KR101934700B1 - Early fire detection system, server and method using image processing and artificial intelligence based on deep learning - Google Patents

Abstract

Disclosed are a system, a server and a method for early fire detection using image processing and artificial intelligence (AI) based on deep learning which can accurately detect a fire in an early stage. The system for early fire detection using image processing and AI based on deep learning comprises at least one CCTV installed in each space of a building and an early fire detection server. At least one CCTV performs image processing on a photographed image to determine whether a fire or smoke is detected, and transmits the photographed image to the early fire detection server when a fire or smoke is detected in the photographed image. The early fire detection server includes: a server communication unit to receive the photographed image from at least one CCTV; a fire determination unit to use AI based on deep learning on the photographed image to re-determine whether a fire occurs; and a fire notification unit to transmit a fire notification to a manager terminal via the server communication unit when the fire determination unit determines that a fire has occurred.

Description

FIELD OF THE INVENTION [0001] The present invention relates to an early fire detection system, server and method using AI based on image processing and deep running,

The present invention relates to an early fire detection system, a server and a method using AI based on image processing and deep learning. More particularly, the present invention detects a fire by processing a CCTV image and detects a fire by using a deep learning based artificial intelligence A server, and a method for detecting an early fire using an AI based on image processing and deep learning that can accurately detect a fire early.

Early fire detection and evacuation route guidance systems are very important in building management. A system for early detection of flame or smoke is needed to minimize the casualties of a fire.

A conventional fire detection system often uses a differential type sensor that senses heat due to thermal expansion of air. However, when the temperature around the sensor is high, it is vulnerable to initial fire detection since the fire has already diffused to some extent .

In addition, although smoke detectors such as photoelectric detectors or ionizing detectors are used, it is difficult to be generalized due to economical problems, and smoke may not be detected due to ventilation or diffusion of outside air, There is a problem that it is malfunction due to dew and the like.

In the case of the flame detection sensor, since the ultraviolet ray detection method is used, the ultraviolet ray is absorbed by the other floating matters and the sensitivity may be lowered.

Accordingly, there is a need for a fire detection system capable of detecting a fire early while reducing malfunctions by solving the problems of the related arts.

Disclosure of Invention Technical Problem [8] Accordingly, the present invention has been made keeping in mind the above problems occurring in the prior art, and an object of the present invention is to provide an image processing apparatus and method capable of detecting a fire first by processing a CCTV image and detecting fire secondarily using a deep learning- An object of the present invention is to provide an early fire detection system, server and method using AI based on deep running.

In order to achieve the above object, an early fire detection system using AI based on image processing and deep learning according to an embodiment of the present invention includes at least one CCTV and an early fire detection server installed in each space of a building The at least one CCTV processes the photographed image to determine whether fire or smoke is detected. If fire or smoke is detected in the photographed image, the photographed image is transmitted to the early fire detection server Wherein the early fire detection server comprises: a server communication unit for receiving the photographed image from the at least one CCTV; a fire determination unit for re-determining whether or not the photographed image is fire based on the deep learning based artificial intelligence; When the fire judging unit judges that a fire has occurred, a fire alarm is transmitted to the administrator terminal through the server communication unit It characterized in that it comprises a.

In order to achieve the above object, an early fire detection server using AI based on image processing and deep learning according to an embodiment of the present invention may be configured to receive an image photographed from at least one CCTV installed in each space of a building The server communication unit processes the photographed image to firstly determine whether fire or smoke has been detected. If it is determined that fire or smoke is detected as a result of the primary determination, the photographed image is subjected to deep learning based artificial intelligence And a fire notification unit that transmits a fire notification to the administrator terminal through the server communication unit when the fire determination unit determines that a fire has occurred.

Preferably, the early fire detection server may further include an evacuation route guidance unit for guiding the evacuation route according to a fire occurrence location and a congestion rate analyzed from the image photographed by the at least one CCTV.

Preferably, the evacuation route guidance unit guides the evacuation route when receiving the fire situation confirmation information from the administrator terminal that has received the fire notification, It is possible to feedback to the deep learning based artificial intelligence by recognizing that the fire determination by the fire determination unit is wrong.

