CN114724330A - A realization method of multi-channel video fire real-time alarm system with adaptive mode switching - Google Patents

Abstract

The invention discloses a method for realizing a multi-channel video fire real-time alarm system with self-adaptive mode switching, which comprises the following steps of one to four. The invention collects the images of the monitoring area in real time through the multi-path cameras and sends the images to the control center for detecting and judging whether a fire alarm occurs, compared with the traditional smoke temperature sensor type fire alarm system, the system is not limited by space, the response is rapid, and the detection accuracy is high; compared with the image type fire alarm system using the traditional pattern recognition, the method adopts the deep learning target detection algorithm to detect the fire, has higher accuracy and stronger generalization capability, and can obtain the position information of the fire; the detection algorithm provided by the invention can be switched in a self-adaptive manner according to different modes of the common infrared camera; the invention provides a fire evaluation module which is matched with pictures for marking flame and smoke to more intuitively show the fire to a user.

Description

A realization method of multi-channel video fire real-time alarm system with adaptive mode switching

technical field

The invention belongs to the technical field of real-time fire alarm and fire protection, and in particular relates to an implementation method of a multi-channel video fire real-time alarm system with adaptive mode switching.

Background technique

Fire is one of the most frequent and common disasters that endanger public safety and threaten social development. The occurrence of fire will cause personal and property damage. According to the relevant national fire protection laws, regulations and technical specifications, in order to prevent fires and reduce the losses caused by fires, all buildings need to design and install fire-fighting facilities. Loss.

In the past fire detection, one or more smoke sensors or temperature sensors are generally used as fire detectors. After receiving the fire alarm signal, it is judged whether it is a real alarm through manual on-site confirmation. In the early stage of the fire, the burning area is not large, and it is limited to the vicinity of the initial ignition point, and the fire is unstable; there is high temperature in and near the burning area, the average indoor temperature is low, and the indoor temperature difference is large, and the temperature alarm device cannot be triggered in time; at the same time, there will be smoke in the early stage of the fire. , but it also takes a long time for the smoke to disperse to trigger the smoke sensor. And according to the "Code for Design of Automatic Fire Alarm System", the installation height of temperature-sensing smoke detectors should not exceed 12 meters, so it cannot be used for indoor and outdoor scenes in large spaces. Fire alarm devices based on temperature and smoke sense generally cannot be visualized themselves, and managers can only receive alarm signals and need to go to the scene to confirm the fire situation, or install additional surveillance cameras to increase costs. Late discovery, late alarm, and late disposal are usually the reasons for huge losses caused by fire, so it is particularly important for a fire alarm system to be able to alarm in time at the early stage of a fire.

At present, there are also image-based fire alarm systems on the market. For example, in the patent application number CN201810061303.2, the foreground image is obtained by the background difference method, and then the area characteristics, frequency characteristics and mass center motion characteristics of each area in the foreground image are used to judge whether a fire occurs; for example; In the patent application number CN201611068303.2, the dynamic characteristics and color characteristics of the flame smoke are used to detect fire, both of which use the traditional mode detection method to detect fire, which has poor generalization ability and is easily affected by environmental changes; for example, the application number CN202110041008.2 uses convolution The neural network realizes fire detection by classifying pictures that contain flames and pictures that do not contain flames. It can only detect whether there is a flame and cannot obtain the ignition point information through the algorithm. Moreover, smoke is often generated in the early stage of the fire. Only detecting the flame and ignoring the smoke detection will cause The alarm time is delayed.

When an ordinary infrared camera switches to infrared mode in a dim environment, the image characteristics of flames and smoke in the night monitoring screen are very different from those in the ordinary mode. We need an initial fire alarm system with fast response, high accuracy and strong generalization ability, which is applicable in both normal mode and night mode of the camera.

SUMMARY OF THE INVENTION

The purpose of the present invention is to provide an implementation method of a multi-channel video fire real-time alarm system for adaptive mode switching, so as to solve the problems in the prior art proposed in the above-mentioned background technology.

