Video real-time monitoring system for fire drill simulation
The fire identification system, which uses multi-dimensional feature analysis, solves the problem of poor identification results in fire simulation, and achieves accurate identification and alarm of fire risks, thereby improving safety assurance.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-12-10
- Publication Date
- 2026-04-03
AI Technical Summary
Existing technologies have poor fire identification performance in simulated fire scenarios and cannot adapt to environmental changes in different drill scenarios, resulting in insufficient accuracy in safety assurance and effect evaluation.
A multi-dimensional feature analysis method is adopted. Real-time image frames are acquired through the acquisition module. Color, grayscale and motion information are used to determine the degree of anomaly of each pixel. The overall anomaly degree is combined to identify risks.
It improves the accuracy of fire identification, reduces the occurrence of accidents, and enables precise assessment and real-time alarm of fire risks.
Smart Images
Figure CN121789129A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of computer vision technology, specifically to a real-time video monitoring system for fire drill simulation. Background Technology
[0002] In diverse scenarios such as fire drills, film and television shooting, and safety education, simulating real fire scenarios is a crucial element (e.g., smoke generated by smoke generators, stage fog machines, and simulated flame lighting). However, while simulated scenarios can provide an immersive experience, safety assurance and effectiveness evaluation cannot be ignored. To ensure the absolute safety of participants during the simulation, while comprehensively and accurately evaluating the drill's effectiveness and fully recording the entire process, a video surveillance system capable of real-time identification and alarm activation is indispensable.
[0003] Traditional methods typically filter pixels that match the characteristics of flames or smoke by setting fixed threshold ranges, and then calculate and judge features such as color, texture, and dynamic shape. While this can provide real-time fire warnings, the assembly-line processing means that if any link (such as color) fails to judge due to changes in ambient light, the entire chain will fail. At the same time, manually set thresholds (such as color range and area change thresholds) cannot adapt to different drill scenarios (indoor / outdoor, day / night, and simulated smoke of different colors), resulting in poor fire recognition performance. Summary of the Invention
[0004] In view of this, embodiments of the present invention provide a real-time video monitoring system for fire drill simulation to solve the problem of poor fire identification effect.
[0005] This invention provides a real-time video monitoring system for fire drill simulation, the system comprising: The acquisition module is used to acquire real-time image frames of the training ground; The first acquisition module is used to determine the first degree of abnormality of each pixel in the real-time image frame based on the color information of the real-time image frame. The second acquisition module is used to determine the second degree of abnormality of each pixel in the real-time image frame based on the grayscale information of the real-time image frame. The third acquisition module is used to determine the third degree of anomaly of each pixel in the real-time image frame based on the motion information of the real-time image frame. The identification module is used to determine the comprehensive anomaly level of each pixel in the real-time image frame based on the first anomaly level, the second anomaly level, and the third anomaly level, and to perform risk identification based on the comprehensive anomaly level.
[0006] In some embodiments, the first acquisition module determines the first anomaly level of each pixel in the real-time image frame based on the color information of the real-time image frame, including: The red (R) channel value, green (G) channel value, blue (B) channel value, hue (H) channel value, saturation (S) channel value, and brightness (V) channel value corresponding to each pixel of the real-time image frame are obtained as the color information of the real-time image frame; The flame characterization value of the pixel is determined based on the R channel value, G channel value, B channel value and H channel value of the pixel; The smoke characterization value of the pixel is determined based on the S-channel value and the V-channel value of the pixel; Based on the flame characterization value and the smoke characterization value, the first degree of anomaly of the pixel is determined.
[0007] In some embodiments, the first acquisition module is further configured to: Obtain the sum of the R channel value, G channel value, and B channel value corresponding to each pixel; Calculate the first difference between the R channel value and the G channel value for each pixel, and the second difference between the R channel value and the B channel value; Based on the first difference, the second difference, the summation result, and the H channel value, the flame characterization value of the corresponding pixel is determined; The smoke characterization value of the corresponding pixel is determined by multiplying the reciprocal of the S channel value and the V channel value. The maximum value is determined from the flame characterization value and the smoke characterization value, and is used as the first anomaly level of the corresponding pixel.
