A flame identification method and device

Abstract

The present application provides a flame recognition method, which is applied to a flame recognition device, the flame recognition device comprising: a binocular camera, the binocular camera comprising an infrared channel and a visible light channel, an infrared filter is installed on the infrared channel, the flame recognition method comprises: using the infrared channel to obtain a first image, and determining the initial flame position based on the first image; using the visible light channel to obtain a second image, and determining the flame area based on the initial flame position and the second image; using a neural network model to identify the flame area, and determine whether the flame area is on fire. The present application realizes accurate identification of flames in the monitored area, reduces false alarms and missed alarms, and is conducive to firefighters to deal with fire crises in the first time.

Description

Technical Field

[0001] The invention belongs to the field of flame identification, and in particular relates to a method and a device for flame identification. Background Art

[0002] In the field of fire detection, most traditional smoke particle sensing systems require smoke particles to enter the sensor to trigger an alarm, and infrared and laser technologies also require smoke to block the alarm. These prerequisites require the venue to be a relatively closed space. In outdoor venues such as refineries and warehouses, traditional pyrotechnic equipment is completely ineffective because of the scattered equipment and facilities and large air flow. Now, people are often on duty to supervise, resulting in increased management costs.

[0003] The current indoor and outdoor flame automatic detection and warning cameras based on video analysis automatically detect abnormal fire signs in the monitoring area, issue alarms and assist firefighters in handling fire crises in the fastest and best way, and minimize false alarms and missed alarms. At the same time, you can also view real-time images on the scene and directly command and dispatch firefighting based on intuitive images.

[0004] However, at present, most fire monitoring devices can only monitor the occurrence of fires, but cannot determine the physical location of the fire point, which brings a certain degree of blindness to fire rescue. Summary of the invention

[0005] In view of this, the present invention provides a method and device for flame identification, which solves the problem of being unable to determine the location of the ignition point and realizes accurate identification of the flame in the monitoring area.

[0006] In a first aspect, a flame recognition method is provided, which is applied to a flame recognition device. The flame recognition device includes: a binocular camera, the binocular camera includes an infrared channel and a visible light channel, and an infrared filter is installed on the infrared channel. The flame recognition method includes: acquiring a first image using the infrared channel, and determining an initial flame position based on the first image; acquiring a second image using the visible light channel, and determining a flame area based on the initial flame position and the second image, wherein the flame area is located on the second image; and identifying the flame area using a neural network model to determine whether the flame area is actually on fire.

[0007] The flame identification method provided in the first aspect obtains the first image and the second image through the infrared channel and the visible light channel of the binocular camera respectively. Among them, the infrared spectrum image is obtained by installing an infrared filter on the infrared channel to obtain the initial position of the flame. This infrared spectrum image obtained by ordinary lenses and sensors reduces the cost of camera lenses. The flame area is determined by using the second image obtained on the visible light channel and the initial position of the flame. According to the image of the flame area, the flame is further identified using a neural network model. In this implementation method, accurate identification of the flame in the monitored area is achieved at a relatively low cost, and the phenomenon of false alarms and missed alarms is reduced, which is conducive to firefighters dealing with fire crises in the first place.

[0008] In a possible implementation method of the first aspect, before determining the flame region based on the initial flame position and the second image, the method further includes: converting the initial flame position to the second image by coordinates. In this implementation, by converting the position on the first image to the second image by coordinates, and determining the flame region on the second image using the position and the second image, accurate positioning of the flame region is achieved.

[0009] In a possible implementation method of the first aspect, a second image is acquired using a visible light channel, and a flame region is determined based on an initial flame position and the second image, including: determining three primary color components of each pixel in the second image based on the second image, the three primary color components including a red component, a green component, and a blue component; and determining the flame region based on the three primary color components, the initial flame position, and a preset first threshold condition of the three primary color components. In this implementation, the pixel region of the flame is finally found by determining the three primary color components of each pixel of the second image and combining the preset thresholds of the three primary color components.

