KR20230107035A - Image recognition-based fire response system and method for power facility - Google Patents

Abstract

A fire response method for power facilities according to the present invention may include the steps of: recognizing the occurrence of a fire based on image recognition; separately extracting flame and smoke from the fire, and calculating the area of the extracted flame part; estimating a fire spread speed as a deviation of the area of the flame part changing with time; and taking measures according to the fire spread speed to protect a target facility.

Description

Fire response method and image recognition based fire response system for power facilities {IMAGE RECOGNITION-BASED FIRE RESPONSE SYSTEM AND METHOD FOR POWER FACILITY}

The present invention is for power facilities that more simply and accurately detects/identifies power facilities or nearby fires using image analysis such as depth maps, and takes measures to protect power facilities through fire separation distance estimation. It relates to fire response methods and systems.

More than 70% of Korea's land area is made up of forests. While economic power has improved since the 1970s, forests have become dense, while many forest fires have occurred due to the carelessness of hikers who have increased on holidays. is happening In addition, the causes of forest fires are diverse.

Recently, fires are increasing nationwide due to various factors. In particular, large and small forest fires in mountainous areas are causing social problems such as damage to forests, physical and human damage, and loss of cultural assets.

Accordingly, the necessity of introducing early fire detection technology based on video technology is increasing in order to detect forest fires early and minimize damage.

Electric power companies use CCTV surveillance systems and forest fires to prepare for early detection of damage that may occur to transmission and distribution facilities such as steel towers and power lines in the fire path due to forest fires that occur near power facilities, and to preemptively shut down the grid and secure alternative grids. We operate our own response system, such as surveillance patrols.

In addition, power companies in various countries as well as the United States are operating CCTV-based monitoring systems in vulnerable areas such as strong winds at all times to prepare for power line disconnection fires caused by bad weather such as strong winds.

However, since this monitoring system is based on manpower-based constant monitoring, it is difficult to detect fire occurrence in real time and determine whether or not the fire is close to the object to be monitored. Therefore, it is urgent to introduce a technology that recognizes objects to be monitored, such as electric power facilities, by applying intelligent technology to the existing CCTV surveillance system, estimates the location of a fire from a forest fire, determines whether a fire occurs, proximity, and sends an alarm.

Currently, the fire detection system mainly used is a method of detecting a fire using CCTV video surveillance, an infrared camera, and a sensor. Fire detection equipment based on sensor technology, such as infrared, requires relatively high investment costs and has a limited effective detection radius of several tens of meters, making it difficult to use in wide areas such as mountainous areas.

In addition, due to the accuracy limit of the sensor technology, there are many cases in which false fire recognition alarms frequently occur, which often reduces the user's work efficiency.

Most of the current CCTV surveillance systems depend on the operator's visual monitoring, and it is difficult to recognize the initial fire occurrence late or to immediately determine how close the fire is to the object to be monitored, such as power facilities or cultural assets. there is a problem

In particular, in the case of electric power facilities in mountainous/forest areas and places where management is difficult due to manpower resident for wide-area system network connection, despite fire recognition and monitoring through CCTV, etc. It is difficult to grasp fire information such as the direction of movement and the speed of spread, so in most cases it is impossible to take initial action to prevent fires in facilities.

Therefore, it is urgent to introduce a technology capable of learning and automatically recognizing an object to be monitored using image sensor-based technology, estimating and determining whether or not a fire occurs and the proximity and separation distance from the object to the fire.

Republic of Korea Publication No. 10-2020-0080579

Even in the case of power facilities in places where resident management is difficult, fire information such as fire size, movement direction, and spread speed can be identified and coped with by recognizing/monitoring fire occurrence through CCTV in case of nearby forest fires and fires. It is intended to provide a fire response method and system for power facilities through image recognition-based fire information extraction and fire separation distance estimation.

A fire response method for a power facility according to an aspect of the present invention includes the steps of recognizing the occurrence of a fire based on image recognition; Separately extracting a flame and smoke portion for the fire that occurs, and calculating an area for the extracted flame portion; estimating a fire spread rate as a deviation of the area of the flame portion that changes with time; and performing a measure according to the fire spreading speed to protect the target facility.

