest (ROI) c _ROI ( t) Color distribution characteristic of the pixel Ω _ROI included in the _ROI c _ROI (t) is the number of pixels included in the _ROI with respect to the number of pixels Ω _ROI belonging to the color distribution interval Ω _C as shown in Equation 11 ρ. Defined by the ratio value divided by (t).

Here, ρ (t) can be defined as in the following equation for the smoke (smoke) and flame (flame).

The color distribution interval Ω _C applies differently to smoke and flame. The flame color gamut is relatively wide around the red color component to effectively cope with the error range of the low quality camera or camera color aberration, and the smoke color color range is set to the gray color range in the overall color gamut. Therefore, the estimation of flame or smoke may be determined according to whether a particular color pixel is concentrated in a specific section or spread in a wide range in the captured image.

In order to determine if the next line with pixels (Ω _ROI) motion distribution characteristics of the smoke or fire contained in the region of interest (ROI), the motion distribution characteristics of the pixel (Ω _ROI) included in the region of interest _(ROI) m ROI ( t) The movement characteristic m _ROI (t) of the pixel Ω _ROI included in the ROI for a predetermined time is defined as in Equation 13 below.

As shown in Equation 13, the motion characteristic m _ROI (t) of the active pixel Ω _ROI region may be expressed as a function of the center position, the distance standard deviation, and the velocity. That is, cm _ROI (t) is the center position of the pixels Ω _ROI included in the _ROI , and dm _ROI (t) represents the standard deviation of the distance away from the center position cm _ROI (t), sm _ROI (t) represents the moving speed of the center position.

In general, in the case of smoke, the ignition point has a form in which the periphery spreads far away while being almost maintained at a certain position, whereas in the case of clouds, the cloud itself moves in its entirety. Can be easily distinguished. Therefore, by calculating the speed and distance at which the center positions of the pixels Ω _ROI included in the region of interest ROI are moved, it may be determined whether they correspond to smoke or flame.

In addition, it may be determined whether the pixel Ω _ROI included in the _ROI corresponds to a shape characteristic of smoke or fire. For example, in the case of flames and car headlights, both have a similar radius and have a reddish hue. In the case of headlights, however, the radius is constant, whereas in the case of flames, the radius is not constant and the radius is irregular. Thus, even in the case of similarly shaped flames and automotive headlights, the circumferential length is much larger for flames. Accordingly, it may be determined whether the smoke or the flame corresponds to the shape change of the pixels Ω _ROI included in the _ROI .

As such, the feature extractor 230 extracts a change characteristic over time such as brightness, size, color, motion, and shape of the ROI, and the pattern matching unit 240 extracts the ROI extracted from the feature extractor 230. It is possible to determine whether a fire detection candidate region is applicable through a characteristic function of the pixels Ω _ROI included in the _ROI .

The pattern matching unit 240 may include a brightness characteristic I _ROI (t), a frequency characteristic f _ROI (t), and an amplitude characteristic α as a characteristic function of the pixels Ω _ROI included in the ROI described above. _ROI (t)), size characteristic (r _ROI (t)), color distribution characteristic (c _ROI (t)), and movement characteristic (m _ROI (t)) are substituted into Equation 14, respectively.

Where x _ROI (t) is the brightness characteristic (I _ROI (t)), the frequency characteristic (f _ROI (t)), the amplitude characteristic (α _ROI (t)), the magnitude characteristic (r _ROI (t)), and the color, respectively. Distribution characteristics (c _ROI (t)), movement characteristics (m _ROI (t)), and μ _low (t) and μ _high (t) are empirically obtained thresholds for determining whether each characteristic value is included in a fire. Value. In addition, I (x, t) is a function indicating whether each characteristic of the pixels Ω _ROI included in the _ROI corresponds to a fire. When I (x, t) is 1 in Equation 14, the corresponding pixel ( Ω _ROI ) is estimated to be similar to the characteristics of the fire, and if I (x, t) is 0, the corresponding pixel Ω _ROI is not assumed to be similar to the characteristics of the fire.