[0008] In some embodiments, the second acquisition module determines the second anomaly level of each pixel in the real-time image frame based on the grayscale information of the real-time image frame, including: The grayscale value of each pixel in the real-time image frame is obtained as the grayscale information of the real-time image frame; Obtain the neighborhood window of each pixel, wherein the neighborhood window contains neighboring pixels within a set range centered on the pixel; Obtain the local binary mode (LBP) value, grayscale entropy value, and gradient direction entropy value of each pixel within the neighborhood window; The texture feature value of the center pixel of the neighborhood window is determined based on the grayscale entropy value and the gradient direction entropy value. Obtain the gradient value of each pixel within the neighborhood window, as well as the maximum and minimum grayscale values of the pixels within the neighborhood window; The grayscale feature value of the center pixel of the neighborhood window is determined based on the gradient value, the maximum grayscale value, and the minimum grayscale value. Based on the LBP value of each pixel within the neighborhood window, determine the LBP standard deviation of the center pixel of the neighborhood window; Based on the LBP standard deviation, the texture feature value, and the grayscale feature value, the second anomaly degree of the center pixel of the neighborhood window is determined.
[0009] In some embodiments, the second acquisition module is further configured to: The product of the grayscale entropy value and the gradient direction entropy value is obtained as the texture feature value; Obtain the difference between the maximum gray value and the minimum gray value, and calculate the average gradient of the pixels within the neighborhood window; The grayscale feature value is determined based on the product of the difference and the mean gradient. Calculate the first product of the texture feature value, the grayscale feature value, and the LBP standard deviation, and determine the second anomaly degree of the center pixel of the neighborhood window based on the first product result.
[0010] In some embodiments, the third acquisition module determines the third anomaly level of each pixel in the real-time image frame based on the motion information of the real-time image frame, including: Obtain the previous image frame of the real-time image frame, and obtain the motion vector of each pixel in the real-time image frame as the motion information of the real-time image frame based on the previous image frame; Obtain the neighborhood window of each pixel, wherein the neighborhood window contains neighboring pixels within a set range centered on the pixel; Obtain the motion vector of each pixel within the neighborhood window and its similarity to the mean motion vector of all pixels within the neighborhood window, and determine the mean similarity value of the neighborhood window based on the similarity. Obtain the divergence of each pixel within the neighborhood window, and obtain the mean divergence of the neighborhood window based on the divergence. Obtain the area of the connected component of each pixel in the real-time image frame and the previous image frame respectively; The third anomaly level of the center pixel of the neighborhood window is determined based on the mean similarity, the mean divergence, and the area of the connected region.
[0011] In some embodiments, the third acquisition module is further configured to: Obtain the area difference between the area of the connected component of each pixel in the real-time image frame and the area of the connected component of the pixel in the previous image frame; Calculate the second product of the area difference and the mean divergence corresponding to the center pixel of the neighborhood window, and determine the third anomaly degree of the center pixel of the neighborhood window based on the second product and the mean similarity.
[0012] In some embodiments, the identification module determines the comprehensive anomaly level of each pixel in the real-time image frame based on the first anomaly level, the second anomaly level, and the third anomaly level, including: The average of the first abnormality level, the second abnormality level, and the third abnormality level of each pixel is obtained as the comprehensive abnormality level of the corresponding pixel.
[0013] In some embodiments, the identification module performs risk identification based on the overall anomaly level, including: Target pixels are selected from the real-time image frames based on the overall anomaly level. Based on the number of target pixels, the current fire risk information is determined, and alarm information is determined based on the fire risk information.
[0014] In some embodiments, the alarm information includes at least a fire area outline, and the identification module is further configured to: Each target pixel is used as a seed point for region generation. The growth criteria are based on the comprehensive anomaly degree of adjacent pixels and the gray-level similarity between the adjacent pixels and the seed point, to obtain the target growth region of the seed point, which serves as the outline of the fire area.
[0015] The beneficial effects of this invention compared to existing technologies are as follows: This system includes a data acquisition module, a first acquisition module, a second acquisition module, a third acquisition module, and an identification module. This embodiment utilizes the dynamic correlation between multi-dimensional features such as the color features, grayscale features, and motion features of flames and smoke to acquire the first, second, and third anomaly levels of each pixel in a real-time image frame. It performs comprehensive feature analysis on each pixel in the real-time image frame, thereby obtaining an accurate and comprehensive comprehensive anomaly level for each pixel. Based on the comprehensive anomaly level, risk identification is performed. Compared to the fixed rules and threshold adjustments in existing methods, this method acquires the anomaly level of pixels based on actual on-site characteristics, thereby performing risk identification, significantly improving the accuracy of fire risk identification during drills and reducing the occurrence of accidents. Attached Figure Description
[0016] To more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0017] Figure 1 This is a structural block diagram of a real-time video monitoring system for fire drill simulation provided in an embodiment of this application. Detailed Implementation
[0018] Embodiments of this disclosure are described in detail below, with examples of these embodiments illustrated in the accompanying drawings. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain this disclosure, and should not be construed as limiting it.
