An integrated fire detection system utilizing AI-based video analysis and flame detection sensors.

Abstract

To provide an integrated fire detection system that uses AI-based video image analysis and a flame detection sensor.SOLUTION: The present invention relates to an integrated fire detection system using AI-based video image analysis and a flame detection sensor. The integrated fire detection system comprises: a sensing value acquisition unit that acquires a flame sensing value in a monitoring region from the flame detection sensor; a video acquisition unit that acquires video capturing the monitoring region from a camera; a video analyzing unit that detects a fire by recognizing a flame and smoke by analyzing the video upon a fire detection via an artificial intelligence (AI), and searching for a fire point in the video; and a control unit that outputs the video of the monitoring region in real time through a screen, but generates a primary alarm in a preset manner upon a flame detection by the flame detection sensor, and generates a secondary alarm containing the fire point in a preset manner upon a flame detection via video analysis. According to the present invention, the efficiency and accuracy of fire detection can be increased by utilizing both of AI-based video image analysis and flame detection results.SELECTED DRAWING: Figure 1

Description

Technical Field

[0001] The present invention relates to an integrated fire detection system using AI-based video analysis and a flame detection sensor. More specifically, the present invention relates to an integrated fire detection system that can improve the accuracy of fire detection using video analysis on an artificial intelligence basis and a flame detection sensor.

Background Art

[0002] Traditional flame detection sensors detect the amount of light or temperature change of a flame to notify a fire.

[0003] By the way, in the case of a flame detection sensor, there is a disadvantage in that infrared rays are absorbed by smoke and other floating substances, resulting in a decrease in sensitivity, and there is a high possibility of false detection such as reacting erroneously to light that is not an actual fire.

[0004] Therefore, recently, the development of a technology for analyzing video data from CCTVs and cameras based on AI to determine the situation of a fire has been actively carried out. When such AI-based video analysis techniques are utilized, the malfunction of a fire monitoring device can be reduced more than when simply using a flame detection sensor.

[0005] Therefore, a technology that can accurately detect a fire by combining AI-based video analysis technology and a flame detection sensor function is required.

[0006] The technology that is the background of the present invention is disclosed in Korean Registered Patent No. 10-2020-0041849 (published on April 22, 2020).

Summary of the Invention

Problems to be Solved by the Invention

[0007] An object of the present invention is to provide an integrated fire detection system using AI-based video analysis and a flame detection sensor that can improve the efficiency and accuracy of fire detection using AI-based video analysis and flame detection results.

Means for Solving the Problems

[0008] The integrated fire detection system according to the present invention includes: a sensing value acquisition unit that acquires flame sensing values ​​within a monitoring area from a flame detection sensor; an image acquisition unit that acquires images of the monitoring area taken by a camera; an image analysis unit that, upon flame detection, analyzes the images through artificial intelligence to recognize flames and smoke, detects a fire, and identifies the fire location in the images; and a control unit that outputs images of the monitoring area in real time through a screen, generates a primary alarm according to a set method when flames are detected by the flame detection sensor, and generates a secondary alarm including the fire location according to a set method when a fire is detected through image analysis.

[0009] Furthermore, the control unit can flash the entire screen on which the video is being output when the primary alarm occurs, and flash the screen when the secondary alarm occurs, while also displaying the detection point as a box within the video being output.

[0010] Furthermore, the control unit can map video from multiple monitoring areas to multiple split screen areas arranged in a matrix and display them simultaneously. When an alarm occurs, it can output the primary or secondary alarm on the split screen area corresponding to the monitoring area in a configurable manner.

[0011] Furthermore, the integrated fire detection system may further include a prediction unit that applies the number of human objects analyzed from the video, the motion values ​​of the human objects, and the distance between the detected location of the human objects and the fire location to a previously trained prediction model to predict the probability of a dangerous situation occurring.

[0012] Furthermore, the video analysis unit can analyze the video when a fire is detected, detect at least one object, classify the human object, and extract the number of human objects, their detection location, and their motion values.

[0013] Furthermore, the control unit may compare the predicted probability of a dangerous situation occurring with a critical value, and if it is equal to or greater than the critical value, it may determine that a dangerous situation has occurred and generate a dangerous situation alarm.

[0014] Furthermore, the integrated fire detection system further includes a critical value determination unit that determines the critical value by applying the number of human objects analyzed from the video, the motion values ​​of the human objects, and the distance between the detection position of the human objects and the fire location to a previously learned critical value determination model, and the control unit can determine the presence or absence of the dangerous situation based on the critical value determined by the critical value determination model.

