AI multi-mode charging shed fire control method and system

Abstract

The invention discloses an AI multi-mode charging shed fire management and control method and system. The method comprises the following steps that a video image of a charging shed is collected through a flame recognition bolt; and analyzing and processing the video image, determining whether a fire occurs or not, if so, sending fire information to a management platform, and executing a corresponding fire processing strategy by the management platform according to the fire information. Compared with the prior art, the system can achieve the timely early warning and extinguishment of a fire, and improves the fire response speed.

Description

Technical Field

[0001] The present invention relates to the field of fire protection technology, and in particular to an AI multimodal charging shed fire control method and system. Background Art

[0002] With the development of society, electric bicycles are widely distributed in cities and are an important part of urban public transportation.

[0003] Electric bicycles have many fire risks. If unqualified batteries are used, or if they are charged for a long time or charged indoors with a power supply, fires may occur. Electric bicycles can enter a fierce combustion stage within 12 seconds, and the flame temperature can reach 1200 degrees in 3 minutes. At the same time, toxic gases and thick smoke spread at a speed of 1 meter per second, which can easily cause poisoning and suffocation to people, posing a threat to people's property and life safety.

[0004] However, there is currently a lack of timely and effective fire warning and handling mechanisms, that is, fires are usually extinguished manually by spraying fire extinguishers or fire sprinklers. This method has a slow response and is prone to delays, resulting in greater economic losses. Summary of the Invention

[0005] The main purpose of this invention is to propose an AI multimodal charging shed fire control method and system, aiming to achieve timely early warning and fire extinguishing of fires and improve response speed.

[0006] To achieve the above objectives, the present invention provides an AI multimodal charging shed fire control method, which includes the following steps:

[0007] Step S10, collecting video images of the charging shed through a flame recognition gun;

[0008] Step S20, analyzing and processing the video image to determine whether a fire occurs, and if so, executing step S30;

[0009] Step S30: Send the fire information to the management platform, and the management platform executes the corresponding fire handling strategy according to the fire information.

[0010] A further technical solution of the present invention is that the step S20 of analyzing and processing the video image to determine whether a fire has occurred includes:

[0011] Step S201 : determining whether pixels of the video image are in a flame area based on an RGB color space analysis method or an HSV color space analysis method; if so, determining that a fire has occurred.

[0012] A further technical solution of the present invention is that, in step S201, the step of determining whether the pixel of the video image is in the flame area based on the RGB color space analysis method includes:

[0013] The red component R, green component G, and blue component B of the pixel of the video image are compared. If they all satisfy equation (1), it is determined that the pixel of the video image is in the flame area:

[0014]

[0015] Among them, R, B, G are red, blue, and green color components respectively, S t is the saturation threshold of the pixel, R t ∈[115, 135], S t ∈[45, 60], G th ∈[235, 250], R th ∈[235, 250], Th1∈[10, 30], Th2∈[10, 30], Th3∈[10, 20], T1=|RB|, T2=|GB|, T3=|RG|.

[0016] A further technical solution of the present invention is that, in step S201, the step of determining whether the pixel of the video image is in the flame area based on the HSV color space analysis method includes:

[0017] Compare the hue H of the video image with the preset threshold range, the saturation S with the preset second preset threshold, and the brightness V with the preset third preset threshold respectively. If Hmin is satisfied at the same time,<H<Hmax,S> Smin, V>Vmin, where Hmin, Hmax, Smin, and Vmin are set thresholds, and it is determined that the pixel of the video image is in the flame area.

[0018] A further technical solution of the present invention is that the step S20 of analyzing and processing the video image to determine whether a fire has occurred includes:

[0019] The spatial correlation between the gray values ​​in the video image is described based on the gray level co-occurrence matrix, and the texture characteristics of the flame are analyzed.

[0020] A further technical solution of the present invention is that the step S20 of analyzing and processing the video image to determine whether a fire has occurred includes:

[0021] The area of ​​each pixel in the video image is compared with the central pixel based on the local binary pattern, and a binary code is generated according to the comparison result to describe the texture feature of the image.

[0022] A further technical solution of the present invention is that the step S20 of analyzing and processing the video image to determine whether a fire has occurred includes:

[0023] The movement speed and direction of the pixels in the video image are calculated based on the optical flow method, and the movement state of the flame is determined according to the characteristics of the optical flow field.

[0024] A further technical solution of the present invention is that the step S20 of analyzing and processing the video image to determine whether a fire has occurred includes:

[0025] The texture features of the image are described based on the deep learning algorithm, and the characteristic histogram of the flame is statistically calculated as the basis for flame recognition.