Here, in the step of estimating the fire spread rate, the fire spread rate can be calculated through the spread area versus time based on the deviation of the flame portion extracted at regular time intervals through time monitoring.

Here, in the step of estimating the fire spread speed, the fire spread speed is estimated by changing the length of the four sides of the rectangular frame including all the recognized flame parts, but at a specific time based on area information collected at regular time intervals. The length of each side of the rectangular frame is extracted, and the fire spread rate can be calculated from the length change of each side after the lapse of time.

Here, in the step of estimating the fire spread rate, the fire spread rate may be presented by applying the maximum value among length variations of each of the four sides of the rectangular frame.

Here, in the step of estimating the fire spread rate, the fire spread rate may be estimated from a change in the circumference or diameter of the circular frame including all of the recognized flame parts.

Here, recognizing the occurrence of a fire may include extracting distance information related to a location of a target facility based on a depth map of the image; Detecting flame and smoke of a fire based on the color image of the image and recognizing a target facility; and determining a degree of separation of the fire from the target facility based on the distance information, the detection result, and the recognition result.

Here, in the step of determining the degree of separation, if the detection area of the detected target facility overlaps with the fire detection area by a predetermined ratio or more and the depth map intensity difference of the detection area is equal to or less than a predetermined reference value, it is regarded as a proximity fire for the target facility. can judge

Here, in the step of performing the measures, if it is determined that the fire reaches the target facility quickly according to the degree of separation between the target facility and the fire location and the speed of fire spread, take immediate action for the target facility. can instruct

A fire response system for an image recognition-based power facility according to another aspect of the present invention includes a fire image analysis device that analyzes an image taken in a fire monitoring area to determine the degree of separation of a fire from a target facility in the monitoring area; a fire spread rate deriving device for deriving a fire spread rate based on the deviation of the flame portion of the image; And it may include a fire action device for performing fire response measures for the target object at a response level determined according to the fire spread rate and the degree of separation of the fire from the target facility.

Here, the fire spread rate derivation device, a flame / smoke classification unit for dividing a flame portion and a smoke portion in the image taken with respect to the fire occurred; a flame portion extraction unit for calculating a maximum area for the extracted flame portion; and a fire spread rate estimator for estimating a fire spread rate as a deviation of the flame portion that changes over time.

Here, the fire spread rate derivation device may further include a flame portion deviation calculation unit that calculates a deviation of the flame portion for each predetermined period of time.

Here, the fire spread rate derivation device comprises the steps of dividing and extracting a flame part and a smoke part for an outbreak fire, and calculating an area for the extracted flame part; and estimating the fire spread rate as a deviation of the area of the flame portion that changes over time.

Here, the fire spread rate estimator estimates the fire spread rate by changing the length of four sides of a rectangular frame including all of the recognized flame parts, and the rectangular frame at a specific time based on area information collected at regular time intervals. The length of each side of is extracted, and the fire spread rate can be calculated from the change in the length of each side after the lapse of time.

Here, the fire action device may instruct immediate measures for the target facility when it is determined that the fire reaches the target facility quickly according to the distance between the target facility and the fire location and the speed of the fire spread. can

Here, the fire image analysis device extracts distance information related to the location of the target facility based on a depth map of the image, detects flames and smoke of a fire based on the color image of the image, and targets the facility Recognizes, if the detection area of the target facility and the detection area of fire or smoke overlap at a predetermined rate or more, and the depth map intensity difference between the overlapping detection areas is less than a predetermined reference value, it can be determined as a proximity fire to the target facility.

Based on the image recognition according to the spirit of the present invention having the above configuration If a fire response method and/or system is implemented for power facilities, even in the case of power facilities in places where resident management is difficult, in case of forest fires and fires in the vicinity, fire recognition/monitoring through CCTV, etc. It has the advantage of being able to deal with fire information such as the speed of spread.