[0015] Furthermore, the critical value determination model is trained so that the critical value decreases as the number of human objects increases, the movement value of the human objects decreases, or the distance between them decreases.

[0016] Furthermore, the prediction unit can predict the probability of the dangerous situation occurring by further applying the size of the fire and the speed at which the fire spreads, which have been analyzed from the video, to the prediction model.

[0017] The integrated fire detection method performed by the integrated fire detection system according to the present invention includes the steps of: acquiring images of a monitoring area through a camera and outputting them in real time through a screen; detecting flames in the monitoring area using a flame detection sensor; generating a primary alarm in a pre-configured manner when flames are detected by the flame detection sensor; analyzing the images of the monitoring area through artificial intelligence to recognize flames and smoke and detect a fire, and detecting the location of the fire in the images; and generating a secondary alarm including the location of the fire in a pre-configured manner when a fire is detected through the image analysis.

[0018] Furthermore, in the stage of generating the primary alarm, the entire screen on which the video is being output will flash, and in the stage of generating the secondary alarm, the screen will flash, and the detection point will be displayed as a box within the video being output.

[0019] In addition, the integrated fire detection method may further include: applying the number of human objects analyzed from the video, the movement value of the human objects, and the distance between the detection position of the human objects and the fire location to a pre-trained prediction model to predict the probability of occurrence of a dangerous situation; and comparing the predicted probability of occurrence of the dangerous situation with a threshold value, and if it is greater than or equal to the threshold value, determining that it is a dangerous situation and generating a dangerous situation alarm.

[0020] In addition, the integrated fire detection method may further include determining the threshold value by applying the number of human objects analyzed from the video, the movement value of the human objects, and the distance between the detection position of the human objects and the fire location to a pre-trained threshold value determination model.

Advantages of the Invention

[0021] According to the present invention, both AI-based video analysis and flame detection results can be utilized to improve the efficiency and accuracy of fire detection.

Brief Description of the Drawings

[0022] [Figure 1] It is a drawing showing the configuration of an integrated fire detection system according to an embodiment of the present invention. [Figure 2] It is a drawing exemplarily showing a camera equipped with a flame detection sensor. [Figure 3] It is a drawing for explaining an integrated fire detection method according to an embodiment of the present invention.

Modes for Carrying Out the Invention

[0023] Hereinafter, embodiments of the present invention will be described in detail so that those skilled in the art can easily implement them with reference to the attached drawings. However, the present invention can be embodied in various different forms and is not limited to the embodiments described herein. And in order to clearly explain the present invention in the drawings, parts not related to the explanation are omitted, and similar parts throughout the specification are denoted by similar reference numerals.

[0024] Throughout the specification, when a part is said to be "connected" to another part, this includes not only cases where it is "directly connected", but also cases where it is "electrically connected" with other elements interposed therebetween. Also, when a part is said to "include" a certain component, this means that, unless otherwise stated to the contrary, it does not exclude other components and can further include other components.

[0025] FIG. 1 is a drawing showing the configuration of an integrated fire detection system according to an embodiment of the present invention, and FIG. 2 is a drawing exemplarily showing a camera equipped with a flame detection sensor.

[0026] As shown in FIG. 1, an integrated fire detection system 100 according to an embodiment of the present invention includes a sensing value acquisition unit 110, an image acquisition unit 120, an image analysis unit 130, and a control unit 140, and may further include a prediction unit 150, a threshold determination unit 160, a camera 170, a flame detection sensor 180, and a display unit 190.

[0027] Here, the operations of the respective units 110 to 130, 150 to 190 and the flow of data between the units are controlled by the control unit 140.

[0028] The sensing value acquisition unit 110 can acquire a flame sensing value within a monitoring area from the flame detection sensor 180. The image acquisition unit 120 can acquire an image of the monitoring area taken by the camera 170. The control unit 140 can provide the video data for the monitoring area acquired by the image acquisition unit 120 through the display unit 190. The display unit 190 may be embodied as being included in the control unit 140, or may be included in a separately provided official server or user terminal connected to a network.

[0029] The flame detection sensor 180 can detect flames occurring on-site based on infrared radiation. The flame detection sensor 180 can detect the intensity of a flame by filtering signals in the infrared wavelength band corresponding to the flame component. In this process, the levels of infrared radiation energy emitted from the flame are measured differently, which enables the detection of the intensity of the flame signal.