[0026] A further technical solution of the present invention is that the step S30 of sending the fire information to the management platform, and the management platform executing a corresponding fire handling strategy according to the fire information includes:

[0027] Activate the alarm of the charging shed, turn off the power with the power-off switch, activate the sprinkler valve to automatically extinguish the fire, and notify the backstage management staff.

[0028] To achieve the above-mentioned objectives, the present invention also proposes an AI multimodal charging shed fire control system, which includes a memory, a processor, and an AI multimodal charging shed fire control program stored on the processor. When the AI ​​multimodal charging shed fire control program is run by the processor, it executes the steps of the method described above.

[0029] The beneficial effects of the AI ​​multimodal charging shed fire control method and system of the present invention are:

[0030] Through the above technical solution, the present invention collects video images of the charging shed through a flame recognition gun; analyzes and processes the video images to determine whether a fire has occurred. If so, step S30 is executed; the fire information is sent to the management platform, and the management platform executes the corresponding fire handling strategy based on the fire information, which can provide timely warning and fire extinguishing, thereby improving the fire response speed. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the structures shown in these drawings without paying any creative work.

[0032] Figure 1 This is a flow chart of a preferred embodiment of the AI ​​multi-modal charging shed fire control method of the present invention;

[0033] Figure 2 This is a schematic diagram of the AI ​​multimodal charging shed fire control system.

[0034] The purpose, features and advantages of the present invention will be further described with reference to the accompanying drawings and in conjunction with the embodiments. DETAILED DESCRIPTION

[0035] 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. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0036] The present invention proposes an AI multimodal charging shed fire control method. The present invention can be applied to an AI multimodal charging shed fire control system, which includes a sprinkler valve, an audible and visual alarm, and a flame recognition gun arranged in the charging shed.

[0037] The preferred embodiment of the AI ​​multimodal charging shed fire control method of the present invention includes the following steps:

[0038] Step S10: Capture video images of the charging shed through a flame recognition gun.

[0039] In this embodiment, the flame recognition gun can be used to collect video images of the charging shed. In order to improve the accuracy of collecting whether a fire has occurred, a smoke sensor or the like can also be used for detection.

[0040] Step S20: Analyze and process the video image to determine whether a fire occurs. If so, execute step S30.

[0041] After the flame recognition gun collects the video image of the charging shed, the video image is sent to the management platform server, and the AI ​​algorithm is used to analyze and process the video image to determine whether a fire has occurred.

[0042] Step S30: Send the fire information to the management platform, and the management platform executes the corresponding fire handling strategy according to the fire information.

[0043] Among them, the processing strategy is, for example, to activate the alarm of the charging shed, turn off the power through the power-off switch, and then start the sprinkler water valve to automatically extinguish the fire. After the flame is extinguished, spray for another 5 minutes. After stopping the spraying, notify the management staff to go to the site to start the power.

[0044] In this embodiment, different AI algorithms can be used to process video images to determine whether a fire has occurred. The details are as follows.

[0045] As an implementation scheme, in this embodiment, a color feature algorithm may be used in step S20 to analyze and process the video image to determine whether a fire has occurred.

[0046] The steps of analyzing and processing the video image using a color feature algorithm to determine whether a fire has occurred include:

[0047] Step S201 : determining whether pixels of the video image are in a flame area based on an RGB color space analysis method or an HSV color space analysis method; if so, determining that a fire has occurred.

[0048] Specifically, in step S201, the step of determining whether the pixel of the video image is in the flame area based on the RGB color space analysis method includes:

[0049] The red component R, green component G, and blue component B of the pixel of the video image are compared. If they all satisfy equation (1), it is determined that the pixel of the video image is in the flame area:

[0050]

[0051] Among them, R, B, G are red, blue, and green color components respectively, S t is the saturation threshold of the pixel, R t ∈[115, 135], S t ∈[45, 60], G th ∈[235, 250], R th ∈[235, 250], Th1∈[10, 30], Th2∈[10, 30], Th3∈[10, 20], T1=|RB|, T2=|GB|, T3=|R-G1.

[0052] Flames are affected by temperature, combustion materials, and the degree of combustion. Flames have a special color distribution pattern, showing significant and bright red and yellow. In RGB and HIS space, each component has a specific relationship. In the prior art, some scholars use RGB and HIS color models to perform dynamic analysis on flames and summarize three expressions, as shown in formula (2):

[0053]

[0054] In formula (2), R, B, and G are red, blue, and green color components respectively, and S t is the saturation threshold of the pixel, R t ∈[115, 135], S t∈[45, 60]. However, this method is prone to causing the internal space of the flame, and is prone to misjudging withered grass, thick smoke produced by burning flames, etc. as flames. Therefore, this embodiment improves this method, summarizes five rules, and establishes a new flame color model, as shown in formula (1).