Preferably, the evacuation route guide unit may automatically guide the evacuation route when the fire status information or the fire false detection information is not received within a predetermined time from the administrator terminal that received the fire notification have.

In order to achieve the above object, an early fire detection method using AI based on image processing and deep learning according to an embodiment of the present invention is characterized in that a server communication unit is provided for detecting an image captured from at least one CCTV installed in each space of a building The method of claim 1, further comprising the steps of: determining whether a fire or smoke is detected by processing a photographed image by the fire determination unit; and if the fire determination unit determines that fire or smoke is detected as a result of the primary determination, Determining whether a fire is occurring by using the artificial intelligence based on the deep learning based on the fire detection result of the fire detection unit; transmitting a fire notification to the administrator terminal through the server communication unit when the fire determination unit determines that the fire has occurred; Receives the fire status confirmation information from the manager terminal that has received the notification, or transmits the fire notification And guiding the evacuation route according to the fire occurrence location and the congestion rate analyzed from the image photographed by the at least one CCTV when the reply from the administrator terminal is not received within a predetermined time .

Preferably, when the fire detection information is received from the administrator terminal that received the fire notification, the fire determination unit recognizes that the fire determination by the fire determination unit is incorrect, and feeds back to the deep learning based artificial intelligence .

According to the present invention, an early fire detection system, server and method using AI based on image processing and deep learning according to the present invention can detect a fire by processing a CCTV image and detect a fire by using a deep learning based artificial intelligence By detecting double fire by detecting, it is possible to minimize malfunction and to detect fire early and accurately.

According to the present invention, an early fire detection system, a server, and a method using AI based on image processing and deep learning can solve the malfunction problem of existing smoke detectors and flame detectors by using image processing technology, It can be a great help to detect and evolve fire.

Since the early fire detection system, server and method using AI based on image processing and deep learning according to the present invention use deep learning based artificial intelligence, the more accurate the learning, the more accurate the recognition and the reliability Can be increased.

The above-described early fire detection system, server and method using AI based on image processing and deep learning according to the present invention can advantageously use CCTV installed in a building as it is, It has the effect of minimizing the blind spot than existing fire detectors.

The early fire detection system, server and method using AI based on the image processing and the deep learning based on the present invention can guide the evacuation route in real time by reflecting the fire occurrence location and the congestion rate of the evacuation route It is effective.

1 is a diagram showing an early fire detection system using AI based on image processing and deep learning according to an embodiment of the present invention.
2 is a diagram showing an early fire detection server using AI based on image processing and deep learning according to an embodiment of the present invention.
3 is a flow chart illustrating an early fire detection method using AI based on image processing and deep learning according to an embodiment of the present invention.
FIGS. 4A and 4B are views illustrating a method of processing an image photographed in CCTV according to an early fire detection method using AI based on image processing and deep learning based on an embodiment of the present invention to determine whether smoke or smoke is detected Fig.
5A and 5B are views for explaining a method of operating the evacuation route guide unit according to an embodiment of the present invention.

Hereinafter, embodiments according to the present invention will be described with reference to the accompanying drawings. It should be noted that, in adding reference numerals to the constituent elements of the drawings, the same constituent elements are denoted by the same reference symbols as possible even if they are shown in different drawings. In the following description of the embodiments of the present invention, a detailed description of known functions and configurations incorporated herein will be omitted when it may make the difference that the embodiments of the present invention are not conclusive. In addition, embodiments of the present invention will be described below, but the technical idea of the present invention is not limited thereto and can be variously modified by those skilled in the art.

Throughout the specification, when a part is referred to as being "connected" to another part, it includes not only "directly connected" but also "indirectly connected" . Throughout the specification, when an element is referred to as "comprising ", it means that it can include other elements as well, without excluding other elements unless specifically stated otherwise.

1 is a diagram showing an early fire detection system using AI based on image processing and deep learning according to an embodiment of the present invention. Referring to FIG. 1, an early fire detection system 1000 using AI based on image processing and deep learning according to an embodiment of the present invention includes at least one CCTV 100_1 to 100_6 installed in each space of a building P, And an early fire detection server 300.