[0019] It should be noted that the terms "first," "second," etc., used in this disclosure and the accompanying drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this disclosure described herein can be implemented in orders other than those illustrated or described herein. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this disclosure. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this disclosure.
[0020] To illustrate the technical solution of the present invention, specific embodiments are described below.
[0021] Figure 1 This application provides a real-time video monitoring system for fire drill simulation. For example... Figure 1 As shown, the system specifically includes: The acquisition module 101 is used to acquire real-time image frames of the training ground.
[0022] Optionally, surveillance cameras can be deployed in the fire drill scenario to capture real-time image frames of the drill site.
[0023] The first acquisition module 102 is used to determine the first degree of abnormality of each pixel in the real-time image frame based on the color information of the real-time image frame.
[0024] Since flames and smoke in the environment at the time of a fire have color, texture and motion characteristics, the characteristics of flames and smoke can be analyzed separately to obtain a more accurate degree of anomaly.
[0025] It is understandable that the core area of a flame typically has the highest Red (R) component, followed by the Green (G) component, and the lowest Blue (B) component. Meanwhile, the flame hue (H) is concentrated in the low-value area (from red to yellow, between 0-60°). Therefore, the flame characterization value of a pixel can be determined by the RGB value and H value of each pixel.
[0026] It is understandable that smoke has a low saturation (S) and a high brightness (V), so the smoke characterization value of a pixel can be determined by the S channel value and V channel value of each pixel.
[0027] In this embodiment, the red (R) channel value, green (G) channel value, blue (B) channel value, hue (H) channel value, saturation (S) channel value, and brightness (V) channel value corresponding to each pixel of the real-time image frame are obtained as the color information of the real-time image frame.
[0028] Optionally, the flame characterization value of a pixel can be determined based on the R-channel, G-channel, B-channel, and H-channel values of that pixel. Specifically, the summation of the R-channel, G-channel, and B-channel values corresponding to each pixel can be obtained; a first difference between the R-channel and G-channel values, and a second difference between the R-channel and B-channel values corresponding to each pixel can be calculated; based on the first difference, the second difference, the summation result, and the H-channel value, the flame characterization value of the corresponding pixel can be determined, and the flame characterization value can be expressed as: ;
[0029] in, For the first The flame characterization value of each pixel; , and The first The R channel value, G channel value, and B channel value of each pixel. To calculate the summation result; This is the first difference; The second difference; For the first The flame index value of each pixel depends on the H channel value of the pixel. If the H channel value is between 0 and 60 degrees, the flame index value is 1; otherwise, if the H channel value is not between 0 and 60 degrees, the flame index value is 0.01. c is an arbitrary constant to prevent the denominator from being 0. In this embodiment, the value is 0.01.
[0030] In this embodiment, the flame characterization value This ensures the dominance of red, meaning that the larger the R channel value of the pixel, the more it matches the color characteristics of a flame; Used for normalization, making it insensitive to changes in brightness.
[0031] Optionally, the smoke characterization value of a pixel can be determined based on the S-channel and V-channel values of the pixel; specifically, the smoke characterization value of the corresponding pixel can be determined based on the product of the reciprocal of the S-channel value and the V-channel value, and the smoke characterization value can be expressed as: ; in, For the first The smoke characterization value of each pixel; For the first The S-channel value of each pixel; For the first The V channel value of each pixel.
[0032] In this embodiment, the lower the saturation and the higher the brightness of a pixel, the more it conforms to the characteristics of white and gray smoke areas, and the greater its corresponding smoke characterization value.
[0033] Furthermore, based on the flame characterization value and the smoke characterization value, the first anomaly level of the pixel is determined. In this embodiment, the maximum value between the flame characterization value and the smoke characterization value is determined as the first anomaly level of the corresponding pixel. The first anomaly level is then expressed as: ; That is to say: ; in, For the first The first anomaly level of each pixel; For the first The flame characterization value of each pixel; For the first The smoke representation value for each pixel; max is the maximum value.
