A fire monitoring method based on vision

Abstract

The present invention discloses a fire monitoring method based on vision, which belongs to the field of security technology. The method comprises the following steps: establishing a background model of a monitoring location on a fire prevention and control center platform; obtaining the proportion of flammable items in the picture based on the picture taken by the camera; determining whether the monitored area is a fire-prone area or a fire-unlikely area; the camera determines the fire picture based on a deep neural network model; the fire prevention and control center platform simulates the fire development trend in the background model, solving the technical problems of using a visual method to perform fire alarms and predict the fire development trend. The present invention determines the fire picture based on the deep neural network model, and can quickly identify flames or smoke within 5 seconds of the appearance of the fire image. It utilizes more than 90% of various models of video monitoring systems currently on the market, has a simple loading method and is easy to operate, simulates the fire development trend in the background model, and provides strong support for fire fighting.

Description

Technical Field

[0001] The invention belongs to the technical field of security and protection, and relates to a fire monitoring method based on vision. Background Art

[0002] Existing fire monitoring generally consists of two types of systems: an automatic fire alarm system and a fire linkage system. The automatic fire alarm system converts physical quantities such as smoke, heat, and flames generated by combustion into electrical signals through fire detectors, transmits them to the fire alarm controller, and simultaneously notifies the entire floor to evacuate in the form of sound or light. The controller records the location and time of the fire, so that people can detect the fire in time and take effective measures to extinguish the initial fire.

[0003] However, the existing fire automatic alarm system has the following deficiencies:

[0004] 1. Fire detectors (probes) are out of service. Domestic probes generally have a lifespan of around 30,000 hours (three and a half years), while world-renowned brand fire detectors can reach up to 60,000 hours (seven years). However, some building alarm systems are in use for too long without replacing, repairing, or cleaning the probes. As a result, the detectors' sensitivity simply cannot meet operational requirements, and the fire information they receive is inaccurate. This can delay the alarm and cause a disaster.

[0005] 2. Common fire detectors collect physical quantities such as smoke, heat, and flames generated by combustion in a certain space, convert them into electrical signals through the fire detector, and transmit them to the fire alarm controller. This is a passive contact method and is easily affected by space, airflow, and temperature. It takes a long time and has low accuracy. By the time the alarm is triggered, a small fire has often turned into a large fire.

[0006] 3. Existing fire detectors are typically installed on the ceiling or walls of the protected space. They require a detected signal (such as smoke) to enter the detector and accumulate to a certain level before generating an alarm. This is a contact detection method. In some large spaces, outdoor areas, and locations with unstable airflow, environmental interference can make it difficult for fire signals to reach the detectors quickly and efficiently. This can cause these fire detectors to malfunction and affect detection precision and accuracy. Summary of the Invention

[0007] The present invention is to solve the above technical problems. The purpose of the present invention is to provide a fire monitoring method based on vision, which solves the technical problems of using visual means to make fire alarms and predict fire development trends.

[0008] To achieve the above object, the present invention adopts the following technical solutions:

[0009] A visual-based fire monitoring method includes deploying several cameras at a monitoring location, with each camera capturing a monitoring area;

[0010] All cameras communicate with the fire prevention and control center platform via data cables;

[0011] Establish a background model of the monitoring location on the fire prevention and control center platform;

[0012] According to the images captured by the camera, obtain the proportion of flammable items in the image;

[0013] Determine fire-prone areas and, based on the determination results, determine whether the monitored area is fire-prone or fire-unprone;

[0014] Add the monitoring area of ​​each camera to the background model;

[0015] When a fire occurs in a certain monitoring area, the camera determines the fire image based on the deep neural network model;

[0016] The fire prevention and control center platform displays the fire image and camera number information, and simultaneously identifies the monitoring area where the fire occurred in the background model and issues a fire alarm;

[0017] The fire prevention and control center platform simulates the fire development trend in the background model.

[0018] Preferably, the fire prevention and control center platform deploys a distributed server cluster or only deploys a separate central server, and the camera communicates with the distributed server cluster or the central server through a digital display.

[0019] Preferably, when determining a fire-prone area, the following steps are specifically included:

[0020] Step 1: Analyze the image captured by the camera and obtain all flammable items in the image through a deep neural network model;

[0021] Step 2: On the screen, mark each flammable item with a box;

[0022] Step 3: Calculate the area ratio of the box in the picture. When it reaches the preset value, the area captured by the camera is judged as a fire-prone area; otherwise, it is judged as a fire-prone area;

[0023] According to the methods of step 1 and step 2, the monitoring area captured by each camera is judged and divided into fire-prone areas and fire-unlikely areas.