Image recognition basis of the present invention A fire response method and/or system for power facilities has the advantage of preventing forest loss through early detection of forest fires and rapid initial action and minimizing human and material damage to cultural properties, important structures, and power facilities caused by fire. there is

Image recognition basis of the present invention The fire response method and/or system for power facilities estimates the distance from the object requiring fire accident prevention to the point of fire occurrence, enabling rapid situation judgment and initial action, and efficiency in manpower operation through AI-based automatic fire detection. has the advantage of increasing

1 is a flowchart illustrating an embodiment of a fire response method for power equipment according to the spirit of the present invention.
2 is an exemplary view of a method for estimating the fire spread rate based on image recognition that can be applied in the present invention.
3 is a specific calculation example showing a flame spread rate estimation method through real-time deviation of a flame part in the fire spread rate estimation scheme of FIG. 2;
Figure 4 is a block diagram showing an embodiment of a fire management system using image recognition information for fire according to the spirit of the present invention.
5 is a block diagram of a configuration for calculating the separation distance from a target power facility to a location where a fire occurs.
6 is an exemplary view of a fire detection screen near power facilities;
7 is a block diagram showing an example of a detailed configuration of the detection area position coordinate calculation module 300 of FIG. 5;
8 is an exemplary diagram of detection results for fire and target power facilities.
FIG. 9 is a block diagram showing an example of a detailed configuration of the fire location final determination module 400 of FIG. 5;
10 is a flowchart illustrating an example of a reoccurrence position final determination logic;

In describing the present invention, terms such as first and second may be used to describe various components, but the components may not be limited by the terms. Terms are only for the purpose of distinguishing one element from another. For example, a first element may be termed a second element, and similarly, a second element may be termed a first element, without departing from the scope of the present invention.

When a component is referred to as being connected or connected to another component, it may be directly connected or connected to the other component, but it may be understood that another component may exist in the middle. .

Terms used in this specification are only used to describe specific embodiments, and are not intended to limit the present invention. Singular expressions may include plural expressions unless the context clearly dictates otherwise.

In this specification, the terms include or include are intended to designate that there is a feature, number, step, operation, component, part, or combination thereof described in the specification, and one or more other features or numbers, It can be understood that the presence or addition of steps, operations, components, parts, or combinations thereof is not precluded.

In addition, shapes and sizes of elements in the drawings may be exaggerated for clearer description.

As described above, the existing video-based fire monitoring system is aimed at minimizing human damage such as fire suppression and evacuation in places adjacent to the fire source for the purpose of fire monitoring in narrow places such as the vicinity of video equipment and inside buildings. As a direction for the safety management system, there is a part that has been developed with limited functions.

However, in the case of power facilities in mountainous/forest areas where manpower access is not easy, it is impossible to move the facilities, and fire prevention for facilities is important in the event of a remote fire at an unspecified point such as a forest fire. There is a need for an intelligent fire monitoring system so that fire information, such as the expected access time from the fire source to the facility, can be linked with the safety management system so that effective first action can be taken.

1 is a flowchart illustrating an embodiment of a fire response method for a power facility according to the spirit of the present invention.

The fire response method for the illustrated power facility includes recognizing the occurrence of a fire through an image recognition-based classification technique (S20); Separately extracting a flame and a smoke part for the generated fire, and calculating a maximum area for the extracted flame part (S40); Dividing the fire scale in stages based on the flame portion and estimating the fire spread rate as a deviation of the fire area that changes with time (S60); and performing a measure according to the fire spreading speed (S80).

In the step of estimating the fire spread information (S60), the fire spread rate can be estimated through the spread area versus time based on the deviation of the flame portion extracted at regular time intervals through time monitoring.

2 illustrates an image recognition-based fire spread rate estimation method applicable to the step S60.

2 is an example of a method of estimating the fire spread rate based on image recognition information about a fire. Through an image recognition-based classification technique, flame and smoke parts are classified for an outbreak fire, and the maximum area for the extracted flame part is calculated. yield Through real-time monitoring, it is possible to estimate the fire spread rate through the spread area versus time based on the deviation of the flame parts extracted at regular time intervals.

FIG. 3 shows a flame spread rate estimation method through real-time deviation of a flame part in the fire spread rate estimation scheme of FIG. 2 .

3 is an example of a method for estimating the flame spread rate through the deviation of the flame part that changes in real time with respect to the recognized flame part, the flame part is based on the maximum area for the area can be calculated.

For example, based on area information collected in real time or at regular time intervals, the lengths of A and B are extracted at a specific point in time, and each area compilation A'-A and B'- for A' and B' after time lapse C Calculate the flame spread rate with (A'-A)/C and (B'-B)/C by reflecting the elapsed time C based on B. In addition, for quick safety management of fire risk, it is possible to apply and present the maximum value of the diffusion speed in the above two cases.