[0034] The second acquisition module 103 is used to determine the second degree of abnormality of each pixel in the real-time image frame based on the grayscale information of the real-time image frame.
[0035] Since flames and smoke not only contain color features, but also have corresponding features during the motion changes of single-frame and multi-frame images, grayscale images reduce the amount of computation and make it easier to analyze images. Therefore, the grayscale value of each pixel in the real-time image frame can be obtained as the grayscale information of the real-time image frame.
[0036] It is understandable that smoke and flame regions are both turbulent regions, with high local texture complexity in a single frame image, corresponding to a high standard deviation of their Local Binary Pattern (LBP) values. Furthermore, considering the varying smoke density and occlusion levels in smoke regions, and the fact that flame colors generally vary with the shape of the burning material, the grayscale values within these regions are chaotic and irregularly arranged, resulting in high local grayscale entropy and gradient direction entropy. Since both flames and smoke have a certain degree of coverage, varying smoke concentrations or flame sizes can blur the edges of objects, leading to lower contrast and smaller grayscale differences within neighboring pixels. This allows for the analysis of the grayscale features of the pixels.
[0037] Optionally, a neighborhood window can be obtained for each pixel. The neighborhood window contains neighboring pixels within a set range centered on the pixel. In this embodiment, the neighborhood window is used as a unit of measurement. Construct pixels from the 7x7 neighborhood pixels at the center. i The neighborhood window.
[0038] Furthermore, the local binary mode LBP value, grayscale entropy value, and gradient direction entropy value of each pixel within the neighborhood window can be obtained; based on the grayscale entropy value and gradient direction entropy value, the texture feature value of the center pixel of the neighborhood window is determined. The entropy calculation is obtained using existing known techniques, which will not be elaborated further.
[0039] Optionally, the product of the grayscale entropy value and the gradient direction entropy value can be obtained as the texture feature value.
[0040] Furthermore, the gradient value of each pixel within the neighborhood window, as well as the maximum and minimum grayscale values of the pixels within the neighborhood window, can be obtained; based on the gradient value, the maximum grayscale value, and the minimum grayscale value, the grayscale feature value of the center pixel of the neighborhood window can be determined.
[0041] Optionally, the difference between the maximum and minimum gray values can be obtained, and the average gradient of pixels within the neighborhood window can be calculated; the gray feature value can be determined based on the product of the difference and the average gradient.
[0042] Furthermore, the LBP standard deviation of the center pixel of the neighborhood window can be determined based on the LBP value of each pixel within the neighborhood window; the second anomaly degree of the center pixel of the neighborhood window can be determined based on the LBP standard deviation, texture feature value, and grayscale feature value; in this embodiment, the corresponding LBP value is calculated using the 3x3 neighborhood pixels of each pixel.
[0043] Optionally, a first product of the texture feature value, grayscale feature value, and LBP standard deviation can be calculated, and a second anomaly degree of the center pixel of the neighborhood window can be determined based on the first product result. The second anomaly degree is expressed as: ; in, For the first The second anomaly level of each pixel; For the first The LBP standard deviation of the neighborhood window centered on a pixel; For the first The grayscale entropy value of pixels within the neighborhood window when the center is a pixel; For the first The gradient direction entropy value of pixels within the neighborhood window when the center is a pixel; For texture feature values; For the first The average gradient within the neighborhood window centered on a pixel; For the first The maximum gray value within the neighborhood window when the center is a pixel; For the first The minimum gray value within the neighborhood window when the center is a pixel; These are grayscale feature values; This is for normalization purposes.
[0044] LBP standard deviation The larger the value, the more chaotic the local texture within the corresponding neighborhood window, and the more it conforms to the characteristics of smoke and flames; and This is used to reflect the disorder and randomness of the local gray-level distribution of the target pixel. The larger the value, the more the local gray-level distribution of the target pixel conforms to the complex and irregular gray-level composition of the flame and smoke area. The larger the value, the higher the contrast within the window, which is less consistent with the low contrast characteristics of smoke and flame areas. The larger the value, the greater the grayscale range within the window, which is less consistent with the characteristic of low contrast in flame and smoke areas.
[0045] The third acquisition module 104 is used to determine the third anomaly level of each pixel in the real-time image frame based on the motion information of the real-time image frame.