At least one CCTV 100_1 to 100_6 may be installed in each space P1, P2, P3 of the building P. [ FIG. 1 is a simplified illustration of a three-story building P and at least one CCTV 100_1 to 100_6 for convenience. At least one CCTV 100_1 to 100_6 may be installed in various spaces inside the building P, The number can also be varied. At least one of the CCTVs 100_1 to 100_6 may communicate with the early fire detection server 300 through a network such as a wired / wireless communication network 500.

At least one of the CCTVs 100_1 to 100_6 may image the photographed image to determine whether fire or smoke is detected. Such a process may be performed through an image processing unit (not shown) provided in at least one of the CCTVs 100_1 to 100_6. As an example, at least one of the CCTVs 100_1 to 100_6 can analyze a photographed image in real time using OpenCV (Open Source Computer Vision) which is a library for real-time image processing.

When at least one of the flame and smoke is detected through the real-time image processing on the photographed image, at least one of the CCTVs 100_1 to 100_6 transmits the photographed image to the early fire detection server 300 using the wired / Lt; / RTI > For example, at least one of the CCTVs 100_1 to 100_6 can transmit a real-time image to the early fire detection server 300 for each predetermined frame when at least one of flame and smoke is detected in the photographed image.

The early fire detection server 300 can receive images photographed from at least one CCTV 100_1 to 100_6. The fact that at least one of the CCTVs 100_1 to 100_6 has transmitted the real-time image to the early-fire detection server 300 means that a flame or smoke is primarily detected in the real-time image. The early fire detection server 300 can re-determine whether or not the image transmitted from at least one of the CCTVs 100_1 to 100_6 is fire based on the deep learning based artificial intelligence.

The early fire detection server 300 may transmit a fire notification to the administrator terminal 700 when it is determined that a fire has occurred as a result of re-determining the fire using the deep learning based artificial intelligence. The early fire detection server 300 can guide the evacuation route when receiving the fire situation confirmation information from the administrator terminal 700 which received the fire notification. Here, the fire status confirmation information may be the information that the administrator confirms if the fire occurrence is confirmed as a result of directly confirming after receiving the fire notification through the administrator terminal 700, and transmits the information to the early fire detection server 300. The early fire detection server 300 can guide the escape route according to the fire occurrence location and the congestion rate analyzed from the images photographed by at least one CCTV 100_1 to 100_6.

When the early fire detection server 300 receives the fire alarm information from the administrator terminal 700 receiving the fire alarm, the early fire alarm server 300 recognizes that the fire alarm is incorrect and can feedback the deep learning based artificial intelligence. Here, the fire-false information may be information that the administrator confirms if the fire is not generated as a result of directly confirming the fire notification received through the administrator terminal 700, and transmits the information to the early fire detection server 300.

The early fire detection server 300 can automatically guide the evacuation route if it fails to receive the fire status confirmation information or the fire false detection information within a predetermined time from the administrator terminal 700 that received the fire notification. In other words, if the administrator is absent or fails to confirm the fire notification, the evacuation route guidance can not be delayed. Therefore, it is determined that the fire is caused by trusting the result of the artificial intelligence based on the deep learning. .

Although the early fire detection server 300 is illustrated as being located outside the building P in FIG. 1, the early fire detection server 300 may be located inside the building P. FIG. The detailed configuration and operation method of the early fire detection server 300 will be described in further detail with reference to FIGS. 2 to 5B, and a detailed description thereof will be omitted.

The wired / wireless communication network 500 is configured as a communication network through which at least one CCTV 100_1 to 100_6, an early fire detection server 300 and an administrator terminal 700 can communicate with each other without specifically selecting a communication mode And can be composed of various communication networks such as a personal area network (PAN), a local area network (LAN), a metropolitan area network (MAN), a wide area network (WAN) . Also, the wired / wireless communication network 500 may be a known World Wide Web (WWW), and may use a wireless transmission technology used for short-distance communication such as infrared (IrDA) or Bluetooth It is possible. The wired / wireless communication network 500 is widely known to those of ordinary skill in the art and will not be described in detail.