In FIG. 3, the flame spread speed is estimated by changing the length of the four sides of a rectangular frame including all recognized flame parts. This method can apply the X-Y coordinate system composed of most image sensor pixels as it is, It has the advantage of being fast.

In other implementations, the flame spread rate is calculated from changes in the area of recognized flame parts, or from changes in the diameter or circumference of a circular frame that includes all recognized flame parts, or includes all recognized flame parts. The flame spread rate can be estimated by the change in the length of the sides of other polygonal frames, such as hexagons.

Depending on the implementation, recognizing the occurrence of a fire (S20) may include extracting distance information based on a gray image; Detecting flames and smoke of a fire based on a color image and recognizing a power facility; and determining a fire occurrence location (separation distance as a degree of separation) for the power facility from the results of the two steps.

In this case, in the step of recognizing the fire (S20), the separation distance between each power facility and the location (region) of the fire is estimated. For example, as the separation distance estimation method, deep learning-based object detection and location coordinates, It can be calculated by estimating an approximate distance through the fusion of stereo camera brightness information and turning it into a table.

For example, in the step of determining the degree of separation, the detection area of the detected target facility and the fire detection area overlap with a predetermined ratio (80%) or more, and the depth map intensity difference of the detection area is less than a predetermined reference value (10%), It can be judged as a proximity fire to the target facility.

In accordance with the idea of the present invention, if the above steps are performed (S80), the separation distance between the fire source and the location of the power facility is estimated based on the image information acquired from the plurality of cameras, and the fire spread speed and distance By reflecting the distance, it is possible to efficiently prevent and respond to fires according to the order of fire access to nearby power facilities. Countermeasures are derived step by step according to fire information.

For example, in the step of performing the action (S80), if it is determined that the fire reaches the power facility quickly according to the distance between the power facility and the location where the fire occurred and the speed of fire spread, the immediate response to the power facility is determined. Actions (e.g., suspension of operation of the power facility and/or line blocking) are instructed, and if it is determined that there is sufficient time for the fire to reach the power facility, only notification and alarm actions are performed to the manager, and direct actions are taken by the manager. Can be done according to instructions.

For example, when there is only one fire area, the priority of fire prevention may be determined according to a separation distance between each power facility and a fire location (source).

For example, the priority of the fire prevention is the separation distance between each power facility and the fire location (source) and the fire spread in each fire area when there are multiple ignition points or the fire spreads to a long distance from the first ignition point and the fire area is two or more. It can be determined by speed.

4 is a block diagram illustrating an embodiment of a fire response system using image recognition information on fire according to the spirit of the present invention.

The illustrated fire management system is a fire image analysis device (10) that determines whether a fire has occurred and the degree of proximity to the target object (electric power facility) to be monitored by applying image recognition technology to an image taken in a fire monitoring area. ; A fire spread rate derivation device 20 for deriving a fire spread rate based on the deviation of the flame portion of the image; A fire action device (40) that performs fire response measures for the object with a response level determined according to the fire spread rate derived for each fire (area) and the degree of proximity to the target object (electric power facility) to be monitored. ) may be included.

The fire image analysis device 10 may recognize/confirm the occurrence of a fire through an image recognition-based classification technique. That is, step S20 of FIG. 1 may be performed.

The fire spread rate deriving device 20 may perform steps S40 and S60 of FIG. 1 .

The illustrated fire spread rate derivation device 20 includes a flame/smoke classification unit 22 that separates and extracts a flame and smoke portion for an outbreak fire; A flame portion extraction unit 24 that calculates a maximum area for the extracted flame portion; It may include a fire spread rate estimating unit 28 that divides the fire scale in stages based on the flame portion and estimates the fire spread rate as a deviation of the fire area that changes over time.

The fire spread rate estimator 28 may estimate the fire spread rate through the spread area versus time based on the deviation of the flame portion extracted at regular time intervals through time monitoring. To this end, as shown, a flame portion deviation calculation unit 26 for calculating the deviation of the flame portion for each predetermined period of time may be further included.

For example, the fire spread rate estimator 28 may use the method for estimating the fire spread rate based on image recognition as shown in FIG. 2 .