[0046] It is understandable that smoke and flames are not static phenomena. That is, if smoke and flames exist in a scene, there are corresponding motion features in consecutive image frames. Considering that the spread of smoke and flames is affected by multiple factors, the motion direction in the image frame is chaotic, and the consistency of the motion vector direction of each pixel is low. At the same time, before smoke and flames are effectively controlled, they usually exhibit expansion characteristics, with a large divergence of motion vectors for each pixel. In addition, in consecutive images, this is reflected by an increase in the area of the connected region to which the pixel belongs. Based on this logic, the motion of each pixel is analyzed to obtain the third degree of anomaly.
[0047] In some embodiments, the previous image frame of the real-time image frame can be obtained, and the motion vector of each pixel in the real-time image frame can be obtained based on the previous image frame as the motion information of the real-time image frame. In this embodiment, the method for obtaining the motion vector of each pixel in the real-time image frame based on the previous image frame and the real-time image frame can be optical flow method, and the specific acquisition process will not be described in detail.
[0048] Obtain the neighborhood window for each pixel. The neighborhood window contains neighboring pixels within a defined range centered on the pixel. In this embodiment, pixels are used as the basis for the calculation. Construct pixels from the 7x7 neighborhood pixels at the center. The neighborhood window.
[0049] In some embodiments, the motion vector of each pixel within a neighborhood window can be obtained, along with its similarity to the mean motion vector of all pixels within the neighborhood window. The mean similarity of the neighborhood window is then determined based on this similarity. In this embodiment, the similarity can be calculated using cosine similarity, i.e., by obtaining the pixel... The mean motion vector of all pixels within the neighborhood window centered at the given point, for each pixel. Get pixels The cosine similarity between the motion vector and the mean motion vector of the corresponding neighborhood window is used as the similarity in this embodiment. The similarity of each pixel within the neighborhood window is obtained using the same method. Therefore, the mean similarity of the neighborhood window is obtained based on the similarity of all pixels within the neighborhood. A higher mean similarity indicates a better current pixel... The more consistent the motion vector direction of each pixel within its neighborhood window, the better.
[0050] In some embodiments, the divergence of each pixel within the neighborhood window can be obtained, and the mean divergence of the neighborhood window can be obtained based on the divergence. The divergence of a pixel can be calculated using the pixel motion vector, for example, by using the vector field divergence formula to obtain the divergence of each pixel within the neighborhood window, and then the mean divergence of the neighborhood window can be calculated. The larger the mean divergence, the more obvious the expansion characteristics of the pixels within the neighborhood window.
[0051] In some embodiments, the area of the connected component of each pixel in the real-time image frame and the previous image frame can also be obtained separately. The connected component to which the pixel belongs can be obtained from the grayscale image of the real-time image frame. The grayscale image is binarized to obtain a binarized image. Then, OpenCV is used to perform connectivity analysis on the binarized image to obtain at least one connected component in the grayscale image. Thus, based on the connected component and the position of the pixel, the connected component to which each pixel belongs can be determined. In this embodiment, the area of the connected component can be the total number of pixels in the connected component.
[0052] Further, based on the mean similarity, mean divergence, and connected component area, the third degree of anomaly of the center pixel of the neighborhood window is determined; optionally, the area difference between the connected component area of each pixel in the real-time image frame and the connected component area of the pixel in the previous image frame can be obtained; the second product of the area difference corresponding to the center pixel of the neighborhood window and the mean divergence is calculated, and the third degree of anomaly of the center pixel of the neighborhood window is determined based on the second product result and the mean similarity.
[0053] Alternatively, the calculation of the third degree of anomaly can be as follows: ; in, For the first The third degree of anomaly of each pixel; For the first The similarity between the mean motion vector of the neighborhood window corresponding to a pixel and the motion vector of the j-th pixel in the neighborhood window; For the first The average similarity of the neighborhood windows corresponding to each pixel, where N is the number of pixels in the neighborhood window; For the first The mean divergence of the neighborhood window corresponding to each pixel; For the first The area of the connected component to which each pixel belongs in a real-time image frame; For the first The area of the connected region to which each pixel belongs in the previous image frame; This represents the area difference; This is the result of the second product; This is for normalization purposes.
[0054] Understandable, The smaller the mean similarity of the neighborhood windows corresponding to the i-th pixel, that is, the lower the consistency of the motion vector direction, the more it conforms to the irregular changes of smoke and flame regions over time; area difference The larger the value, the more it matches the characteristics of smoke and flame areas expanding over time.