The administrator terminal 700 is a terminal capable of receiving a fire notification from the early fire detection server 300 and may be various types of terminals capable of data communication by wire or wireless such as a user computer 700_1 or a mobile terminal 700_2, Equipment, and may be variously changed within a range known to those of ordinary skill in the art. The administrator terminal 700 may be a user computer 700_1 installed in a central control room or the like and may be a mobile terminal 700_2 directly owned by an administrator. The administrator can receive the fire notification through the administrator terminal 700. If it is determined that the actual fire situation is detected, the manager can transmit the fire situation confirmation information to the early fire detection server 300. [ If the manager receives the fire notification through the administrator terminal 700 but it is determined that it is not a fire, the manager can transmit the fire false detection information to the early fire detection server 300.

2 is a diagram showing an early fire detection server using AI based on image processing and deep learning according to an embodiment of the present invention. Referring to FIG. 2, an early fire detection server 300 using AI based on image processing and deep learning according to an embodiment of the present invention includes a server communication unit 310, a fire determination unit 330, and a fire notification unit 350, and may further include an evacuation route guide unit 370 and a database unit 390 according to an embodiment.

The server communication unit 310 can receive images photographed from at least one CCTV 100_1 to 100_6 installed in each space of the building P. [ The server communication unit 310 may transmit the fire notification received from the fire notification unit 350 to the administrator terminal 700 and may send fire status confirmation information and fire false detection information transmitted from the administrator terminal 700 to fire (330) and the evacuation route guide unit (370). The server communication unit 310 can communicate with at least one of the CCTVs 100_1 to 100_6 and the administrator terminal 700 through the wired / wireless communication network 500 and can be implemented as a communication module supporting various wired / wireless communication methods, But may be varied within the scope known to the skilled artisan.

The fire determination unit 330 can receive a real-time image in which flame or smoke is firstly sensed from at least one CCTV 100_1 to 100_6 and can deep-run the image transmitted from at least one CCTV 100_1 to 100_6, Based artificial intelligence can be used to judge whether or not the fire has occurred.

According to the embodiment, the fire determination unit 330 processes the photographed image to firstly determine whether fire or smoke is sensed. If it is determined that the fire or smoke is sensed as a result of the primary determination, Based on the deep learning based artificial intelligence, the captured images can be judged in the second place. In this case, at least one of the CCTVs 100_1 to 100_6 may transmit an image photographed at a predetermined frame interval to the early fire detection server 300 without performing real-time image processing for detecting flame or smoke in the photographed image . That is, according to the embodiment, at least one CCTV 100_1 to 100_6 may perform a primary judgment for detecting a flame or smoke in the photographed image, or may be performed by the fire determination unit 330 of the early fire detection server 300, . ≪ / RTI >

The deep learning based artificial intelligence performed by the fire determination unit 330 may operate in the following manner.

The image information captured by at least one CCTV 100_1 to 100_6 installed in the building P may be transmitted to the early fire detection server 300 and the early fire detection server 300 may transmit the received CCTV image to the database unit 390). In this manner, the normal CCTV image stored in the database unit 390 can be a standard data value. The normal CCTV images may be classified and stored according to at least one of the CCTVs 100_1 to 100_6.

When a fire or smoke is detected in an image photographed by at least one of the CCTVs 100_1 to 100_6 according to the result of the first real-time image processing, the fire determination unit 330 judges whether or not the deep learning based artificial intelligence It is possible to judge once again whether the object of the photographed image is certainly fire or smoke. For example, TensorFlow may be used as the deep learning based artificial intelligence engine, and Inception v3 model may be used. In order to achieve the goal of early fire detection, we can use map learning, which is a method of learning a machine learning model using training data that knows the correct answer. The structure, weight, parameter, data set, etc. of the artificial intelligence model are variously set . Images transmitted from at least one CCTV 100_1 to 100_6 can be stored in the database unit 390 for each frame and learning can be performed by various source codes.