To achieve the above object, the present invention has adopted the following technical solutions:

A kind of realization method of the multi-channel video fire real-time alarm system of self-adaptive mode switching, is characterized in that, comprises the steps:

Step 1, the multi-channel cameras simultaneously capture the videos of multiple monitoring areas;

In step 2, the collected video is sent to the control center for processing, and fire information is output;

Step 3, alarm locally, save the fire information, and upload the alarm information to the cloud server;

Step 4, the user receives the alarm information on the mobile phone APP;

It also includes a multi-channel video fire real-time alarm system combined with smoke and flame detection algorithms, including image acquisition terminals, hard disk recorders, switches, control centers, cloud servers, and mobile phone terminals;

The image acquisition terminal is installed in the monitored area and used to collect the video of the monitored area;

The hard disk video recorder, switch, and control center are installed in the central control room. The hard disk recorder is used to store surveillance video, and the switch is used to transmit the images transmitted by multi-channel cameras to the control center. In the control center, smoke and flame detection algorithms are used to determine whether the image exists. Smoke and flame, if there is, the flame or smoke information output by the algorithm will be sent to the fire evaluation module to get the fire level, and the local alarm will be reported. At the same time, the fire information will be stored in the hard disk and sent to the cloud server. If there is no fire In case of fire, it continues to detect the next frame; the cloud server organizes the fire information and detection results and transmits it to the mobile terminal, and the mobile terminal displays the detection results and alarms.

Preferably, it also includes a fire detection algorithm for adaptive mode switching, and the algorithm as a whole includes an adaptive mode switching module, a fire detection algorithm module and a fire situation assessment module.

Preferably, the adaptive mode switching module includes:

Before detecting the flame, the mode of the camera is judged every 30s. The algorithm is as follows:

BEGIN (input image img):

Bm, Gm, Rm = the average value of each channel of the image;

DIFF1=Bm-Gm; DIFF2=Bm-Rm; DIFF3=Rm-Gm;

IF(SUM(DIFF1,DIFF2,DIFF3)/3<=1):

THEN return 1;

ELSE return 0;

Since Opencv reads the image of the camera in BGR format, it can only be judged by judging whether the average value of each channel value pixel is equal. If it returns 1, it is normal mode, and if it returns 0, it is night mode.

Preferably, the fire detection algorithm module includes:

Step 1, build a flame smoke sample library;

Step 2, train the flame smoke target detection model;

Step 3, fire detection algorithm.

Preferably, step 2 of the fire detection algorithm module includes:

The algorithm model used at night is trained, and the flame and smoke data are extracted from the flame and smoke sample library. Since it is difficult to obtain the fire scene image captured by the infrared camera in the night mode, we simulate it through the following algorithm:

BEGIN (input image img):

img = img.gray();

img=img.point()*0.7;

return img;

Preferably, step 3 of the fire detection algorithm module includes the following steps:

Obtain the image from the image acquisition terminal, and determine whether to use the night mode or the normal mode through the adaptive mode switching module;

After the adaptive mode switching module, if the normal mode is entered, the third frame is used to detect smoke. In the normal mode, the smoke detection algorithm calls the daytime smoke detection model when the frame counter is 3. If there is smoke, the smoke information is sent to the fire. If the situation evaluation module does not exist, it will detect the next frame;

The 6th frame detects the flame. The normal mode flame detection algorithm calls the daytime flame detection model when the frame counter is 6, and resets the frame counter to 1. If there is a flame, the flame information is sent to the fire evaluation module. Then detect the next frame detection;

If you enter the night mode, the structure of the smoke and flame detection algorithm at night is the same as the algorithm in the normal mode. The flame smoke target detection algorithm model uses the algorithm model of night smoke and flame trained in step 2. After the detection, the information of the flame and smoke is sent to the fire situation. Evaluation module.

Preferably, the fire assessment module includes:

The fire situation divides the flame grade into 4 grades according to the area and area change rate of the smoke and flame area, and divides the smoke grade into 3 grades, and then determines the alarm mode according to the classification situation.

Preferably, judging the alarm level in the fire assessment module includes:

If it is smoke, judge whether the smoke area ratio S1 is greater than or equal to 0.3 at this time: if yes, then the smoke level is level 3;

If it is otherwise, compare it with the area proportion of the previous smoke: if it is greater than that, add 1 to the smoke level; if it is less than, keep the current smoke level unchanged;

If it is a flame, judge whether the flame area ratio S2 is greater than or equal to 0.1 at this time: if it is, the flame level is 4, if it is otherwise, it is compared with the area ratio of the previous flame. If it is greater than that, the flame level is increased by 1 , if it is less than, keep the current flame level unchanged.

Preferably, the smoke level situation is divided into 3 levels, and the flame level situation is divided into 4 levels;

Determining the alarm mode includes determining the alarm mode through the following steps after obtaining the fire level through the fire evaluation module:

When there is only a smoke alarm, if the smoke level is 1 or 2, it will enter the pre-alarm state, and if no other alarm occurs within 5s, the pre-alarm will be automatically reset;

When the smoke level is level 3, or when a flame is detected, a fire alarm is issued immediately and cannot be reset automatically.