When the characteristics of the pixels Ω _ROI included in the ROI shown in Equation 14 are combined and represented as one equation, Equation 15 is obtained.

M (x, t) = I (I, t) -I (f, t) -I (α, t) -I (r, t) -I (c, t) -I (m, t)

Here, M (x, t) is a fire detection function indicating whether a pixel Ω _ROI included in the _ROI corresponds to a fire, and I (I, t) represents a brightness characteristic of the pixel Ω _ROI . I (f, t) is the frequency characteristic of pixel Ω _ROI , I (α, t) is the amplitude characteristic of pixel Ω _ROI , and I (r, t) is the magnitude characteristic of pixel Ω _ROI . , I (c, t) is a function indicating the color distribution characteristic of the pixel Ω _ROI , and I (m, t) is a motion characteristic of the pixel Ω _ROI .

As shown in Equation 15, if any one of the characteristic functions I (x, t) for the pixel Ω _ROI included in the ROI in Equation 14 has a value of 0, M (x, t) will have a value of 0. On the other hand, in Equation 14, brightness characteristic (I (I, t)), frequency characteristic (I (f, t)), amplitude characteristic (I (α, t)), magnitude characteristic (I (r, t)), color When the values of the distribution characteristics I (c, t) and the motion characteristics I (m, t) are all 1, M (x, t) has a value of 1.

Accordingly, the pattern matching unit 240 determines whether the pixel Ω _ROI included in the _ROI corresponds to the fire detection candidate region through Equation 15. That is, the pattern matching unit 240 recognizes the fire detection candidate region when M (x, t) is 1.

If M (x, t) is 1, the pattern matching unit 240 designates a corresponding pixel Ω _ROI as a candidate region (S350), and transmits coordinate information of the corresponding pixel to the ROI determiner 220. Transfer (S360). The ROI determiner 220 measures statistical movement of pixels corresponding to the candidate area over time, and sets the ROI as the ROI when the measurement result of the pixels is larger than a threshold. The pixel set as the ROI is determined to continuously correspond to a fire through steps S330 to S350.

Meanwhile, the pattern matching unit 240 may calculate the fire alarm level Q (t) by using Equation 16 (S370).

Q (t) = Q (t-1) + v (2M (x, t) -1), 0≤Q (t) ≤Q _T

Here, Q (t) represents an alarm level for a fire, v is a speed constant, and Q _T can be arbitrarily set by experience as a threshold value. Hereinafter, Equation 16 will be described in more detail with reference to FIG. 4.

4 is an exemplary view for explaining a fire alarm graph according to an embodiment of the present invention. As shown in FIG. 4, the fire alarm level Q (t) increases or decreases according to the output value of M (x, t).

According to FIG. 4, the pattern matching unit 240 transmits a fire alarm signal to the alarm generating device 300 when the fire alarm level Q (t) is equal to or greater than Q ₀ , and Q (t) maintains a Q _T value. The type of fire alarm such as an initial alarm, a caution alarm, or an emergency alarm may be determined according to time (S380). That is, the longer the time Q (t) maintains the Q _T value means that a fire is spreading, the emergency alert signal is transmitted to the alarm generating device 300.

As described above, according to the first exemplary embodiment of the present invention through FIGS. 1 to 4, an optimal candidate region is detected from an image change of a monitoring target photographed in real time, and the features of the candidate region are continuously extracted and analyzed to thereby fire. This can eliminate false noise and reduce the malfunction rate of fire detection.

On the other hand, the fire detection system according to an embodiment of the present invention can be modified in various forms, it will be described below with reference to FIGS.

5 is a view showing the configuration of a fire detection system according to a second embodiment of the present invention. As shown in FIG. 5, the fire detection system according to the second embodiment of the present invention may include an image photographing device 102, a fire detection device 202, and an alarm generating device 302.