The fire determination unit 330 compares the CCTV image primarily determined to have detected fire or smoke with the normal CCTV image stored in the database unit 390. If the degree of similarity with the CCTV image is significantly different It can be judged that it is a fire. For example, if the degree of similarity between the CCTV image and the normal CCTV image is 90% or more, it is determined that the primary judgment result is malfunctioning, and the feedback is given and the fire notification is not generated have. If the degree of similarity between the CCTV image and the normal CCTV image is less than 90%, the fire escape system can be operated. That is, the fire determination unit 330 can set a predetermined similarity reference value. The fire determination unit 330 compares the similarity of the CCTV image, which is determined to be primarily fire or smoke, It is possible to re-determine whether the result is correct or not. The predetermined similarity reference value may be variously changed according to each building P, each zone P1, P2, P3, and each CCTV 100_1 to 100_6.

Since the fire determination unit 330 uses the deep learning based artificial intelligence, the normal CCTV images of at least one of the CCTVs 100_1 to 100_6 are continuously accumulated in the database unit 390 to provide a large amount of training data, The accuracy of fire judgment is increased. Accordingly, as the deep learning-based artificial intelligence of the fire determination unit 330 learns more, the accuracy and reliability of recognition of the fire situation of the building P becomes higher. Therefore, Early fire detection and evacuation route guidance may be possible.

Meanwhile, according to the embodiment, the deep learning based artificial intelligence performed by the fire determination unit 330 may operate as a method of learning a flame or smoke image. That is, the database unit 390 can store a large amount of flame images and smoke images, and the deep learning-based artificial intelligence of the fire determination unit 330 can be stored in the database unit 390 through flame images and smoke images, The ability to recognize smoke can be improved. The fire determination unit 330 analyzes the CCTV image primarily determined to have detected fire or smoke according to the deep learning-based artificial intelligence, and determines whether or not there is a flame or smoke in the corresponding CCTV image have.

The fire notification unit 350 may transmit a fire notification to the administrator terminal 700 through the server communication unit 310 when the fire determination unit 330 determines that a fire has occurred. The fire notification can be transmitted to the administrator terminal 700 in the form of a text message or a push message, and the location of the suspected fire zone, the image of the CCTV in which a fire is detected, etc. can be transmitted together with the manager terminal 700 .

The administrator terminal 700 can notify the manager of the reception of the fire notification through a sound of a notification. When an administrator receives a fire notification, the administrator can confirm the fire situation again by eye to see if the fire situation is right. If the fire situation is right, the fire escape route guidance system is activated so that people in the building can safely escape through the shortest route Can be done.

The evacuation route guide unit 370 can guide the evacuation route according to the fire occurrence location and the congestion rate analyzed from the images photographed by at least one CCTV 100_1 to 100_6. The congestion rate can be calculated according to whether the crowd is crowded by analyzing the CCTV image.

The evacuation route guidance unit 370 can guide the evacuation route when receiving the fire situation confirmation information from the administrator terminal 700 having received the fire notification.

The evacuation route guide unit 700 can automatically guide the evacuation route when it fails to receive the fire status confirmation information or the fire false detection information within a predetermined time from the administrator terminal 700 having received the fire notification. That is, if the notification can not be confirmed due to the absence of an administrator, the evacuation route guidance unit 370 can automatically guide the evacuation route after a specific time.

When the fire determination unit 330 receives the fire detection information from the administrator terminal 700 having received the fire notification, the fire determination unit 330 recognizes that the fire determination by the fire determination unit 330 is wrong, Feedback can be made. The deep learning based artificial intelligence of the fire determination unit 330 recognizes that the CCTV image in which the fire false detection information is received is not a fire, and can newly learn it and reflect it in the next fire judgment.