[0055] The identification module 105 is used to determine the comprehensive degree of anomaly of each pixel in the real-time image frame based on the first degree of anomaly, the second degree of anomaly, and the third degree of anomaly, and to identify risks based on the comprehensive degree of anomaly.
[0056] Optionally, the average of the first, second, and third anomaly levels for each pixel can be obtained as the overall anomaly level for that pixel. The overall anomaly level is expressed as: ; in, For the first The overall anomaly level of each pixel.
[0057] After determining the overall anomaly level of each pixel in the real-time image frame, target pixels can be selected from the real-time image frame based on the overall anomaly level. In this embodiment, pixels with an overall anomaly level greater than or equal to 0.8 are selected as target pixels. In other embodiments, target pixels can be determined based on different conditions.
[0058] The higher the number of target pixels, the greater the likelihood of a fire in the current real-time image frame. Therefore, the current fire risk information can be determined based on the number of target pixels, and alarm information can be determined based on the fire risk information.
[0059] For example, when the number of target pixels in a real-time image frame is greater than or equal to 50, the current fire risk information is determined to be that a fire has occurred in the exercise environment; otherwise, the current fire risk information is determined to be that no fire has occurred in the exercise environment.
[0060] Furthermore, when the fire risk information is determined to be a fire, an alarm message is generated and sent as a warning. In this embodiment, the alarm message includes at least the fire area outline. The fire area outline can be generated by using each target pixel as a seed point. The growth criteria are based on the comprehensive anomaly degree of adjacent pixels and the gray-level similarity between adjacent pixels and the seed point to obtain the target growth area of the seed point, which serves as the fire area outline. The specific method is existing technology and will not be described in detail here.
[0061] After determining the outline of the fire area, the fire area outline and the current fire risk information can be sent together as the content of the alarm message to warn of the fire and to more conveniently assist relevant personnel in analyzing and controlling the fire.
[0062] In summary, this embodiment proposes a real-time video monitoring system for fire drill simulation, including an acquisition module, a first acquisition module, a second acquisition module, a third acquisition module, and an identification module. Compared with the fixed rules and threshold adjustments in existing methods, this embodiment utilizes the dynamic correlation between multi-dimensional features such as the color features, grayscale features, and motion features of flames and smoke to obtain the comprehensive anomaly degree of each pixel in the real-time image frame, thereby determining the target pixel and generating the contour of smoke or flame based on the region, improving the accuracy of fire identification during the drill and reducing the occurrence of accidents.
[0063] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.
Claims
1. A real-time video monitoring system for fire drill simulation, characterized in that, The system includes: The acquisition module is used to acquire real-time image frames of the training ground; The first acquisition module is used to determine the first degree of abnormality of each pixel in the real-time image frame based on the color information of the real-time image frame. The second acquisition module is used to determine the second degree of abnormality of each pixel in the real-time image frame based on the grayscale information of the real-time image frame. The third acquisition module is used to determine the third degree of anomaly of each pixel in the real-time image frame based on the motion information of the real-time image frame. The identification module is used to determine the comprehensive anomaly level of each pixel in the real-time image frame based on the first anomaly level, the second anomaly level, and the third anomaly level, and to perform risk identification based on the comprehensive anomaly level.
2. The real-time video monitoring system for fire drill simulation according to claim 1, characterized in that, The first acquisition module determines the first anomaly level of each pixel in the real-time image frame based on the color information of the real-time image frame, including: The red (R) channel value, green (G) channel value, blue (B) channel value, hue (H) channel value, saturation (S) channel value, and brightness (V) channel value corresponding to each pixel of the real-time image frame are obtained as the color information of the real-time image frame; The flame characterization value of the pixel is determined based on the R channel value, G channel value, B channel value and H channel value of the pixel; The smoke characterization value of the pixel is determined based on the S-channel value and the V-channel value of the pixel; Based on the flame characterization value and the smoke characterization value, the first degree of anomaly of the pixel is determined.
3. The real-time video monitoring system for fire drill simulation according to claim 2, characterized in that, The first acquisition module is further configured to: Obtain the sum of the R channel value, G channel value, and B channel value corresponding to each pixel; Calculate the first difference between the R channel value and the G channel value for each pixel, and the second difference between the R channel value and the B channel value; Based on the first difference, the second difference, the summation result, and the H channel value, the flame characterization value of the corresponding pixel is determined; The smoke characterization value of the corresponding pixel is determined by multiplying the reciprocal of the S channel value and the V channel value. The maximum value is determined from the flame characterization value and the smoke characterization value, and is used as the first anomaly level of the corresponding pixel.
