Data abnormal value elimination method and system for fire monitoring

Abstract

The invention relates to the technical field of image recognition, and discloses a data abnormal value elimination method and system for fire monitoring. The method comprises the following steps: acquiring an original video stream of a fire monitoring area; extracting an image frame at the current time from the original video stream; extracting a foreground motion area of the image frame from the image frame; extracting fire features of the foreground motion area; identifying a fire area from the foreground motion area according to the fire features of the foreground motion area; calculating a feature mutation trend of each pixel in the fire area; according to the feature mutation trend of each pixel in the fire area, identifying interference pixels from the fire area; and carrying out elimination processing on the interference pixels. The method makes remarkable progress in the aspects of reducing the false alarm rate, improving the adaptability to complex scenes, retaining key feature dynamic information and the like, and provides effective technical support for reliable operation of a fire monitoring system.

Description

Technical Field

[0001] The present invention relates to the field of image recognition technology, and more particularly, to a method and system for eliminating abnormal values ​​in fire monitoring data. Background Art

[0002] Fire is a highly prevalent disaster threatening public safety. Early detection and early warning are crucial for reducing casualties and property losses. With the widespread adoption of video surveillance technology, computer vision-based fire detection systems have become a key technology in the intelligent fire protection field due to their intuitiveness and wide coverage. These systems typically collect raw video streams from the fire monitoring area and utilize image processing and pattern recognition techniques to automatically identify fire characteristics (such as flame color, brightness, flicker frequency, and smoke diffusion patterns), enabling real-time fire monitoring and alarming.

[0003] However, in practical applications, a core challenge facing fire detection systems is how to accurately distinguish true fire signatures from false fire signals generated by various interference factors. Due to the complexity of the monitored scene, the raw video stream often contains a large number of dynamic changes not related to fires. These include: periodic changes in natural light (such as sunrise and sunset, cloud cover), noise fluctuations in monitoring equipment, random movements of small objects such as insects and birds, light and shadow disturbances from human activities (such as light switches and vehicle lights), and even other types of aerosols (such as dust and water mist). These interference factors may manifest in the image as sudden or continuous changes in pixel values ​​in localized areas. If not effectively eliminated, the system may mistakenly identify the interference areas as fire areas, generating a large number of false alarms. Conversely, excessive filtering may miss the true initial characteristics of a fire, resulting in underreporting.

[0004] Therefore, how to accurately remove interference pixels from video data and retain the true fire characteristics is a key technical difficulty in improving the reliability of the fire monitoring system. Summary of the Invention

[0005] In order to solve the above technical problems, the present application is proposed to provide a method and system for removing data outliers for fire monitoring, which can accurately remove interfering pixel data in video data and retain the true fire characteristics.

[0006] In a first aspect, the present invention provides a method for eliminating data outliers for fire monitoring, comprising: obtaining an original video stream of a fire monitoring area; extracting an image frame at a current time from the original video stream; extracting a foreground motion area of ​​the image frame from the image frame; extracting fire features of the foreground motion area; identifying a fire area from the foreground motion area based on the fire features of the foreground motion area; calculating a characteristic mutation trend of each pixel in the fire area; identifying interference pixels from the fire area based on the characteristic mutation trend of each pixel in the fire area; and eliminating the interference pixels.

[0007] Preferably, in the aforementioned method for eliminating data outliers for fire monitoring, the foreground motion area of ​​the image frame is extracted from the image frame, including: taking any pixel in the image frame as a target pixel, obtaining a color histogram of the target pixel, and calculating a dense optical flow field of the target pixel; extracting the color feature of the target pixel from the color histogram of the target pixel; calculating the optical flow direction feature of the target pixel based on the dense optical flow field of the target pixel; splicing the color feature and the optical flow direction feature of the target pixel to form a comprehensive feature of the target pixel; generating a foreground mask based on the comprehensive feature of each pixel in the image frame; and processing the image frame using the foreground mask to obtain the foreground motion area of ​​the image frame.

[0008] Preferably, in the aforementioned method for eliminating abnormal values ​​in data for fire monitoring, the foreground mask is generated according to the comprehensive features of each pixel in the image frame, including: setting a pixel property function for the target pixel ,in is the preset color weight, represents dot product, The function is a natural exponential function, represents the position of the target pixel, is the comprehensive feature of the target pixel, is the mean value of the comprehensive features of all pixels in the image frame, is the variance of the comprehensive features of all pixels in the image frame, is the dense optical flow field of the target pixel, is the mean of the dense optical flow fields of all pixels in the image frame; sets a foreground indicator value for the target pixel , set the pixel property function at the target pixel When the foreground indicator value of the target pixel Takes 1, otherwise takes 0, where is a preset segmentation threshold; and the foreground mask is generated according to the foreground indication value of each pixel in the image frame.

[0009] Preferably, in the aforementioned method for eliminating data outliers for fire monitoring, extracting the fire features of the foreground motion area includes: obtaining the historical flame area in the historical image frame of the previous time before the current time; selecting a candidate flame area in the foreground motion area based on the historical flame area and a preset magnification ratio; using a preset U-Net network to identify the flame area from the candidate flame area; and calculating the area change rate and edge jaggedness of the flame area as the fire features of the foreground motion area.