According to the embodiment, when the congestion degree of the suspected fire zone is increased or the moving speed of the people is higher than the normal speed in the suspected fire zone in a state in which there is no response from the administrator terminal 700 receiving the fire alert, May determine that a fire has occurred and guide the evacuation route through the evacuation route guide unit 370. [ That is, the fire determination unit 330 analyzes the images photographed in the CCTV of the fire suspected area, and when a phenomenon that the number of people in the relevant area increases to increase the congestion degree or the moving speed of people is detected, It can be guided to the evacuation route. That is, the fire determination unit 330 according to an embodiment of the present invention can guide the evacuation route by analyzing the congestion degree of the suspected fire zone analyzed through the CCTV and the moving speed of the people.

The database unit 390 can store photographed images and the like transmitted from at least one CCTV 150_1 to 150_6 and can also store information about user terminals held by people in the building P for evacuation route guidance.

3 is a flow chart illustrating an early fire detection method using AI based on image processing and deep learning according to an embodiment of the present invention. The method 2000 for detecting an early fire using AI based on image processing and deep learning in accordance with an embodiment of the present invention shown in FIG. 3 may be applied to an embodiment of the present invention shown in FIGS. 1 to 2 And an early fire detection server 300 using an AI based on deep processing.

Referring to FIG. 3, a method 2000 for detecting an early fire using AI based on image processing and deep learning based on an embodiment of the present invention includes a server communication unit 310, And receiving a photographed image from the CCTVs 100_1 to 100_6 in step S21. According to the embodiment, at least one of the CCTVs 100_1 to 100_6 may determine the photographed image as a first order, and may transmit the photographed image to the early fire detection server 300 only when fire or smoke is detected. In this case, step S22 performed by the early fire detection server 300 may be omitted.

Next, an early fire detection method (2000) using AI based on image processing and deep learning according to an embodiment of the present invention detects an early fire when the fire determination unit (330) processes the captured image and detects fire or smoke (Step S22). As described above, the fire determination unit 330 can detect fire or smoke according to the OpenCV-based real-time image processing method.

Next, an early fire detection method 2000 using AI based on image processing and deep learning according to an embodiment of the present invention determines whether fire or smoke is detected as a result of the primary determination (S23) of judging whether or not to fire by using the artificial intelligence based on the deep learning based on the photographed image.

If no fire or smoke is detected as a result of the primary determination, the process returns to step S21 to newly transmit an image of at least one CCTV 100_1 to 100_6 and perform step S22 again.

Next, an early fire detection method (2000) using AI based on image processing and deep learning according to an embodiment of the present invention is performed by the server communication unit 310 when the fire determination unit 330 determines that a fire has occurred, And sending a fire notification to the administrator terminal 700 through the firewall (S24).

If it is determined that the fire or smoke is detected as a result of the primary determination but the fire determination unit 330 determines that the fire or smoke is not detected as a result of the secondary determination, the fire determination unit 330 may return to step S21, can do.

Next, an early fire detection method 2000 using AI based on image processing and deep learning according to an embodiment of the present invention receives fire status confirmation information from the administrator terminal 700 that has received a fire notification, If the escape route guidance unit 370 fails to receive a response within a predetermined time from the administrator terminal 700 having received the notification, the congestion rate analyzed from the fire location and the images photographed by at least one CCTV 100_1 to 100_6 (S25) of guiding the evacuation route in accordance with the guidance of the evacuation route.

When the fire determination unit 330 receives the fire detection information from the administrator terminal 700 having received the fire notification, the fire determination unit 330 recognizes that the fire determination by the fire determination unit 330 is wrong, Feedback can be made.

FIGS. 4A and 4B are views illustrating a method of processing an image photographed in CCTV according to an early fire detection method using AI based on image processing and deep learning based on an embodiment of the present invention to determine whether smoke or smoke is detected Fig.

Referring to FIG. 4A, it is determined that a portion processed with a red color box is detected as a flame by the fire determination unit 330. Referring to FIG. 4B, a portion processed with a red color box is a portion where smoke is detected by the fire determination unit 330. As shown in FIGS. 4A and 4, the fire determination unit 330 can detect fire (smoke) and smoke in real time through image processing using OpenCV or the like. As an example, the OpenCV can use an AI learning model called Caffe.

5A and 5B are views for explaining a method of operating the evacuation route guide unit according to an embodiment of the present invention. FIGS. 5A and 5B are schematic diagrams of a building, in which R1 to R10 represent spaces such as offices, and S1 to S4 represent steps that can be escaped when a fire occurs.