That is, the fire spread rate estimator 28 estimates the fire spread rate by changing the length of the four sides of the rectangular frame including all the recognized flame parts, but at a specific time point based on area information collected at regular time intervals. The length of each side of the rectangular frame of is extracted, and the fire spread rate can be calculated from the change in the length of each side after the lapse of time.

The fire action device 40 is for performing step S80 of FIG. 1, for example, a plurality of response levels set according to the separation distance between each target facility and the fire location (source) and the fire spread speed of each fire area. Response measures can be performed according to, and actions to be performed at each response level can be written in the corresponding storage areas 41 to 43.

For example, the fire action device 40 takes immediate action on the target facility when it is determined that the fire reaches the target facility quickly according to the distance between the target facility and the fire location and the speed of the fire spread. can be instructed.

The fire image analysis device 10 may be implemented with various known technologies.

The illustrated fire image analysis device 10 includes a fire recognition unit 11 for recognizing whether or not a fire has occurred by analyzing an image captured in an area to be monitored; When a fire occurs, a fire origin estimation unit 12 for estimating the location of the fire by analyzing the image; A facility location estimating unit 14 for estimating the location of an object to be protected (power facility) by analyzing the image when a fire occurs; and a separation distance calculation unit 16 that calculates a separation distance between the target object and the location of the fire.

For example, the fire image analysis device 10 extracts distance information related to the target facility location based on a depth map for the image, detects flames and smoke of a fire based on a color image of the image, , When the target facility is recognized, the detection area of the target facility and the flame or smoke detection area overlap at a predetermined rate or more, and the depth map intensity difference between the overlapped detection areas is less than a predetermined reference value, it is judged as a proximity fire to the target facility. can

Whether or not a fire is determined by the fire recognition unit 11 is provided to the fire spread rate deriving device 20, and a process of deriving a fire spread rate according to the spirit of the present invention is initiated.

The fire source estimating unit 12, the facility location estimating unit 14, and the separation distance calculating unit 16 may be implemented by applying various known technologies, and may be implemented in the form of a block diagram of FIG. 5 to be described later. can

The main idea of the present invention is to estimate the fire spread rate based on the deviation of the flame part of the image, but the fire spread rate estimated in this way is combined with the location information of the power facility to be protected, and countermeasures (eg, step S80) ) can be used as a basis for determining In particular, it is necessary to know the separation distance from the location (region) of the fire among the location information of power facilities.

The separation distance may be obtained from various image-based analysis methods performed in the fire image analysis device 100 of FIG. 4 . The following exemplifies a separation distance estimation process between a target (power facility) and a fire that can be applied to the present invention.

5 is a block diagram for calculating the separation distance from a power facility (object) to a location where a fire occurs.

6 illustrates a fire detection screen near power facilities.

In the separation distance estimation process supported by the illustrated block diagram configuration, the electric power company designates the object to be protected from fire as a power facility, and after learning equipment such as electric poles, switches, insulators, and transformers in advance using image recognition technology, fire detection And it can be applied in a way to determine the location of the fire. With various embodiments of the proposed technology, the object to be recognized can be variously applied to objects requiring protection from fire, such as cultural assets, mechanical facilities, and structures, as well as the power facilities shown in FIG. 6 .

The configuration of FIG. 5 has a structure in which color (RGB) image analysis and gray image analysis are performed in parallel on images captured by two or more cameras to calculate the separation distance based on the depth map. It does not constitute a relationship that matches or is included in the fire image analysis device 100, but may have corresponding components that partially perform the same role as each other.

The depth map-based distance information extraction process unit 200 of FIG. 5 applies the existing depth map extraction algorithm and then defines the distance according to the brightness value of the depth map.

For example, brightness values for a corresponding object may be defined in advance in units of 50 m, such as 100 m / 150 m / 200 m, and a table as shown in Table 1 for a range of brightness values for measured values for each distance to the object may be defined.

At this time, the actual distance of the specific object may be referenced using a distance measurement sensor (radar, lidar, map information, etc.) in advance.

FIG. 7 shows an example of a detailed configuration of the detection area position coordinate calculation module 300 of FIG. 5 .