[0030] As shown in Figure 2, the flame detection sensor 180 can be implemented as a device built into the camera 170. Such a camera 170 can be installed in the monitoring area to collect video and flame sensing values ​​in real time and provide them to the sensing value acquisition unit 110 and the video acquisition unit 120.

[0031] Here, camera 170 corresponds to a network camera installed at the site. Camera 170 can provide video footage of the monitored site and flame sensing values ​​to the integrated fire detection system 100 in real time via a wired, wireless, or mixed wired / wireless network.

[0032] The flame detection sensor 180 can be integrated into the camera 170, connected externally to the camera 170, or operated separately from the camera 170. While the diagram illustrates a configuration where the flame detection sensor 180 and camera 170 are connected externally to the integrated fire detection system 100, they may also be included within the integrated fire detection system 100.

[0033] If the flame sensing value obtained through the flame detection sensor 180 exceeds a predetermined critical value, the control unit 140 determines that a flame has been detected, generates a primary alarm, and controls the video analysis unit 130 to perform video analysis on the artificial intelligence platform.

[0034] The video analysis unit 130 can analyze the video through artificial intelligence when flames are detected, recognize flames and smoke to detect a fire, and locate the fire's location within the video.

[0035] The video analysis unit 130 can input video footage into a pre-trained artificial intelligence model to detect the presence or absence of a fire. For example, the artificial intelligence model can detect features (hue, shape, pattern, etc.) related to flames and smoke within the input video to determine the location of the fire or ignition point within the video. The video analysis unit 130 can then provide the results of the video analysis to the control unit 140.

[0036] The control unit 140 can generate a primary alarm on the screen in a predetermined manner (e.g., screen flashing) when a flame is detected by the flame detection sensor 180, and can generate a secondary alarm including the fire location on the screen in a predetermined manner (e.g., displaying the fire location in the video as a box) when a fire is detected through video analysis.

[0037] In this way, the control unit 140 flashes the entire screen of the display unit 190, which is currently outputting video, when a primary alarm occurs, and flashes the screen when a secondary alarm occurs, and also displays the fire detection location as a box within the video for guidance.

[0038] The integrated fire detection system 100 can simultaneously display video from different locations on a single screen, enabling multi-level monitoring of different areas.

[0039] In this case, the control unit 140 maps the video from multiple monitoring areas to multiple split screen areas arranged in a matrix and displays them simultaneously. When an alarm occurs, it can output a primary or secondary alarm on the split screen area corresponding to the monitoring area using a configurable method.

[0040] For example, the control unit 140 can divide the display screen into 4 rows x 5 columns and output the video of a total of 20 monitoring areas in a one-to-one matching manner to 20 divided screen areas. If a flame is detected in the nth monitoring area, only the corresponding nth divided screen area will flash. If, as a result of the video analysis, a fire is detected at the upper left corner of the nth monitoring area, the real-time video of the nth monitoring area currently being output in the nth divided screen area can be displayed with a box indicating the fire detection point at the upper left corner.

[0041] In an embodiment of the present invention, when a fire is detected, the video analysis unit 130 analyzes the video of the detection area to detect at least one object, then classifies the human object, and can extract the number of human objects, the detection position in the video, and the motion value based on the object classification result in the video.

[0042] In this process, the video analysis unit 130 uses an artificial intelligence-based algorithm to detect, classify, and track objects within the video, thereby detecting at least one object within the video and classifying it as a human object. This allows the unit to obtain the number of human objects in the video and their detection locations.

[0043] Furthermore, the motion values ​​of human objects can be confirmed by tracking their movements. When a person is lying down or has fallen, a low motion value is detected, while when they are walking or running, a high motion value is detected.

[0044] The prediction unit 150 can predict the probability of a dangerous situation occurring by applying the number of human objects analyzed from the video, the motion values ​​of the human objects, and the distance between the detected location of the human objects in the video and the fire location to a pre-trained prediction model.

[0045] In this case, the prediction unit 150 can further apply the size of the fire and the speed at which the fire spreads, analyzed from the video, to the prediction model to predict the probability of a dangerous situation occurring, and based on this, it can improve the accuracy of the prediction. Of course, for this purpose, the video analysis unit 130 can further calculate the size of the fire and the speed at which the fire spreads by analyzing the video of the monitored area based on deep learning when a fire occurs.