[0024] Preferably, when simulating the fire development trend in the background model, the following steps are specifically included:

[0025] Step A1: calibrate the monitoring area where the fire occurs in the background model, and set the monitoring area as monitoring area A; Step A2: obtain a monitoring area B adjacent to monitoring area A;

[0026] Step A3: Determine whether the monitoring area B is a fire-prone area: if yes, determine that the fire is spreading to the monitoring area B; if no, determine that the fire does not want to spread to the monitoring area B;

[0027] Step A4: According to the method of step A2 and step A3, all monitoring areas adjacent to monitoring area A are judged;

[0028] Step A5: Based on the result of step A4, a fire development trend diagram is obtained and calibrated in the background model.

[0029] Beneficial effects of the present invention:

[0030] The vision-based fire monitoring method described in the present invention solves the technical problems of using visual methods to conduct fire alarms and predict fire development trends. The present invention judges fire images based on a deep neural network model, and can quickly identify flames or smoke within 5 seconds of the appearance of a fire image, and simultaneously issue a fire alarm signal. The detection speed is much faster than that of conventional fire detectors. It can simultaneously detect flames and smoke in the monitoring area and issue corresponding alarm signals. Whether it is smoldering smoke or obvious flames, it can be quickly identified and an alarm can be quickly issued. The present invention can utilize more than 90% of various models of video surveillance systems currently on the market. It only needs to obtain the camera image to make a judgment. The loading method is simple, the operation is convenient, and it has wide applicability and compatibility. The present invention can simulate the development trend of a fire in a background model, providing strong support for firefighting. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] Figure 1 is a flow chart of the present invention;

[0032] Figure 2 It is a flow chart of the present invention for calibrating a monitoring area into a fire-prone area or a fire-unprone area;

[0033] Figure 3 This is the monitoring picture actually taken by the present invention of the area where fire is not likely to occur;

[0034] Figure 4 It is a surveillance image of a fire-prone area captured by the world of the present invention;

[0035] In the figure: flammable material 1, non-flammable material 2, frame 4, and border 5. DETAILED DESCRIPTION

[0036] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0037] like Figures 1-4 A visual-based fire monitoring method is shown, which includes deploying several cameras at a monitoring location, with each camera capturing a monitoring area;

[0038] All cameras communicate with the fire prevention and control center platform via data cables;

[0039] Establish a background model of the monitoring location on the fire prevention and control center platform;

[0040] According to the images captured by the camera, obtain the proportion of flammable items in the image;

[0041] Determine fire-prone areas and, based on the determination results, determine whether the monitored area is fire-prone or fire-unprone;

[0042] Add the monitoring area of ​​each camera to the background model;

[0043] When a fire occurs in a certain monitoring area, the camera determines the fire image based on the deep neural network model;

[0044] The use of deep neural network models to identify open flames is an existing technology and will not be described in detail.

[0045] The fire prevention and control center platform displays the fire image and camera number information, and simultaneously identifies the monitoring area where the fire occurred in the background model and issues a fire alarm;

[0046] The fire prevention and control center platform simulates the fire development trend in the background model.

[0047] Preferably, the fire prevention and control center platform deploys a distributed server cluster or only deploys a separate central server, and the camera communicates with the distributed server cluster or the central server through a digital display.

[0048] Preferably, when determining a fire-prone area, the following steps are specifically included:

[0049] Step 1: Analyze the image captured by the camera and obtain all flammable items in the image through a deep neural network model;

[0050] The deep neural network model is an existing technology and is used to determine the type of specific objects in the picture, so it will not be described in detail.

[0051] Step 2: On the screen, mark each flammable item with a box;

[0052] like Figure 3 and Figure 4 As shown, in the embodiment, the box of the flammable item needs to frame the flammable item in the picture. In specific implementation, after obtaining the picture of the flammable item, the box used by this application is larger than the area occupied by the flammable item itself in the picture.

[0053] Step 3: Calculate the area ratio of the box in the picture. When it reaches the preset value, the area captured by the camera is judged as a fire-prone area; otherwise, it is judged as a fire-prone area;

[0054] According to the methods of step 1 and step 2, the monitoring area captured by each camera is judged and divided into fire-prone areas and fire-unlikely areas.

[0055] Preferably, when simulating the fire development trend in the background model, the following steps are specifically included:

[0056] Step A1: calibrate the monitoring area where the fire occurs in the background model, and set the monitoring area as monitoring area A; Step A2: obtain a monitoring area B adjacent to monitoring area A;

[0057] Step A3: Determine whether the monitoring area B is a fire-prone area: if yes, determine that the fire is spreading to the monitoring area B; if no, determine that the fire does not want to spread to the monitoring area B;

[0058] Step A4: According to the method of step A2 and step A3, all monitoring areas adjacent to monitoring area A are judged;

[0059] Step A5: Based on the result of step A4, a fire development trend diagram is obtained and calibrated in the background model.