The illustrated detection area location coordinate calculation module 300 is for detecting the location of a fire area and an object (eg, power facility), and performs fire and object detection by applying, for example, a deep learning-based object detection algorithm. can do. While calculating the location coordinates of the detected area and calculating the distance of the detected area, the relationship between the actual measured distance and the depth map information brightness can be approximated.

8 illustrates the results of detecting a fire and a target object (power facility).

Depending on the implementation, the detection area location coordinate calculation module 301, the fire location detection module 302, and the target detection module 303 constituting the detection area location coordinate calculation module 300 of FIG. 6 are implemented as learning modules. It can be.

For example, each of the modules 301, 302, and 303 can train a vast amount of data images for fire and objects (power facilities) for the development of this proposed technology. That is, after calculating the coordinates and area using the final result of AI-based fire and object detection FIG. You can write and save.

FIG. 9 shows an example of a detailed configuration of the fire location final determination module 400 of FIG. 5 .

The illustrated fire occurrence location final determination module 400 is a module for determining whether a target object (power facility) is adjacent to or distant from a fire, and includes a part 404 for calculating the coordinates of a fire/object detection area, AI-based fire and object detection It can be composed of an overlap check part 401 with the area, a fire detection distance based on depth map and a target object (power facility) detection distance calculation part 402, and a final determination part 403 of the fire location.

The fire and object detection area overlap checking module 401 can check the overlap between the two areas using each coordinate of the fire and object detection area (power facility) from the table results stored in Tables 2 and 3. In other words, if the detection area of the detected target and the detection area of fire (smoke and flame) overlap by more than 80%, it is recognized as a fire close to the target.

Table 4 below shows examples of area values for overlapping of each target area and fire area.

From the table results of Table 4 above, the following results can be predicted. For example, although a fire broke out near an electric pole, transformer, and switchgear, the areas where the detection area overlaps by more than 80% are (electric pole, flame 2) and (transformer, smoke 1). In other words, it can be estimated that a fire broke out adjacent to an electric pole.

It cannot be concluded that the fire has occurred from the target object only with the fire and object detection area overlap confirmation module 401 alone. The reason is that distance information cannot be known in a 2D image.

The Depth Map-based fire detection distance and object detection distance calculation module 402 uses the Depth Map-based distance information extraction process unit 200 of FIG. 5 to refer to the depth map-based distance information for the fire area and the target area. . As for the distance information, only the distance between (firework 2, electric pole) and (smoke 1, transformer) is calculated as in the table of Table 4 above.

In the illustrated case, the proposed technology defines only when the overlapping area is 80% or more in the fire and object detection area overlap check module 401.

Table 5 below is a table illustrating distance information for fire and object detection areas showing a difference in intensity of depth map information for this distance.

The fire area final determination module 403 may derive fire proximity information for each fire area and each power facility. As the proximity information, a separation distance can be calculated as a continuous value between each fire area and each power facility, or a close or distant fire for each power facility (target) can be calculated as a binary value. The former case is advantageous in assigning an order of action according to more specific circumstances, and the latter case has the advantage of low computational burden and fast computational speed.

FIG. 10 is a flowchart illustrating an example of a fire occurrence location final determination logic performed by the fire occurrence area final determination module 403 in the latter case.

The fire occurrence area final determination module 403 may finally determine whether the fire occurred at a distance from the target object or a fire that occurred close to the target object.

That is, the results (distance and overlapping area) extracted from the AI-based fire and object detection area overlap check part 401 and the depth map-based fire detection distance and object (power facility) detection distance calculation part 402 are shown. Applied to the same logic as in 6, it is possible to determine proximity or long-distance fire to the target.

The above description is merely an example of the technical idea of the present invention, and various modifications and variations can be made to those skilled in the art without departing from the essential characteristics of the present invention. Therefore, the embodiments disclosed in the present invention are not intended to limit the technical spirit of the present invention, but to explain, and the scope of the technical spirit of the present invention is not limited by these embodiments. The protection scope of the present invention should be construed according to the following claims, and all technical ideas within the equivalent range should be construed as being included in the scope of the present invention.