First, when the image is captured using the image capturing device 102 such as a pan / tilt rotating camera, a dome rotating camera, a general fixed camera, and a dome fixed camera, the image is transmitted to the fire detection device 202. According to the second embodiment of the present invention, the image captured by the image capturing apparatus 102 is transmitted to the fire detection apparatus 202 through the image cable. When the fire detection device 202 is recognized as a fire and transmits a fire alarm signal to the alarm generating device (302). Here, the alarm generating device 302 includes a warning light siren, a speaker, a display device, and the like as shown in FIG. 5.

6 is a view showing the configuration of a fire detection system according to a third embodiment of the present invention. According to the third embodiment of the present invention, the image photographed by the image capturing apparatus 103 is transmitted to the fire detection apparatus 203 at the speed of light through the light conversion transmitter and the optical cable.

7 is a view showing the configuration of a fire detection system according to a fourth embodiment of the present invention. According to the fourth embodiment of the present invention, the image captured by the image capturing apparatus 104 is transmitted to the fire detection apparatus 204 in a compressed form through an image compression transmission unit and a network.

8 is a view showing the configuration of a fire detection system according to a fifth embodiment of the present invention. According to the fifth embodiment of the present invention, the image photographed by the image capturing apparatus 105 is wirelessly transmitted to the fire detection apparatus 205 through the image compression transmitter and the wireless transmitter.

9 is a view showing the configuration of a fire detection system according to a sixth embodiment of the present invention. The fire detection system according to the sixth embodiment of the present invention further includes a control unit 406 unlike the first to fifth embodiments. The integrated monitoring device 406 includes a central control device and a camera control device, and is connected to the fire detection device 206 by a network to enable remote monitoring. When the central control unit receives a fire detection signal from the fire detection unit 206 or the alarm generating unit 306, the camera control unit can take a pan / tilt rotating camera or dome rotating camera to capture the image of the fire. Rotate and lock it so that you can focus on the fire scene. Thus, the operator of the fire detection system can recognize the fire situation more quickly and accurately.

The control unit 406 may be located outside the fire detection device 206 and connected to the fire detection device 206 through a network as shown in FIG. 9, or may be included in the fire detection device 206.

10 is a view showing the configuration of a fire detection system according to a seventh embodiment of the present invention. According to the seventh exemplary embodiment of the present invention, the plurality of fire detection systems 110a including the image photographing device 107, the fire detection device 207, and the alarm generating device 307 are connected to each other through a network. , The network is connected to the integrated monitoring device 407. Therefore, remote monitoring of the plurality of fire detection systems 110a through the network is possible through the central monitoring PC included in the integrated monitoring device 407. Here, the network is based on TCP / IP such as LAN, ADSL, dedicated line, and can transmit information to the integrated monitoring device 407 at high speed in real time.

Meanwhile, in the embodiments of the present invention, a fire prevention system for fire prevention purposes has been mainly described. However, by using the methods described above, the movements, clothes, and movement paths of a thief who break into a house may be analyzed and applied to a security system for security purposes. It may be.

Although the preferred embodiments of the present invention have been described in detail above, the scope of the present invention is not limited thereto, and various modifications and improvements of those skilled in the art using the basic concepts of the present invention defined in the following claims are also provided. It belongs to the scope of rights.

1 is a view showing the configuration of a fire detection system according to a first embodiment of the present invention.

2 is a view showing the configuration of a fire detection apparatus according to an embodiment of the present invention.

3 is a flowchart illustrating a fire detection method of a fire detection system according to an exemplary embodiment of the present invention.

4 is an exemplary view for explaining a fire alarm graph according to an embodiment of the present invention.

5 is a view showing the configuration of a fire detection system according to a second embodiment of the present invention.

6 is a view showing the configuration of a fire detection system according to a third embodiment of the present invention.

7 is a view showing the configuration of a fire detection system according to a fourth embodiment of the present invention.

8 is a view showing the configuration of a fire detection system according to a fifth embodiment of the present invention.

9 is a view showing the configuration of a fire detection system according to a sixth embodiment of the present invention.

10 is a view showing the configuration of a fire detection system according to a seventh embodiment of the present invention.