Technical effect and advantage of the present invention: the realization method of a kind of self-adaptive mode switching multi-channel video fire real-time alarm system proposed by the present invention has the following advantages compared with the prior art:

The invention provides an implementation method of a multi-channel video fire real-time alarm system with adaptive mode switching. The images of the monitoring area are collected in real time by multi-channel cameras and sent to the control center for detection and judgment whether there is a fire alarm. Compared with the image-based fire alarm system, it is not restricted by space, has a rapid response, and has a high detection accuracy; compared with the image-based fire alarm system using traditional pattern recognition, the present invention adopts the deep learning target detection algorithm to detect fire, and the accuracy is higher. , the generalization ability is stronger, and the location information of the fire can be obtained;

The detection algorithm provided by the present invention can switch the detection algorithm adaptively according to the different modes of the common infrared camera, and train four sets of target detection algorithm models to detect the flame smoke during the day and the flame smoke at night respectively. In the process of training the algorithm model, the present invention obtains part of the smoke data set by means of image synthesis, uses the color image to simulate the grayscale image monitored in the night mode by means of image processing, trains the detection model used at night, and aims at different application places. , the actual picture of the monitoring area is used as the background to train together to improve the accuracy of the model;

The invention proposes a fire situation evaluation module, which calculates the target area according to the position frame of the flame and smoke obtained by the fire detection algorithm, evaluates the fire situation level according to the change of the area, adopts different alarm methods according to the fire situation level, and cooperates with marking the flame. and smoke pictures more intuitively show the fire to the user;

The present invention can be directly connected to the original video monitoring system, and can be put into use only by configuring the equipment of the relevant control center, which can reduce the personnel and equipment costs for repetitively laying monitoring equipment.

Other features and advantages of the present invention will be set forth in the description which follows, and, in part, will be apparent from the description, or may be learned by practice of the invention. The objectives and other advantages of the invention may be realized and attained by the structure pointed out in the description and drawings.

Description of drawings

Fig. 1 is a kind of overall flow chart of the realization method of the multi-channel video fire real-time alarm system in conjunction with smoke and flame detection algorithm of the present invention;

Fig. 2 is one of the overall frameworks of a kind of multi-channel video fire real-time alarm system in conjunction with smoke and flame detection algorithm of the present invention;

Fig. 3 is a kind of overall framework of the multi-channel video fire real-time alarm system in conjunction with smoke and flame detection algorithm of the present invention;

4 is a flowchart of a smoke and flame detection algorithm in an embodiment of the present invention;

5 is a schematic diagram of a smoke and flame detection algorithm in an embodiment of the present invention;

FIG. 6 is a schematic flowchart of a fire situation assessment module in an embodiment of the present invention.

Labels in the figure: 1, image acquisition terminal; 2, hard disk video recorder; 3, switch; 4, control center; 5, cloud server; 6, mobile phone terminal; 7, storage unit; 8, processing unit; 9, alarm unit.

Detailed ways

The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. The specific embodiments described herein are only used to explain the present invention, and are not intended to limit the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative work shall fall within the protection scope of the present invention.

The present invention provides embodiments as shown in Figures 1-6:

A method for implementing a multi-channel video fire real-time alarm system combined with smoke and flame detection algorithms, the implementation steps of which are shown in Figure 1 and include:

Step 1, the multi-channel cameras simultaneously capture the videos of multiple monitoring areas;

In step 2, the collected video is sent to the control center for processing, and fire information is output;

Step 3, alarm locally, save the fire information, and upload the alarm information to the cloud server;

Step 4, the user receives the alarm information on the mobile phone APP;

A kind of realization method of the multi-channel video fire real-time alarm system in conjunction with smoke and flame detection algorithm of the present application will be elaborated below in detail:

As shown in Figure 2-3, it is an overall framework of a multi-channel video fire real-time alarm system combined with smoke and flame detection algorithms, including image acquisition terminal 1, hard disk recorder 2, switch 3, control center 4, cloud server 5, Mobile terminal 6;

The image acquisition terminal 1 is connected with the hard disk video recorder 2, the hard disk video recorder 2 is connected with the switch 3, the switch 3 is connected with the control center 4, the control center 4 is connected with the cloud server 5, and the cloud server 5 is connected with the mobile terminal 6; It includes a storage unit 7 , a processing unit 8 and an alarm unit 9 .