[0055] In step S201, the step of determining whether the pixel of the video image is in the flame area based on the HSV color space analysis method includes:

[0056] The hue H of the video image is compared with a preset threshold range, the saturation S is compared with a preset second preset threshold, and the brightness V is compared with a preset third preset threshold. If Hmin<H<Hmax, S>Smin, V>Vmin are satisfied at the same time, where Hmin, Hmax, Smin, and Vmin are the set thresholds, it is determined that the pixel of the video image is in the flame area.

[0057] In the HSV color space, the hue of a flame typically falls within a certain range, and its saturation (S) and lightness (V) also have specific values. Generally speaking, the hue (H) of a flame might range from 0 to 30 degrees, with saturation (S) exceeding a certain threshold and lightness (V) also high.

[0058] As a second implementation scheme, in this embodiment, in step S20, a texture feature-based algorithm may be used to analyze and process the video image to determine whether a fire has occurred.

[0059] The steps of analyzing and processing the video image using a texture feature algorithm to determine whether a fire has occurred include:

[0060] The spatial correlation between gray values ​​in the video image is described based on the gray level co-occurrence matrix (GLCM), and the texture characteristics of the flame are analyzed.

[0061] This embodiment analyzes the texture characteristics of the flame by calculating some characteristic parameters of the gray level co-occurrence matrix (GLCM), such as energy E, contrast CON, and entropy ENT.

[0062] Where, energy E = ∑ i ∑ j p(i, j) 2 , contrast CON = ∑ i ∑ j (ij) 2 p(i, j), entropy ENT = -∑ i ∑ j p(i, j)log p(i, j), where p(i, j) is the probability of gray levels i and j appearing simultaneously in the gray level co-occurrence matrix, μ i 、μ jis the mean of i and j, σ i , σ j is the standard deviation of i and j. These texture feature parameters of flame images are usually significantly different from those of other background objects.

[0063] The step of analyzing and processing the video image using a texture feature algorithm to determine whether a fire has occurred can also be based on a local binary pattern (LBP) to compare the area of ​​each pixel in the video image with the central pixel, and generate a binary code based on the comparison result to describe the texture characteristics of the image.

[0064] Specifically, for a central pixel g in the video image c , its neighborhood pixel g p , p=0,1,...,P-1, P is the number of neighborhood pixels, and the LBP encoding formula is: in, The LBP feature histogram of the flame image can be used as a basis for flame recognition.

[0065] As a third implementation scheme, in this embodiment, in step S20, the movement speed and direction of the pixel points in the video image can be calculated based on the optical flow method, and the movement state of the flame can be determined according to the characteristics of the optical flow field.

[0066] Specifically, in the optical flow algorithm based on the constant brightness assumption, I(x, y, t) = I(x + Δ x，y +Δ y，t +Δ t ), perform Taylor expansion on it and ignore the higher-order terms, and we can get I x u+I y v+I t =0, where I x , I y , I t are the partial derivatives of image I with respect to x, y, and t, respectively. is the component of the optical flow in the x and y directions. The optical flow field of a flame usually has a unique distribution and variation pattern, which is significantly different from the optical flow of a static background or other stable objects.

[0067] As a fourth implementation scheme, in this embodiment, in step S20, the texture features of the image can be described based on a deep learning algorithm, and the characteristic histogram of the flame can be statistically calculated as a basis for flame recognition.

[0068] Taking the classic AlexNet as an example, the convolutional neural network (CNN) consists of multiple convolutional layers, pooling layers, and fully connected layers. The clipping layer extracts features by convolving the kernel with the input image, such as in, is the convolution kernel weight from layer l-1 to layer l, is the input image element of the l-1th layer, b l is the bias term. The pooling layer downsamples the output of the convolutional layer, such as the maximum pooling operation Where s is the pooling step size. The fully connected layer classifies the output of the pooling layer and calculates the probability of the flame category through the softmax function. Among them, z k is the value of the kth node output by the fully connected layer, and C is the total number of categories.

[0069] Furthermore, in this embodiment, the step S30 of sending the fire information to the management platform, and the management platform executing the corresponding fire handling strategy according to the fire information includes:

[0070] Activate the alarm of the charging shed, turn off the power with the power-off switch, activate the sprinkler valve to automatically extinguish the fire, and notify the backstage management staff.

[0071] Specifically, in this embodiment, after sending the fire information to the management platform, the management platform can automatically activate the alarm of the charging shed, and start the sprinkler valve to automatically extinguish the fire after the power is turned off by the power switch. After the flame is extinguished, the water valve will continue to spray for 5 minutes. After the spraying stops, the management personnel will be notified to go to the scene to start the power supply, thereby achieving timely warning and fire extinguishing of the fire and improving the response speed.