Claims (17)

Receiving an image captured by the image capturing apparatus, Setting a region of interest by statistically measuring a change amount of pixels of the captured image according to time; Extracting characteristics of pixels included in the ROI; Determining whether a fire detection candidate region corresponds to a characteristic of pixels included in the ROI; Storing coordinate information of pixels set as the fire detection candidate region, calculating fire alarm levels of pixels set as the fire detection candidate region, and Performing a predetermined alarm operation in response to the fire alarm level, Determining whether the fire detection candidate area corresponds to, Image-based fire detection method characterized in that determined using the following equation: M (x, t) = I (I, t) -I (f, t) -I (α, t) -I (r, t) -I (c, t) -I (m, t) Here, M (x, t) is a function indicating whether a pixel included in a region of interest corresponds to a fire detection candidate region, I (I, t) denotes a brightness characteristic of the pixel, and I (f, t) denotes the pixel. Where I (α, t) is the amplitude characteristic of the pixel, I (r, t) is the size characteristic of the pixel, I (c, t) is the color distribution characteristic of the pixel, and I ( m, t) is a function representing the motion characteristic of the pixel. The method of claim 1, Converting the photographed image into an image on a YCbCr color space; The setting of the ROI may include obtaining a Gaussian distribution for the pixel variation of the captured image and setting the ROI as the ROI when the pixel variation is greater than a threshold. delete delete delete The method of claim 1, The fire alarm level calculation, The pixel set as the fire detection candidate region is calculated using the following equation, Q (t) = Q (t-1) + v (2M (x, t) -1), Where Q (t) represents the fire alarm level at time t, v represents the rate constant, And performing the alarm operation when the fire alarm level (Q (t)) maintains a value greater than or equal to a threshold value for a predetermined period of time. delete Image acquisition unit for receiving an image captured by the image capture device, A region of interest determiner configured to set a region of interest by statistically measuring the amount of change in pixels of the captured image; A feature extractor which extracts features of pixels included in the ROI, and It is determined whether the fire detection candidate region corresponds to the fire detection candidate region from the characteristics of the pixels included in the ROI, the fire alarm level of the pixels set as the fire detection candidate region is calculated, and a predetermined alarm operation corresponding to the fire alarm level is performed. Including a pattern matching unit to perform, The pattern matching unit, An image-based fire detection device, characterized in that it is determined whether the fire detection candidate region corresponds to the following equation: M (x, t) = I (I, t) -I (f, t) -I (α, t) -I (r, t) -I (c, t) -I (m, t) Here, M (x, t) is a function indicating whether a pixel included in a region of interest corresponds to a fire detection candidate region, I (I, t) denotes a brightness characteristic of the pixel, and I (f, t) denotes the pixel. Where I (α, t) is the amplitude characteristic of the pixel, I (r, t) is the size characteristic of the pixel, I (c, t) is the color distribution characteristic of the pixel, and I ( m, t) is a function representing the motion characteristic of the pixel. The method of claim 8, The image acquisition unit converts the captured image into an image on a YCbCr color space, And the ROI determiner obtains a Gaussian distribution with respect to the pixel variation of the captured image, and sets the ROI when the pixel variation is greater than a threshold. delete delete delete The method of claim 8, The fire alarm level is calculated for a pixel set as the fire detection candidate region by using the following equation, Q (t) = Q (t-1) + v (2M (x, t) -1) Where Q (t) represents the fire alarm level at time t, v represents the rate constant, And an alarm operation when the fire alarm level Q (t) maintains a value greater than or equal to a threshold value for a predetermined time. delete The method of claim 8, The pattern matching unit, And the coordinate information of pixels set as the fire detection candidate region is transmitted to the ROI determiner. delete Imaging device for taking a picture of the monitoring target, Sets a region of interest by measuring the amount of change in pixels of the photographed image statistically over time and extracts the characteristics of pixels included in the region of interest to determine whether the fire is a candidate region from the characteristics of the pixels included in the region of interest. A fire detection device for determining a fire alarm level of pixels set as the fire detection candidate area; And an alarm generating device configured to perform an alarm operation in response to the fire alarm level.

Images