[0010] In a possible implementation method of the first aspect, a neural network model is used to identify a flame area to determine whether the flame area is on fire, including: extracting motion features and contour features of the flame; generating a feature code based on the motion features and the contour features; and identifying the flame based on the feature code and a second threshold condition. In this implementation method, deep learning neural network technology is used to collect sample data, perform model training, and extract motion features and contour features of the flame in combination with a deep learning algorithm to further determine whether there is really a flame in the video.

[0011] In a possible implementation method of the first aspect, there is a first reference object on the first image, and there is a second reference object on the second image, and the initial flame position is converted to the second image through coordinates, including: calibrating the first reference object and the second reference object, and converting the initial flame position to the second image through coordinates. In this implementation method, by placing reference objects on the two images respectively, calibrating the same overlapping area, and transferring the position information on the first image to the second image, the initial flame position and the flame area are unified, and the accuracy of the flame area judgment is increased.

[0012] In a second aspect, a flame identification device is provided, the flame identification device comprising a binocular camera, the binocular camera comprising an infrared channel and a visible light channel, an infrared filter is installed on the infrared channel. The device is used to execute the above first aspect or any possible implementation method of the first aspect.

[0013] In a third aspect, a flame identification device is provided, the device comprising at least one processor and a memory, the at least one processor being configured to execute the method in the above first aspect or any possible implementation of the first aspect.

[0014] In a fourth aspect, a flame identification device is provided, the flame identification device comprising at least one processor and an interface circuit, the at least one processor being used to execute the method in the above first aspect or any possible implementation of the first aspect.

[0015] In a fifth aspect, a flame recognition device is provided, and the flame recognition processing device includes any flame recognition device provided in the second aspect, the third aspect or the fourth aspect.

[0016] According to a sixth aspect, a computer program product is provided. The computer program product includes a computer program. When the computer program is executed by a processor, the computer program is used to execute the method in the first aspect or any possible implementation of the first aspect.

[0017] In a seventh aspect, a computer-readable storage medium is provided, in which a computer program is stored. When the computer program is executed, it is used to execute the method in the first aspect or any possible implementation of the first aspect.

[0018] In an eighth aspect, a chip or an integrated circuit is provided, which includes: a processor, configured to call and run a computer program from a memory, so that a device equipped with the chip or the integrated circuit executes the method in the first aspect or any possible implementation of the first aspect.

[0019] The technical effects of the device provided in the present application can refer to the technical effects of the above-mentioned first aspect or each implementation method of the first aspect, and will not be repeated here.

[0020] Compared with the prior art, the present invention has the following beneficial effects:

[0021] The present application provides a method and device for flame recognition, which respectively obtains a first image and a second image through the infrared channel and the visible light channel of a binocular camera. Among them, an infrared spectrum image is obtained by installing an infrared filter on the infrared channel to obtain the initial position of the flame. This infrared spectrum image obtained by ordinary lenses and sensors reduces the cost of camera lenses. The flame area is determined by using the second image obtained on the visible light channel and the initial position of the flame. According to the image of the flame area, the flame is further identified using a neural network model. In this implementation method, accurate identification of the flame in the monitored area is achieved at a relatively low cost, reducing false alarms and missed alarms, which is conducive to firefighters dealing with fire crises in the first place. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0023] Figure 1 A schematic diagram of the implementation process of the flame identification method 100 provided in an embodiment of the present application is shown;

[0024] Figure 2 A schematic block diagram of a flame identification device 200 according to an embodiment of the present application is shown;

[0025] Figure 3 A schematic block diagram of a flame identification device 300 according to an embodiment of the present application is shown. DETAILED DESCRIPTION

[0026] Fires cause great losses to people's lives and property safety. Since the flames of fires generally have obvious visual characteristics, most of them currently use high-definition cameras and flame detection to achieve the purpose of flame recognition by identifying flame characteristics. It has non-contact characteristics, which is different from traditional contact detection technology. And because it detects quickly and the image information is intuitive and rich, image-based fire detection has become a powerful helper for humans to detect fires in large spaces and open spaces such as warehouses, large factories, and wild forests. However, this type of image-based fire detector also has shortcomings: similar to the need to solve the problem of whether there is interference near the sensor, the interference of non-contact fire detectors will come from a much larger range, and it is impossible to accurately and efficiently obtain flame recognition, and objects with shapes similar to flames cannot be eliminated. Therefore, the development of a new flame recognition method has become a problem that people are eager to solve.