Among them, the image acquisition terminal 1 is installed in the monitored area to collect the video of the monitored area; the hard disk video recorder 2, the switch 3, and the control center 4 are installed in the central control room, the hard disk video recorder 2 is used to store the monitoring video, and the switch 3 is used to The image transmitted by the camera is sent to the control center 4, and the smoke and flame detection algorithm is used in the control center 4 to determine whether there is smoke or flame in the image. If there is, the flame or smoke information output by the algorithm is sent to the fire evaluation module. Fire level, and alarm locally, and store the fire information in the hard disk and send it to the cloud server 5. If there is no fire, continue to detect the next frame; the cloud server 5 organizes the fire information and detection results and other data. It is transmitted to the mobile terminal 6, and the mobile terminal 6 displays the detection result and alarms;

In the implementation method of a multi-channel video fire real-time alarm system combined with smoke and flame detection algorithms proposed by the present invention, an adaptive mode switching fire detection algorithm is designed. The algorithm as a whole includes: (1) an adaptive mode switching module , (2) fire detection algorithm module, (3) fire assessment module

(1) In the embodiment of this application, the first one is an adaptive mode switching module, which includes:

Before detecting the flame, the mode of the camera is judged every 30s. The algorithm is as follows:

BEGIN (input image img):

Bm, Gm, Rm = the average value of each channel of the image;

DIFF1=Bm-Gm; DIFF2=Bm-Rm; DIFF3=Rm-Gm;

IF(SUM(DIFF1,DIFF2,DIFF3)/3<=1):

THEN return 1;

ELSE return 0;

Since Opencv reads the image of the camera in BGR format, it can only be judged by judging whether the average value of each channel value pixel is equal. If it returns 1, it is normal mode, and if it returns 0, it is night mode.

(2) In the embodiment of the present application, the second is a fire detection algorithm module, which includes:

Step 1, build the flame smoke sample library:

Before training, we need to build a fire and smoke sample library. We collect a large number of fire and smoke images from the network and images obtained from burning experiments in the field. These images include a variety of indoor, rooftop, office and other scenes, strong light, weak Lighting, various experimental environments during the day and night, wood, plastic, paper, gasoline and other materials, because most of the fire smoke images on the Internet are the background of thick smoke in the later stage of the fire, and most of the images are of poor quality and insufficient quantity. Smoke images in different scenes are obtained by synthesizing smoke and background images, and these data are manually marked to obtain a flame smoke sample library.

Step 2, train the flame smoke target detection model:

Train the algorithm model used during the day, extract the flame and smoke data from the flame and smoke sample library, in order to improve the model generalization ability, Gaussian noise, contrast enhancement and other methods for data expansion, and in order to make the model learn more to monitor the situation on site , in addition to preparing the labeled smoke data set, the monitoring screen of the on-site monitoring area is added as the background, accounting for about 10% of the training data. Adding the background can effectively reduce false positives. We divide the obtained data into training set and verification Set two parts, then build a training environment on the server, and finally train the flame and smoke through the data set to obtain the target detection algorithm model. After obtaining the algorithm model, deploy it in the control center to complete the detection function;

The algorithm model used at night is trained, and the flame and smoke data are extracted from the flame and smoke sample library. Since it is difficult to obtain the fire scene image captured by the infrared camera in the night mode, we simulate it through the following algorithm:

BEGIN (input image img):

img = img.gray();

img=img.point()*0.7;

return img;

By converting the images into grayscale images and darkening, the nighttime monitoring images are simulated, and the monitoring images of the on-site monitoring area collected at night are also used as the background, accounting for 10% of the training samples, and the obtained training samples are divided into training sets and validation sets. , and then build a training environment on the server, and finally train the flame and smoke through the data set to obtain the target detection algorithm model. After the algorithm model is obtained, it is deployed in the control center to complete the detection function;

Step 3, fire detection algorithm:

As shown in Figure 4 is the flow chart of the detection algorithm in the fire alarm system, in this embodiment, comprises the following steps:

①. Obtain the image from the image acquisition terminal, and determine whether to use the night mode or the normal mode through the adaptive mode switching module.