10: fire image analysis device 20: fire spread rate derivation device
22: flame / smoke classification unit 24: flame portion extraction unit
26: flame partial deviation calculation unit 28: fire spread rate estimation unit
40: fire action device

Claims (15)

Recognizing the occurrence of a fire based on image recognition;
Separately extracting a flame and smoke portion for the fire that occurs, and calculating an area for the extracted flame portion;
estimating a fire spread rate as a deviation of the area of the flame portion that changes with time; and
Performing measures according to the fire spread rate to protect the target facility
A fire response method for power facilities comprising a.
According to claim 1,
The step of estimating the fire spread rate,
A fire response method for power facilities that calculates the fire spread rate through the spread area versus time based on the deviation of the flame portion extracted at regular time intervals through time monitoring.
According to claim 1,
In the step of estimating the fire spread rate,
Estimating the fire spread speed by changing the length of the four sides of the rectangular frame including all the recognized flame parts,
A fire response method for power facilities that extracts the length of each side of the rectangular frame at a specific time based on area information collected at regular time intervals and calculates the fire spread rate from the change in the length of each side after the lapse of time.
According to claim 3,
In the step of estimating the fire spread rate,
A fire response method for a power facility that presents a fire spread rate by applying a maximum value among length changes of each of the four sides of the rectangular frame.
According to claim 1,
In the step of estimating the fire spread rate,
A fire response method for power facilities that estimates a fire spread rate from a change in the circumference or diameter of a circular frame including all of the recognized flame parts.
According to claim 1,
The step of recognizing the occurrence of a fire is,
extracting distance information related to a target facility location based on a depth map of the image;
Detecting flame and smoke of a fire based on the color image of the image and recognizing a target facility; and
Determining the degree of separation of the fire from the target facility from the distance information, detection result, and recognition result
A fire response method for power facilities comprising a.
According to claim 6,
In the step of determining the degree of separation,
The detection area of the detected target facility and the fire detection area overlap at a predetermined rate or more,
When the depth map intensity difference of the detection area is less than a predetermined reference value,
A fire response method for a power facility that determines a proximity fire to the target facility.
According to claim 6,
In the step of carrying out the above measures,
According to the distance between the target facility and the fire location and the rate of fire spread, when it is determined that the fire reaches the target facility quickly, instructing immediate measures for the target facility Fire response method for a power facility.
A fire image analysis device that analyzes an image taken in a fire monitoring area to determine the degree of separation of fire from target facilities in the monitoring area;
a fire spread rate deriving device for deriving a fire spread rate based on the deviation of the flame portion of the image; and
A fire action device that performs fire response measures for the target object at a response level determined according to the fire spread rate and the degree of separation of the fire from the target facility
An image recognition-based fire response system that includes a.
According to claim 9,
The fire spread rate derivation device,
A flame/smoke classification unit for classifying a flame part and a smoke part in an image taken for a fire that has occurred;
a flame portion extraction unit for calculating a maximum area for the extracted flame portion; and
A fire spread rate estimator for estimating the fire spread rate as a deviation of the flame portion that changes over time.
Image recognition-based fire response system including a.
According to claim 10,
The fire spread rate derivation device,
Flame part deviation calculation unit for calculating the deviation of the flame part for a certain period of time
A fire response system further comprising a.
According to claim 9,
The fire spread rate derivation device,
Separately extracting a flame part and a smoke part for an outbreak fire, and calculating an area for the extracted flame part; and
A fire response system for performing a fire spread rate estimation method comprising the step of estimating a fire spread rate as a deviation of the area of the flame portion that changes over time.
According to claim 10,
The fire spread rate estimation unit,
Estimating the fire spread speed by changing the length of the four sides of the rectangular frame including all the recognized flame parts,
A fire response system that extracts the length of each side of the rectangular frame at a specific time based on area information collected at regular time intervals and calculates the fire spread rate from the change in length of each side after the lapse of time.
According to claim 9,
The fire action device,
A fire response system instructing an immediate action for the target facility when it is determined that the fire reaches the target facility quickly according to the degree of separation between the target facility and the fire location and the rate of fire spread.
According to claim 9,
The fire image analysis device,
Based on the depth map of the image, distance information related to the location of the target facility is extracted, fire flame and smoke are detected based on the color image of the image, the target facility is recognized, and the target facility is detected. A fire response system that determines that a fire is close to the target facility when the area and the detection area of flame or smoke overlap at a predetermined rate or more, and the depth map intensity difference between the overlapping detection areas is less than a predetermined reference value.

Images