The evacuation route guide unit 370 can guide the evacuation route according to the fire occurrence location and the congestion rate analyzed from the images photographed by at least one CCTV 100_1 to 100_6. The evacuation route guide unit 370 can analyze the congestion rate by measuring the degree of population density of each evacuation route using OpenCV. In addition, the evacuation route guide unit 370 compares the images photographed by at least one of the CCTVs 100_1 to 100_6 with the images photographed when there is no fire and no population movement, and the congestion rate is calculated It is possible. For example, when a fire occurs and a large number of people gather in the exit area of the corresponding area, the degree of similarity with the previously stored image information is considerably reduced, so that it can be estimated that a large number of people are gathered. On the other hand, when there are not many people in the exit area of the area, it can be assumed that the population is not crowded because the similarity with the image information stored in the area is not greatly different. This can lead to a smooth evacuation where the congestion rate is not high.

Referring to FIG. 5A, when a fire occurs in the S1 step, the evacuation route guide unit 370 moves the evacuation route to S2 and S3 steps located at the shortest distance to R1, R2, R7, R8, R9, And guide the evacuation route to S4, to the personnel of R3, R4, R5, R6.

Next, referring to FIG. 5B, it can be determined that the congestion rate of the corresponding zone is increased through the images taken by the CCTV installed in the vicinity of the S2 step and the S3 step when the escape is made and the personnel are gathered in the S2 step and the S3 step . Accordingly, the evacuation route guide unit 370 evacuates the personnel of R1, R2, R7, R8, R9, and R10 to the S4 step rather than the S2 step and the S3 step in consideration of the increased congestion rate of the S2 step and S3 step. . This congestion rate can be calculated in real time for each zone in which at least one CCTV 100_1 to 100_6 is installed in order to calculate an optimal evacuation route in case of a fire.

The evacuation route guide unit 370 can guide the evacuation route based on the shortest path based on the fire occurrence location and the congestion rate analyzed from the images photographed by at least one CCTV 100_1 to 100_6. For example, the shortest path-based evacuation route guidance method may use a Dijkstra algorithm or an A * algorithm. The weight of each algorithm may be the distance from the CCTV installed in each zone to the evacuation route.

According to the embodiment, the evacuation route guide unit 370 can calculate the evacuation route from areas where it is difficult to evacuate, such as a high-rise building or a location far from the exit. The evacuation route guide unit 370 sets a route from the current location to the nearest exit to the evacuation route. If the fire has proceeded from the current position to the exit, the emergency exit is excluded and the route to the emergency site or the exit If the congestion rate is high, the exits can be excluded. The evacuation route guide unit 370 can set the evacuation route guiding sequence in consideration of the distance from the fire occurrence location and the like.

According to the embodiment, the evacuation route guide unit 370 can guide the evacuation route to the finally retrieved emergency exit, increase the congestion rate for the exit of the other layers connected to the escape route guided to the evacuation route, It is possible to prevent the rolling phenomenon.

According to the embodiment, the evacuation route guide unit 370 can guide the evacuation route through a broadcasting speaker installed in the building P, an escape route guide lamp, or the like. Also, the evacuation route guide unit 370 may guide the evacuation route by using a circuit or the like capable of guiding the exit through the fluorescent lamp, as in a general emergency exit guide lamp. Also, according to the embodiment, the evacuation route guide unit 370 may transmit a message regarding the evacuation route to the user terminal of the people in the building P.

While the present invention has been described with reference to exemplary embodiments, it is to be understood that the invention is not limited to the disclosed exemplary embodiments, but, on the contrary, is intended to cover various modifications and equivalent arrangements included within the spirit and scope of the appended claims. The above-described embodiments can be variously combined. The true scope of the present invention should be determined by the technical idea of the appended claims.