[0046] Here, the predictive model is pre-trained by a learning unit (not shown) based on big data collected for various monitoring domains. For example, the learning unit can be trained by setting the number of human subjects, motion values, and the distance between the human subjects and the fire location as input data, and the presence or absence of the actual dangerous situation in the case as output data.

[0047] Here, the control unit 140 compares the probability of a dangerous situation occurring, derived through the prediction model, with a critical value, and if the probability of a dangerous situation occurring is greater than or equal to the critical value, it can generate a dangerous situation alarm. Such a dangerous situation alarm is output through the screen of the display unit 190. In this way, the present invention can provide not only primary and secondary alarms related to simple fires, but also dangerous situation alarms that take into account the degree of danger to human subjects.

[0048] In embodiments of the present invention, the alarm can be not only a visual alarm through the display unit 190, but also an auditory alarm through a speaker built into the display unit 190 or a separate output means.

[0049] Here, the critical value determination unit 160 can determine the critical value by applying the number of human subjects analyzed from the video of the current monitoring area, the motion values ​​of the human subjects, and the distance between the human subjects and the fire location to a previously trained critical value determination model.

[0050] Here, the critical value determination model is trained to determine a lower critical value for generating a hazard alarm when there are more human objects, when the movement values ​​of the human objects are small, or when the distance between the human objects and the fire location is short.

[0051] Generally, the more people present at a fire, and the closer the distance between the fire and the people, the greater the loss of life. Furthermore, a small amount of movement in the human body indicates situations such as falling, tripping, being trapped under structures, momentary loss of consciousness, or cardiac arrest, which also constitute a critical situation requiring rapid response. Therefore, the more such situations are detected, the lower the critical threshold can be set, increasing the sensitivity of the danger alarm.

[0052] The control unit 140 can determine whether or not a dangerous situation exists based on the critical value determined by the critical value determination model. The control unit 140 can determine that a dangerous situation has occurred if the probability of a dangerous situation occurring is greater than or equal to the critical value, and that a dangerous situation did not occur if the probability of a dangerous situation occurring is less than the critical value.

[0053] Thus, according to the present invention, a critical value is determined according to the surrounding conditions, and by recognizing dangerous situations and providing a dangerous situation alarm, false alarms are reduced, and a rapid response to emergency situations is possible.

[0054] Figure 3 is a diagram illustrating an integrated fire detection method according to an embodiment of the present invention.

[0055] First, the integrated fire detection system 100 can acquire video of the monitoring area through the camera 170 and output it in real time through the screen of the display unit 190 (step S310). Then, the integrated fire detection system 100 can detect flames occurring within the monitoring area using the flame detection sensor 180 (step S320).

[0056] The integrated fire detection system 100 generates a primary alarm when a flame is detected by the flame detection sensor 180 (step S330), analyzes the video being captured in the monitoring area based on artificial intelligence to recognize flames and smoke, detect the presence or absence of a fire, and locate the fire location in the video (step S340).

[0057] The integrated fire detection system 100 can generate a secondary alarm containing information about the fire location if a fire is actually detected through AI video analysis (step S350). Here, during the primary alarm, a method can be used in which the entire screen on which the video is being output flashes, and during subsequent secondary alarms, the screen can flash while displaying the actual detection location in the video as a box.

[0058] In step S340, human detection within the video can also be performed, and human object detection may occur after step S350. Furthermore, through human object detection within the video, the number of human objects present in the video, the movement values ​​of the humans, and the distance between the human objects and the fire location can be obtained.

[0059] The integrated fire detection system 100 can predict the probability of a dangerous situation occurring by applying the number of human objects, motion values, and distance between the human objects and the fire location, which are analyzed from the video in this way, to a pre-trained predictive model (step S360).

[0060] In this case, if multiple human objects are detected, the motion value may be the largest of the motion values ​​detected for each human object, or the average value may be adopted. Similarly, in the case of distance, the shortest distance value among the distances between fire locations determined for each human object may be adopted, or the average distance value may be adopted.

[0061] In this case, the integrated fire detection system 100 can determine that a dangerous situation has occurred and generate a dangerous situation alarm if the probability of a dangerous situation occurring is above a critical value (step S370). In the embodiment of the present invention, the critical value for generating a dangerous situation alarm can be adaptively adjusted by considering the number of human objects detected, the motion value, and the distance between the human object and the fire location. That is, the critical value can be set to be lower the higher the degree of danger.