②. After the adaptive mode switching module, if the normal mode is entered, the third frame is used to detect smoke. The smoke detection algorithm in the normal mode is shown in (a) in Figure 5. When the frame counter is 3, the daytime smoke detection model is called. If there is smoke, send the smoke information to the fire evaluation module. If it does not exist, the next frame will be detected; the 6th frame will detect the flame, of which the normal mode flame detection algorithm is shown in Figure 5 (b) when the frame counter is 6 Call the daytime flame detection model, reset the frame counter to 1, if there is a flame, send the flame information to the fire evaluation module, if not, detect the next frame detection;

If the night mode is entered, the structure of the nighttime smoke and flame detection algorithm is the same as that in the normal mode, and the flame smoke target detection algorithm model uses the nighttime smoke and flame algorithm model trained in step 2. After the detection, the information of the flame and smoke is sent to the fire assessment module.

(3) In the embodiment of the application, the third is a fire assessment module, which includes:

As shown in Figure 6, the fire evaluation module, the fire situation divides the flame level into 4 levels according to the area of smoke and flame area and the area change rate, and divides the smoke level into 3 levels, and then determines the alarm mode according to the classification situation.

Among them, judging the alarm level in the fire assessment module includes: 1. If it is smoke, then judge whether the proportion of smoke area S1 is greater than or equal to 0.3 at this time. Compared with the ratio, if the ratio is greater than that, the smoke level will be increased by 1; if it is less than the ratio, the current smoke level will remain unchanged. ②If it is a flame, judge whether the proportion of the flame area S2 is greater than or equal to 0.1 at this time. If so, the flame level is 4. If it is, it is compared with the area proportion of the previous flame. If it is greater than that, the flame level will be added 1, if less than, keep the current flame level unchanged. In the present invention, the smoke level situation is divided into 3 levels, and the flame level situation is divided into 4 levels;

Determining the alarm mode includes determining the alarm mode through the following steps after obtaining the fire level through the fire evaluation module:

①When there is only a smoke alarm, if the smoke level is 1 or 2, it will enter the pre-alarm state, and if no other alarm occurs within 5s, the pre-alarm will be automatically reset;

② When the smoke level is level 3, or when a flame is detected, a fire alarm will be issued immediately and cannot be reset automatically.

To sum up, the invention of the present application can monitor the abnormal fire situation in the environment in real time, and push the abnormal fire situation to the mobile phone terminal in the form of an intuitive image, so that the user can intuitively see the actual situation of the scene, and can prevent false alarms, false alarms, etc. False alarms and fire situations are accurately handled to achieve multi-party coordinated fire management and improve reliability. The adopted smoke and flame detection algorithm improves the response speed, judgment sensitivity and judgment accuracy of the system, and can be retrofitted on the existing video surveillance system, reducing hardware costs.

Finally, it should be noted that the above descriptions are only preferred embodiments of the present invention, and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, for those skilled in the art, the The technical solutions described in the foregoing embodiments can be modified, or some technical features thereof can be equivalently replaced, and any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included. within the protection scope of the present invention.

Claims (9)