4. The real-time video monitoring system for fire drill simulation according to claim 1, characterized in that, The second acquisition module determines the second anomaly level of each pixel in the real-time image frame based on the grayscale information of the real-time image frame, including: The grayscale value of each pixel in the real-time image frame is obtained as the grayscale information of the real-time image frame; Obtain the neighborhood window of each pixel, wherein the neighborhood window contains neighboring pixels within a set range centered on the pixel; Obtain the local binary mode (LBP) value, grayscale entropy value, and gradient direction entropy value of each pixel within the neighborhood window; The texture feature value of the center pixel of the neighborhood window is determined based on the grayscale entropy value and the gradient direction entropy value. Obtain the gradient value of each pixel within the neighborhood window, as well as the maximum and minimum grayscale values of the pixels within the neighborhood window; The grayscale feature value of the center pixel of the neighborhood window is determined based on the gradient value, the maximum grayscale value, and the minimum grayscale value. Based on the LBP value of each pixel within the neighborhood window, determine the LBP standard deviation of the center pixel of the neighborhood window; Based on the LBP standard deviation, the texture feature value, and the grayscale feature value, the second anomaly degree of the center pixel of the neighborhood window is determined.
5. A real-time video monitoring system for fire drill simulation according to claim 4, characterized in that, The second acquisition module is further configured to: The product of the grayscale entropy value and the gradient direction entropy value is obtained as the texture feature value; Obtain the difference between the maximum gray value and the minimum gray value, and calculate the average gradient of the pixels within the neighborhood window; The grayscale feature value is determined based on the product of the difference and the mean gradient. Calculate the first product of the texture feature value, the grayscale feature value, and the LBP standard deviation, and determine the second anomaly degree of the center pixel of the neighborhood window based on the first product result.
6. The real-time video monitoring system for fire drill simulation according to claim 1, characterized in that, The third acquisition module determines the third anomaly level of each pixel in the real-time image frame based on the motion information of the real-time image frame, including: Obtain the previous image frame of the real-time image frame, and obtain the motion vector of each pixel in the real-time image frame as the motion information of the real-time image frame based on the previous image frame; Obtain the neighborhood window of each pixel, wherein the neighborhood window contains neighboring pixels within a set range centered on the pixel; Obtain the motion vector of each pixel within the neighborhood window and its similarity to the mean motion vector of all pixels within the neighborhood window, and determine the mean similarity value of the neighborhood window based on the similarity. Obtain the divergence of each pixel within the neighborhood window, and obtain the mean divergence of the neighborhood window based on the divergence. Obtain the area of the connected component of each pixel in the real-time image frame and the previous image frame respectively; The third anomaly level of the center pixel of the neighborhood window is determined based on the mean similarity, the mean divergence, and the area of the connected region.
7. A real-time video monitoring system for fire drill simulation according to claim 6, characterized in that, The third acquisition module is also used for: Obtain the area difference between the area of the connected component of each pixel in the real-time image frame and the area of the connected component of the pixel in the previous image frame; Calculate the second product of the area difference and the mean divergence corresponding to the center pixel of the neighborhood window, and determine the third anomaly degree of the center pixel of the neighborhood window based on the second product and the mean similarity.
8. A real-time video monitoring system for fire drill simulation according to any one of claims 1-7, characterized in that, The identification module determines the comprehensive anomaly level of each pixel in the real-time image frame based on the first anomaly level, the second anomaly level, and the third anomaly level, including: The average of the first abnormality level, the second abnormality level, and the third abnormality level of each pixel is obtained as the comprehensive abnormality level of the corresponding pixel.
9. A real-time video monitoring system for fire drill simulation according to claim 8, characterized in that, The identification module performs risk identification based on the overall anomaly level, including: Target pixels are selected from the real-time image frames based on the overall anomaly level. Based on the number of target pixels, the current fire risk information is determined, and alarm information is determined based on the fire risk information.
10. A real-time video monitoring system for fire drill simulation according to claim 9, characterized in that, The alarm information includes at least the outline of the fire area, and the identification module is further used for: Each target pixel is used as a seed point for region generation. The growth criteria are based on the comprehensive anomaly degree of adjacent pixels and the gray-level similarity between the adjacent pixels and the seed point, to obtain the target growth region of the seed point, which serves as the outline of the fire area.