[0062] According to the present invention as described above, the efficiency and accuracy of fire detection can be improved by utilizing both AI-based video analysis and flame detection results.

[0063] Although the present invention has been described with reference to the embodiments shown in the drawings, these are merely illustrative, and those skilled in the art will understand that a variety of modifications and equivalent other embodiments are possible. Therefore, the true scope of technical protection of the present invention must be determined by the technical idea of ​​the claims.

Claims

1. A sensing value acquisition unit that acquires flame sensing values ​​within the monitoring area from a flame detection sensor, A video acquisition unit that acquires video footage of the aforementioned monitoring area from a camera, The aforementioned video analysis unit analyzes the video through artificial intelligence when a flame is detected to recognize flames and smoke, detect a fire, and locate the fire location in the video. The system further includes a prediction unit that applies the number of human objects analyzed from the video, the motion values ​​of the human objects, and the distance between the detected location of the human objects and the fire location to a previously trained prediction model to predict the probability of a dangerous situation occurring. A critical value determination unit that determines a critical value by applying the number of human objects analyzed from the aforementioned video, the motion values ​​of the human objects, and the distance between the detection position of the human objects and the fire location to a previously learned critical value determination model, The control unit outputs video of the monitoring area in real time via a screen, generates a primary alarm according to a set method when a flame is detected by the flame detection sensor, generates a secondary alarm including the fire location according to a set method when a fire is detected through video analysis, compares the predicted probability of a dangerous situation occurring with the critical value, and if it is greater than or equal to the critical value, determines that a dangerous situation has occurred and generates a dangerous situation alarm. An integrated fire detection system, including...

2. The control unit, When the primary alarm is triggered, the entire screen on which the video is being output will flash. The integrated fire detection system according to claim 1, wherein the screen flashes when the secondary alarm occurs, and the detection location is displayed as a box within the video output to provide notification.

3. The control unit, The integrated fire detection system according to claim 2, wherein video from multiple monitoring areas is mapped to multiple split screen areas arranged in a matrix and displayed simultaneously, and when an alarm occurs, the primary alarm or secondary alarm is output on the split screen area corresponding to the monitoring area in a pre-configured manner.

4. The aforementioned video analysis unit, The integrated fire detection system according to claim 1, wherein when a fire is detected, the video is analyzed to detect at least one object, then the human object is classified, and the number of the human object, the detection location, and the motion value are extracted.

5. The aforementioned critical value determination model is, The integrated fire detection system according to claim 1, wherein the system learns to lower the critical value for generating a danger alarm as the number of human objects increases, the movement value of the human objects decreases, or the distance between them decreases.

6. The prediction unit, The integrated fire detection system according to claim 1, further applying the size of the fire and the speed at which the fire spreads, analyzed from the aforementioned video footage, to the prediction model to predict the probability of the dangerous situation occurring.

7. In an integrated fire detection method performed by an integrated fire detection system, The process involves acquiring video footage of the surveillance area via a camera and outputting it in real time through a screen, A step of detecting a flame in the monitoring area using a flame detection sensor, The step of generating a primary alarm according to a set method when a flame is detected by the flame detection sensor, The process involves analyzing the video footage of the aforementioned monitoring area using artificial intelligence to recognize flames and smoke, detecting a fire, and identifying the location of the fire within the video footage. The process involves generating a secondary alarm, including the location of the fire, in a pre-configured manner upon fire detection through the aforementioned video analysis, The process involves applying the number of human objects analyzed from the aforementioned video, the motion values ​​of the human objects, and the distance between the detected location of the human objects and the fire location to a previously trained predictive model to predict the probability of a dangerous situation occurring. The process involves applying the number of human objects analyzed from the aforementioned video, the motion values ​​of the human objects, and the distance between the detection position of the human objects and the fire location to a previously trained critical value determination model to determine the critical value, The steps include comparing the predicted probability of a dangerous situation occurring with the critical value, and if it is greater than or equal to the critical value, determining that a dangerous situation has occurred and generating a dangerous situation alarm, An integrated fire detection method, including...

8. The integrated fire detection method according to claim 7, wherein the step of generating the primary alarm involves flashing the entire screen on which the video is being output, and the step of generating the secondary alarm involves flashing the screen and displaying the detection point as a box within the video being output.

Images