1. the realization method of the multi-channel video fire real-time alarm system of self-adaptive mode switching, is characterized in that, comprises the steps: Step 1, the multi-channel cameras simultaneously capture the videos of multiple monitoring areas; In step 2, the collected video is sent to the control center for processing, and fire information is output; Step 3, alarm locally, save the fire information, and upload the alarm information to the cloud server; Step 4, the user receives the alarm information on the mobile phone APP; It also includes a multi-channel video fire real-time alarm system combined with smoke and flame detection algorithms, including an image acquisition terminal (1), a hard disk video recorder (2), a switch (3), a control center (4), a cloud server (5), and a mobile phone terminal. (6); Wherein the image acquisition terminal (1) is installed in the monitored area, and is used for collecting the video of the monitored area; The hard disk video recorder (2), the switch (3), and the control center (4) are installed in the central control room. The hard disk video recorder (2) is used to store surveillance video, and the switch (3) is used to transmit the images transmitted by the multi-channel cameras to the control center ( 4), in the control center (4), the smoke and flame detection algorithm is used to judge whether there is smoke and flame in the image, if there is, the flame or smoke information output by the algorithm is sent to the fire evaluation module to obtain the fire level, and in the Local alarm, at the same time, the fire information is stored in the hard disk and sent to the cloud server (5). If there is no fire, the next frame will continue to be detected; A mobile phone terminal (6), the mobile phone terminal (6) displays the detection result and gives an alarm. 2. the realization method of the multi-channel video fire real-time alarm system of a kind of self-adaptive mode switching according to claim 1, is characterized in that: also comprises the fire detection algorithm of self-adaptive mode switching, the algorithm as a whole comprises self-adaptive mode switching module , Fire detection algorithm module and fire assessment module. 3. the realization method of the multi-channel video fire real-time alarm system of a kind of adaptive mode switching according to claim 2, is characterized in that: described adaptive mode switching module comprises: Before detecting the flame, the mode of the camera is judged every 30s. The algorithm is as follows: BEGIN (input image img): Bm, Gm, Rm = the average value of each channel of the image; DIFF1=Bm-Gm; DIFF2=Bm-Rm; DIFF3=Rm-Gm; IF(SUM(DIFF1,DIFF2,DIFF3)/3<=1): THEN return 1; ELSEreturn 0; Since Opencv reads the image of the camera in BGR format, it can only be judged by judging whether the pixel average value of each channel value is equal. If it returns 1, it is normal mode, and if it returns 0, it is night mode. 4. the realization method of the multi-channel video fire real-time alarm system of a kind of adaptive mode switching according to claim 2, is characterized in that: described fire detection algorithm module comprises: Step 1, build a flame smoke sample library; Step 2, train the flame smoke target detection model; Step 3, fire detection algorithm. 5. the realization method of the multi-channel video fire real-time alarm system of a kind of self-adaptive mode switching according to claim 4, is characterized in that: in step 2 of described fire detection algorithm module, comprise: The algorithm model used at night is trained to extract flame and smoke data from the flame and smoke sample library. Since the fire scene images captured by infrared cameras in night mode are difficult to obtain, we use the following algorithm to simulate: BEGIN (input image img): img = img.gray(); img=img.point()*0.7; return img. 6. the realization method of the multi-channel video fire real-time alarm system of a kind of self-adaptive mode switching according to claim 4, is characterized in that: the step 3 of described fire detection algorithm module comprises the following steps: Obtain the image from the image acquisition terminal, and determine whether to use the night mode or the normal mode through the adaptive mode switching module; After the adaptive mode switching module, if the normal mode is entered, the third frame is used to detect smoke. In the normal mode, the smoke detection algorithm calls the daytime smoke detection model when the frame counter is 3. If there is smoke, the smoke information is sent to the fire. If the situation evaluation module does not exist, it will detect the next frame; The 6th frame detects the flame. The normal mode flame detection algorithm calls the daytime flame detection model when the frame counter is 6, and resets the frame counter to 1. If there is a flame, the flame information is sent to the fire evaluation module. Then detect the next frame detection; If you enter the night mode, the structure of the smoke and flame detection algorithm at night is the same as the algorithm in the normal mode. The flame smoke target detection algorithm model uses the algorithm model of the night smoke and flame trained in step 2. After the detection, the information of the flame and smoke is sent to the fire situation. Evaluation module. 7. the realization method of the multi-channel video fire real-time alarm system of a kind of adaptive mode switching according to claim 2, is characterized in that: described fire situation assessment module comprises: The fire situation divides the flame grade into 4 grades according to the area and area change rate of the smoke and flame area, and divides the smoke grade into 3 grades, and then determines the alarm mode according to the classification situation. 8. the realization method of the multi-channel video fire real-time alarm system of a kind of adaptive mode switching according to claim 7, is characterized in that: Among them, judging the alarm level in the fire assessment module includes: If it is smoke, then judge whether the proportion of smoke area S1 is greater than or equal to 0.3 at this time: if so, the smoke level is level 3; If it is otherwise, compare it with the area proportion of the last smoke: if it is greater than, add 1 to the smoke level, if it is less than, keep the current smoke level unchanged; If it is a flame, judge whether the flame area ratio S2 is greater than or equal to 0.1 at this time: if it is, the flame level is 4, if it is otherwise, it is compared with the area ratio of the previous flame. If it is greater than that, the flame level is increased by 1 , if it is less than, keep the current flame level unchanged. 9. the realization method of the multi-channel video fire real-time alarm system of a kind of self-adaptive mode switching according to claim 8, is characterized in that: The smoke level situation is divided into 3 levels, and the flame level situation is divided into 4 levels; Determining the alarm mode includes determining the alarm mode through the following steps after obtaining the fire level through the fire evaluation module: When there is only a smoke alarm, if the smoke level is 1 or 2, it will enter the pre-alarm state, and if no other alarm occurs within 5s, the pre-alarm will be automatically reset; When the smoke level is level 3, or when a flame is detected, a fire alarm is issued immediately and cannot be reset automatically.

Images