[0027] In view of the above problems, the present invention provides a method and device for flame recognition, which respectively obtains a first image and a second image through the infrared channel and the visible light channel of a binocular camera. The infrared spectrum image is obtained by installing an infrared filter on the infrared channel to obtain the initial position of the flame. The flame area is determined by using the second image obtained on the visible light channel and the initial position of the flame. According to the image of the flame area, the flame is further identified using a neural network model. In this implementation method, accurate identification of the flame in the monitored area is achieved at a relatively low cost, and the phenomenon of false alarms and missed alarms is reduced, which is conducive to firefighters dealing with fire crises in the first place.

[0028] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0029] First of all, before introducing the embodiments of the method and device for flame identification provided by the present application, it is necessary to explain some of the terms that will be mentioned below. When the volume terms "first", "second" and other ordinal numbers in the present application are used only for descriptive purposes, they cannot be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" and "second" can explicitly or implicitly include one or more of the features. In the description of the embodiments of the present application, unless otherwise specified, the meaning of "multiple" is two or more.

[0030] Embodiments of the present invention are described in detail below, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and cannot be understood as limiting the present invention.

[0031] These and other aspects of the embodiments of the present invention will become clear with reference to the following description and drawings. In these descriptions and drawings, some specific implementations of the embodiments of the present invention are specifically disclosed to represent some ways of implementing the principles of the embodiments of the present invention, but it should be understood that the scope of the embodiments of the present invention is not limited thereto. On the contrary, the embodiments of the present invention include all changes, modifications and equivalents that fall within the spirit and connotation of the appended claims.

[0032] Secondly, the flame recognition method provided in this application is applicable to a flame recognition device with a binocular camera. Now, monitoring binocular cameras have appeared in the security market. They are new products developed to meet the monitoring needs in different monitoring environments. They use the analysis of pixel points to identify the monitoring scene, and then use new scientific methods to restore the monitoring scene. And compared with ordinary cameras, it overcomes the influence of light intensity and can restore the quality of monitoring more realistically. In this application, the binocular camera includes an infrared channel and a visible light channel, wherein an infrared lens is not installed on the infrared channel, but an ordinary lens is used. By installing an infrared filter between the ordinary lens and the sensor to obtain an infrared spectrum image, the effect consistent with the infrared lens is achieved.

[0033] The flame recognition method provided in the present application is exemplarily described below in conjunction with the illustrated embodiments.

[0034] See also Figure 1 , is a flow chart of an embodiment of a flame identification method provided by the present application. Figure 1 As shown, the method 100 includes S101 to S103.

[0035] S101, acquiring a first image using an infrared channel, and determining an initial flame position based on the first image;

[0036] By adding an infrared filter to the infrared channel of the binocular camera, infrared light that is invisible to the naked eye can be used to illuminate the object being photographed. The infrared light is diffusely reflected and received by the surveillance camera to form a first image. By processing the first image in black and white, it can be determined that the white area is the initial flame position.

[0037] Understandably, considering that the temperature of the flame determines the color of the flame, the flame is burning, and at low temperatures it is infrared. As the temperature rises, the flame changes color from red and orange (3000 degrees) to yellow and white (4000 degrees) to cyan and blue (5000-6000 degrees) to purple (above 7000) to the invisible ultraviolet (tens of thousands of degrees). From the perspective of high-energy physics, infrared and colored spectrum flames are low-energy flames. As the temperature continues to rise, the color of the flame changes from ultraviolet to x-rays to gamma rays, and so on.

[0038] Optionally, as a possible implementation method of the present application, in the infrared channel, an infrared filter is added between the visible light lens and the sensor to allow only the infrared spectrum to enter the sensor as a black and white image. At the location where the flame is burning, the infrared channel video will appear white, and other places will be black. The higher the temperature, the whiter the location. The location of the burning flame can be obtained by detecting the horizontal brightness value.