[0060] like Figure 3 and Figure 4 As shown, when the present invention determines whether the monitored area is a fire-prone area, it only needs to calculate the area of ​​the box in the picture. In practical applications, the threshold standard for the proportion of judgment in this implementation is 60%, as shown in FIG. Figure 3 As shown in the picture, the picture frame is frame 5, the flammable object 1 is framed by frame 4, and the non-flammable object 2 is not framed by frame 4. Figure 3 In the figure, box 4 accounts for less than 30% of the entire image, which means that the monitored area is not prone to fire. Figure 4As shown, all the flammable objects 1 are framed by boxes 4. Finally, it is calculated that the total proportion of all the boxes 4 in the picture exceeds 60%, and the monitored empty area is determined to be a fire-prone area.

[0061] In this implementation, the monitoring areas captured by all cameras are calibrated in the background model. When a fire occurs in a certain area, the monitoring personnel can quickly determine the development trend of the fire through the background model, which provides certain guidance for firefighting work.

[0062] The vision-based fire monitoring method described in the present invention solves the technical problems of using visual methods to conduct fire alarms and predict fire development trends. The present invention judges fire images based on a deep neural network model, and can quickly identify flames or smoke within 5 seconds of the appearance of a fire image, and simultaneously issue a fire alarm signal. The detection speed is much faster than that of conventional fire detectors. It can simultaneously detect flames and smoke in the monitoring area and issue corresponding alarm signals. Whether it is smoldering smoke or obvious flames, it can be quickly identified and an alarm can be quickly issued. The present invention can utilize more than 90% of various models of video surveillance systems currently on the market. It only needs to obtain the camera image to make a judgment. The loading method is simple, the operation is convenient, and it has wide applicability and compatibility. The present invention can simulate the development trend of a fire in a background model, providing strong support for firefighting.

[0063] Those skilled in the art will understand that all or part of the steps in the method of the above embodiment can be completed by instructing related hardware through a program, and the program can be stored in a computer-readable storage medium. When the program is executed, it includes one or a combination of the steps of the method embodiment.

[0064] In addition, the functional units in the various embodiments of the present invention may be integrated into a single processing module, or each unit may exist physically separately, or two or more units may be integrated into a single module. The aforementioned integrated modules may be implemented in the form of hardware or in the form of software functional modules. If the integrated modules are implemented in the form of software functional modules and sold or used as independent products, they may also be stored in a computer-readable storage medium.

[0065] Although the embodiments of the present invention have been shown and described above, it will be understood that the above embodiments are illustrative and are not to be construed as limitations on the present invention. A person skilled in the art may change, modify, replace and modify the above embodiments within the scope of the present invention.

Claims

1. A fire monitoring method based on vision, characterized by: This involves deploying several cameras at the monitoring location, with each camera capturing a single monitoring area; All cameras communicate with the fire prevention and control center platform via data cables; Establish a background model of the monitoring location on the fire prevention and control center platform; According to the images captured by the camera, obtain the proportion of flammable items in the image; Determine fire-prone areas and, based on the determination results, determine whether the monitored area is fire-prone or fire-unprone; Add the monitoring area of ​​each camera to the background model; When a fire occurs in a certain monitoring area, the camera determines the fire image based on the deep neural network model; The fire prevention and control center platform displays the fire image and camera number information, and simultaneously identifies the monitoring area where the fire occurred in the background model and issues a fire alarm; The fire prevention and control center platform simulates the fire development trend in the background model; The specific steps include: Step 1: Analyze the image captured by the camera and obtain all flammable items in the image through a deep neural network model; Step 2: On the screen, mark each flammable item with a box; Step 3: Calculate the area ratio of the box in the picture. When it reaches the preset value, the area captured by the camera is judged as a fire-prone area; otherwise, it is judged as a fire-prone area; According to the methods of steps 1 and 2, the monitoring area captured by each camera is judged and divided into fire-prone areas and fire-unprone areas; When simulating the fire development trend in the background model, the specific steps include: Step A1: calibrate the monitoring area where the fire occurs in the background model, and set the monitoring area as monitoring area A; Step A2: Acquire a monitoring area B adjacent to the monitoring area A; Step A3: Determine whether the monitored area B is a fire-prone area: if yes, determine that the fire is spreading to the monitored area B; if no, determine that the fire is not spreading to the monitored area B; Step A4: According to the method of step A2 and step A3, all monitoring areas adjacent to monitoring area A are judged; Step A5: Based on the result of step A4, a fire development trend diagram is obtained and calibrated in the background model.

2. The visual-based fire monitoring method according to claim 1, wherein: The fire prevention and control center platform deploys a distributed server cluster or only deploys a separate central server, and the camera communicates with the distributed server cluster or central server through digital display.

Images