[0072] To sum up, the AI ​​multimodal charging shed fire control method of the present invention uses the above-mentioned technical solution to collect video images of the charging shed through a flame recognition gun; analyze and process the video images to determine whether a fire has occurred. If so, execute step S30; send the fire information to the management platform, and the management platform executes the corresponding fire handling strategy according to the fire information, which can provide timely warning and fire extinguishing, thereby improving the fire response speed.

[0073] To achieve the above objectives, the present invention also proposes an AI multi-modal charging shed fire control system, such as Figure 2 As shown, the system includes a flame recognition gun arranged in the charging shed, a management platform server, a sprinkler valve, a memory, a processor, and an AI multimodal charging shed fire control program stored on the processor. The AI ​​multimodal charging shed fire control program is executed by the processor when it is run as described in the above embodiment, which will not be repeated here.

[0074] The above description is only a preferred embodiment of the present invention and does not limit the patent scope of the present invention. All equivalent structural transformations made by using the contents of the present invention description and drawings under the concept of the present invention, or direct / indirect application in other related technical fields are included in the patent protection scope of the present invention.

Claims

1. An AI multimodal charging shed fire control method, characterized in that: The method comprises the following steps: Step S10, collecting video images of the charging shed through a flame recognition gun; Step S20, analyzing and processing the video image to determine whether a fire occurs, and if so, executing step S30; Step S30: Send the fire information to the management platform, and the management platform executes the corresponding fire handling strategy according to the fire information.

2. The AI ​​multimodal charging shed fire control method according to claim 1 is characterized in that: The step S20 of analyzing and processing the video image to determine whether a fire has occurred includes: Step S201 : determining whether pixels of the video image are in a flame area based on an RGB color space analysis method or an HSV color space analysis method; if so, determining that a fire has occurred.

3. The AI ​​multimodal charging shed fire control method according to claim 2 is characterized in that: In step S201, the step of determining whether the pixel of the video image is in the flame area based on the RGB color space analysis method includes: The red component R, green component G, and blue component B of the pixel of the video image are compared. If they all satisfy equation (1), it is determined that the pixel of the video image is in the flame area: Among them, R, B, G are red, blue, and green color components respectively, S t is the saturation threshold of the pixel, R t ∈[115,135],S t ∈[45, 60], G th ∈[235, 250], R th ∈[235, 250], Th1∈[10, 30], Th2∈[10, 30], Th3∈[10, 20], T1=|RB|, T2=|GB|, T3=|RG|.

4. The AI ​​multimodal charging shed fire control method according to claim 2 is characterized in that: In step S201, the step of determining whether the pixel of the video image is in the flame area based on the HSV color space analysis method includes: The hue H of the video image is compared with a preset threshold range, the saturation S is compared with a preset second preset threshold, and the brightness V is compared with a preset third preset threshold. If Hmin<H<Hmax, S>Smin, V>Vmin are satisfied at the same time, where Hmin, Hmax, Smin, and Vmin are the set thresholds, it is determined that the pixel of the video image is in the flame area.

5. The AI ​​multimodal charging shed fire control method according to claim 1 is characterized in that: The step S20 of analyzing and processing the video image to determine whether a fire has occurred includes: The spatial correlation between the gray values ​​in the video image is described based on the gray level co-occurrence matrix, and the texture characteristics of the flame are analyzed.

6. The AI ​​multimodal charging shed fire control method according to claim 1 is characterized in that: The step S20 of analyzing and processing the video image to determine whether a fire has occurred includes: The area of ​​each pixel in the video image is compared with the central pixel based on the local binary pattern, and a binary code is generated according to the comparison result to describe the texture feature of the image.

7. The AI ​​multimodal charging shed fire control method according to claim 1 is characterized in that: The step S20 of analyzing and processing the video image to determine whether a fire has occurred includes: The movement speed and direction of the pixels in the video image are calculated based on the optical flow method, and the movement state of the flame is determined according to the characteristics of the optical flow field.

8. The AI ​​multimodal charging shed fire control method according to claim 1 is characterized in that: The step S20 of analyzing and processing the video image to determine whether a fire has occurred includes: The texture features of the image are described based on the deep learning algorithm, and the characteristic histogram of the flame is statistically calculated as the basis for flame recognition.

9. The AI ​​multimodal charging shed fire control method according to any one of claims 1 to 8, characterized in that: The step S30 of sending the fire information to the management platform, and having the management platform execute a corresponding fire handling strategy based on the fire information, includes: Activate the alarm of the charging shed, turn off the power with the power-off switch, activate the sprinkler valve to automatically extinguish the fire, and notify the backstage management staff.

10. An AI multi-modal charging shed fire control system, characterized by: The system includes a memory, a processor, and an AI multimodal charging shed fire management and control program stored on the processor. When the AI ​​multimodal charging shed fire management and control program is run by the processor, the steps of the method described in any one of claims 1 to 9 are executed.

Images