[0039] S102: Acquire a second image using a visible light channel, and determine a flame area based on the initial flame position and the second image.

[0040] The second image is obtained according to the determined initial flame position and the visible light channel, and the second image is processed to obtain the three primary color component values, and the area satisfying the flame pixel is obtained through the three primary color component values. Then, the flame area is determined in combination with the initial flame position.

[0041] Optionally, as a possible implementation of step S102, before processing the second image, it also includes a process of converting the initial flame position determined by the first image to the second image: considering that the field of view of the dual-channel video on the flame recognition device is different, it is necessary to calibrate the field of view of the two channels. By pre-placing two reference objects on the first image and the second image respectively acquired by the binocular camera, and then drawing rectangles with the two reference objects as diagonals in the two videos, it is equivalent to marking the overlapping parts in the two videos, and establishing the coordinate transformation relationship of the overlapping area of ​​the two channels by linear calculation. The preliminary flame position on the infrared channel can be marked on the visible light channel. By converting the position on the first image to the second image through coordinates, the flame area is determined on the second image using the position and the second image, and the accurate positioning of the flame area is achieved.

[0042] Optionally, as a possible implementation manner, the second image acquired by the visible light channel is processed to determine three primary color components, namely, a red (R) component, a green (G) component, and a blue (B) component.

[0043] It is understandable that for flames, the red component (R) and green component (G) of the RGB model will be very large, and the green component (G) will be greater than the blue component (B). The thresholds of the red component (R) and the green component (G) are set, and the visible light channel separates the image into three RGB channels. The red component (R) and the green component (G) need to be greater than the corresponding thresholds. The specific thresholds are not limited in this application. When the R, G, and B of each pixel in the image meet the above threshold relationship, it can be determined to be a flame pixel. The set of these flame pixels and the flame position preliminarily determined above jointly determine the flame area. By judging the three primary color components of each pixel in the flame area, the accuracy of the flame area judgment is increased.

[0044] S103: Using a neural network model to identify the flame area, and determining whether the flame area is on fire.

[0045] By using the changes in the shape of the flame when it burns, deep learning neural network technology is used to detect flames in the visible light channel. By default, there is one or several burning points in the flame, and the flame spreads outward layer by layer from these burning points. The shape of the flame is more variable and continuous as it gets to the outer layer. Based on the large amount of flame sample data collected, deep learning neural network technology is used to collect sample data, train the model, and combine it with deep learning algorithms to detect whether there is a flame in the video. Using the characteristics of the movement of the flame when it burns, deep learning neural network technology is used to identify flame similarity in the visible light channel.

[0046] Optionally, as a possible implementation of the present application, considering that the flame movement, energy and brightness changes are uneven when the flame burns, key points such as motion features and contour features are extracted for each detected flame area, and a feature code is generated through key points of the same detection area in adjacent frames. The generated feature code is calculated by difference to obtain a similarity value. If the similarity is higher than a certain threshold, it can be determined that a real flame exists.

[0047] In this embodiment, the first image and the second image are obtained through the infrared channel and the visible light channel of the binocular camera, respectively. Among them, an infrared spectrum image is obtained by installing an infrared filter on the infrared channel to obtain the initial position of the flame. This infrared spectrum image obtained by ordinary lenses and sensors reduces the cost of camera lenses. The flame area is determined by using the second image obtained on the visible light channel and the initial position of the flame. According to the image of the flame area, the neural network model is used to further identify the flame. In this implementation method, accurate identification of the flame in the monitored area is achieved at a relatively low cost, and the phenomenon of false alarms and missed alarms is reduced, which is conducive to firefighters to deal with fire crises in the first time.

[0048] An embodiment of the present application further provides a flame recognition device, which includes a binocular camera, the binocular camera includes an infrared channel and a visible light channel, and an infrared filter is installed on the infrared channel.