100_1 ~ 100_6: CCTV 300: early fire detection server
310: server communication unit 330:
350: fire notification unit 370: evacuation route guide unit
390: Database part 500: Wired / wireless communication network
700: administrator terminal

Claims (10)

An early fire detection system comprising at least one CCTV and early fire detection server installed in each space of a building,
Wherein the at least one CCTV processes the photographed image to determine whether fire or smoke is detected and transmits the photographed image to the early fire detection server when fire or smoke is detected in the photographed image,
The early fire detection server comprises:
A server communication unit receiving the photographed image from the at least one CCTV;
A fire determination unit for re-determining whether or not the photographed image is fire using a deep learning-based artificial intelligence;
A fire notification unit for transmitting a fire notification to the administrator terminal through the server communication unit when the fire determination unit determines that a fire has occurred; And
And a evacuation route guide unit for guiding a evacuation route according to a fire occurrence location and a congestion rate analyzed from an image photographed by the at least one CCTV,
The evacuation route guide unit,
And analyzing the images photographed by the at least one CCTV to calculate the congestion rate in real time by measuring the density of populations in each of the zones in which the at least one CCTV is installed. Early Fire Detection System. delete [2] The method according to claim 1,
When the fire status confirmation information is received from the administrator terminal that received the fire notification,
The fire determination unit
Wherein the fire detection information is fed back to the deep learning based artificial intelligence by recognizing that the fire determination by the fire determination unit is wrong when the fire alarm detection information is received from the administrator terminal having received the fire notification, Based fire detection system using AI. 4. The method of claim 3, wherein the evacuation route guide comprises:
When the fire status notification information or the fire false detection information is not received within a predetermined time from the administrator terminal having received the fire notification, the escape route is automatically guided. Early fire detection system using. A server communication unit for receiving images photographed from at least one CCTV installed in each space of a building;
And a controller for determining whether fire or smoke is sensed by performing image processing on the photographed image, and if it is determined that fire or smoke is detected as a result of the primary determination, A fire judging unit for judging the fire in a second order;
A fire notification unit for transmitting a fire notification to the administrator terminal through the server communication unit when the fire determination unit determines that a fire has occurred; And
And a evacuation route guide unit for guiding a evacuation route according to a fire occurrence location and a congestion rate analyzed from an image photographed by the at least one CCTV,
The evacuation route guide unit,
And analyzing the images photographed by the at least one CCTV to calculate the congestion rate in real time by measuring the density of populations in each of the zones in which the at least one CCTV is installed. Early fire detection server used. delete 6. The navigation system according to claim 5,
When the fire status confirmation information is received from the administrator terminal that received the fire notification,
The fire determination unit
Wherein the fire detection information is fed back to the deep learning based artificial intelligence by recognizing that the fire determination by the fire determination unit is wrong when the fire alarm detection information is received from the administrator terminal having received the fire notification, Based fire detection server using AI. 8. The navigation system according to claim 7,
When the fire status notification information or the fire false detection information is not received within a predetermined time from the administrator terminal having received the fire notification, the escape route is automatically guided. Fire detection server using. A server communication unit receiving an image photographed from at least one CCTV installed in each space of a building;
A step of firstly determining whether fire or smoke is detected by processing a photographed image by a fire judging unit;
If the fire determination unit determines that fire or smoke has been detected as a result of the primary determination, determining the fire based on the deep learning based artificial intelligence by using the artificial intelligence;
Transmitting a fire notification to the administrator terminal through the server communication unit when the fire determination unit determines that a fire has occurred; And
Wherein when the fire status notification information is received from the administrator terminal having received the fire notification or if the fire alarm notification is not received within a predetermined time from the administrator terminal having received the fire notification, And guiding the evacuation route according to the congestion rate analyzed from the image photographed by the imaging device,
The evacuation route guide unit,
And analyzing the images photographed by the at least one CCTV to calculate the congestion rate in real time by measuring the density of populations in each of the zones in which the at least one CCTV is installed. Detection of early fire using. 10. The method according to claim 9,
Wherein the fire detection information is fed back to the deep learning based artificial intelligence by recognizing that the fire determination by the fire determination unit is wrong when the fire alarm detection information is received from the administrator terminal having received the fire notification, Based fire detection method using AI.

Images