[0049] The flame identification device comprises a receiving unit and a processing unit. Figure 2 A schematic block diagram of a flame identification device 200 according to an embodiment of the present application is shown. Each unit of the flame identification device 200 is used to execute each action or processing performed by the flame identification device in each embodiment of the above-mentioned method 101 to method 103, such as Figure 2 As shown, the flame identification device 200 may include: a receiving unit 201 and a processing unit 202. The flame device 200 may also include a storage unit for storing instructions executed by the receiving unit 201 and the processing unit 202. The storage unit stores instructions, the processing unit 202 is used to execute the instructions stored in the storage unit, the receiving unit is used to receive the image storage command on the flame identification device, and the processing unit is used to process the flame identification command received above.

[0050] The storage unit may be a memory, and the processing unit 202 may be implemented by a processor. Figure 3 FIG. 3 is a schematic block diagram of a flame identification device 300 according to an embodiment of the present application. Figure 3 As shown, the flame identification device 300 may include a processor 301 and a memory 302, and the at least one processor is used to execute the flame identification method in the above method 100. The processor 301 is connected to a binocular camera 303, and is used to execute instructions stored in a binocular camera storage unit. The binocular camera 303 includes an infrared channel 3031 and a visible light channel 3032, and an ordinary camera is installed on the infrared channel, and an infrared filter 3033 is installed between the ordinary camera and the sensor.

[0051] The embodiment of the present application further provides a computer readable medium for storing computer program code, the computer program including instructions for executing the flame identification method of the embodiment of the present application in the above method 100. The embodiment of the present application does not limit the readable medium.

[0052] The present application also provides a computer program product, which includes instructions. When the instructions are executed, the flame identification device performs operations in the flame identification device corresponding to the above method 100.

[0053] The embodiment of the present application further provides a chip or integrated circuit, which includes: a processor, which is used to call and run a computer program from a memory, so that a device equipped with the chip or integrated circuit executes the flame identification method in the above method 100.

[0054] The embodiments described above are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, a person skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. Such modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application, and should all be included in the protection scope of the present application.

Claims

1. A flame identification method, characterized in that: Applied to a flame recognition device, the flame recognition device comprises: a binocular camera, the binocular camera comprises an infrared channel and a visible light channel, an infrared filter is installed on the infrared channel, and the method comprises: acquiring a first image using the infrared channel, and determining an initial flame position based on the first image; The second image is acquired by using the visible light channel, and the flame area is determined based on the initial flame position and the second image, and the flame area is located on the second image; the initial flame position determined by the first image is converted to the second image; two reference objects are placed in advance on the first image and the second image respectively acquired by the binocular camera, and then rectangles with the two reference objects as diagonals are drawn in the two videos, which is equivalent to marking the overlapping parts of the two videos, and a coordinate conversion relationship of the overlapping areas of the two channels is established according to linear calculation; the preliminary flame position on the infrared channel is marked on the visible light channel; the position on the first image is converted to the second image through coordinates, and the flame area is determined on the second image using the position and the second image, thereby achieving accurate positioning of the flame area; Using a neural network model to identify the flame area and determine whether the flame area is on fire includes: Extracting the motion characteristics and contour feature key points of the flame in the flame area; A feature code is generated by the key points of the same detection area of ​​several adjacent frames, and the difference calculation is performed on the generated feature code to obtain a similarity value; if the similarity is higher than a certain threshold, it is determined that there is a real flame.

2. The flame identification method according to claim 1, characterized in that: The method of acquiring a second image by using a visible light channel and determining a flame area based on the initial flame position and the second image includes: Determine three primary color components of each pixel in the second image based on the second image, the three primary color components including a red component, a green component, and a blue component; The flame area is determined based on the three primary color components, the initial flame position and a first threshold condition of the three primary color components.

3. A flame identification device, characterized in that include: A binocular camera, the binocular camera comprising an infrared channel and a visible light channel, an infrared filter being installed on the infrared channel, and the device being used to execute the flame recognition method as described in any one of claims 1 to 2.

4. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 2 are implemented.

5. A chip, characterized in that: include: A processor is used to call and run a computer program from a memory so that a flame identification device equipped with the chip executes the method according to any one of claims 1 to 2.

Images