[0010] Preferably, in the aforementioned method for eliminating data outliers for fire monitoring, the edge jaggedness of the flame area is calculated, including: setting a grid size according to a preset fire monitoring accuracy; dividing the flame area into K grids of the same size according to the grid size; and counting the number of pixels in the grid whose color gradient exceeds a preset color threshold. The number of grids ,in, The function is a conditional function, which takes 1 when the condition is met, otherwise it takes 0. The function is used to get the maximum value. is the i-th grid among K grids, for Any pixel in for Calculate the fractal dimension of the edge curve of the flame area ; Calculate the edge jaggedness of the flame area .

[0011] Preferably, in the aforementioned method for eliminating data outliers for fire monitoring, the method of extracting the fire features of the foreground motion area further includes: extracting the color features and texture features of the foreground motion area; splicing the color features and texture features of the foreground motion area into fusion features of the foreground motion area; inputting the fusion features of the foreground motion area into a preset neural network, and allowing the neural network to identify a smoke area from the foreground motion area; and extracting smoke diffusion features from the smoke area as fire features of the foreground motion area.

[0012] Preferably, in the aforementioned method for eliminating abnormal values ​​in data for fire monitoring, the extraction of smoke diffusion features from the smoke area comprises: obtaining a texture histogram of the smoke area at the current time; And the historical texture histogram of the previous time of the current time ; Calculate the texture histogram The probability distribution of , the historical texture histogram The probability distribution of And the texture histogram With the historical texture histogram The joint probability distribution of ; Calculate the texture continuity of smoke features ; Calculate the energy standard deviation of the texture histogram ; Calculate the optical flow vector field of the smoke area at the current time ; Calculate the optical flow vector field Speed ​​variance , speed standard deviation and velocity gradient ; Calculate the motion smoothness of the smoke area ,in, Represents dot product; calculates the physical continuity of the smoke area , as the smoke diffusion characteristic.

[0013] Preferably, in the aforementioned method for eliminating data outliers for fire monitoring, interfering pixels are identified from the fire area based on the characteristic mutation trend of each pixel in the fire area, including: generating an anomaly probability matrix based on the characteristic mutation trend of each pixel in the fire area; inputting the anomaly probability matrix into a trained neural network to output an environmental noise mask corresponding to the fire area; and using the environmental noise mask to identify interfering pixels from the fire area.

[0014] Preferably, in the aforementioned method for eliminating data outliers for fire monitoring, before eliminating the interfering pixels, it also includes: obtaining the abnormal confidence of the interfering pixels fed back by a preset sensor; detecting whether the abnormal confidence of the interfering pixels is higher than a preset confidence threshold; when the judgment result is yes, entering the step of eliminating the interfering pixels.

[0015] In a second aspect, the present invention provides a data outlier elimination system for fire monitoring, comprising: a video stream acquisition module for acquiring the original video stream of the fire monitoring area; an image frame extraction module for extracting the image frame of the current time from the original video stream; a motion area extraction module for extracting the foreground motion area of ​​the image frame from the image frame; a fire feature extraction module for extracting the fire features of the foreground motion area; a fire area extraction module for identifying the fire area from the foreground motion area based on the fire features of the foreground motion area; a feature mutation calculation module for calculating the feature mutation trend of each pixel in the fire area; an interference pixel identification module for identifying interference pixels from the fire area based on the feature mutation trend of each pixel in the fire area; and a elimination processing module for eliminating the interference pixels.

[0016] The above one or more technical solutions of the present invention have at least one or more of the following beneficial effects:

[0017] According to the technical solution of the present invention, through the progressive processing flow of "foreground motion area extraction-fire feature focusing-pixel-level mutation trend analysis-interference pixel elimination", significant progress has been made in reducing false alarm rates, improving adaptability to complex scenarios, and retaining key feature dynamic information, providing effective technical support for the reliable operation of the fire monitoring system. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] The above and other purposes, features, and advantages of the present application will become more apparent through a more detailed description of the embodiments of the present application in conjunction with the accompanying drawings. The accompanying drawings are intended to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation of the present application. In the drawings, the same reference numerals generally represent the same components or steps.

[0019] Figure 1 Flowchart of a method for removing abnormal values ​​from fire monitoring data according to an embodiment of the present application;

[0020] Figure 2 This is a partial flow chart of a method for eliminating abnormal values ​​in fire monitoring data according to an embodiment of the present application;

[0021] Figure 3 is another layout flow chart of a method for eliminating abnormal values ​​in fire monitoring data according to an embodiment of the present application;

[0022] Figure 4 is another partial flow chart of a method for eliminating abnormal values ​​in fire monitoring data according to an embodiment of the present application;

[0023] Figure 54 is a block diagram of a data outlier removal system for fire monitoring according to an embodiment of the present application. DETAILED DESCRIPTION

[0024] Some embodiments of the present invention are described below with reference to the accompanying drawings. Those skilled in the art should understand that these embodiments are only used to explain the technical principles of the present invention and are not intended to limit the scope of protection of the present invention.

[0025] like Figure 1 As shown, in one embodiment of the present invention, a method for removing abnormal values ​​from data for fire monitoring is provided, comprising:

[0026] Step S110: Acquire the original video stream of the fire monitoring area.

[0027] Step S120: extracting the image frame at the current time from the original video stream.

[0028] Step S130: extracting a foreground motion region of the image frame from the image frame.

[0029] In this embodiment, the foreground motion area of ​​the image frame is extracted through step S130, which effectively eliminates interference from static background (such as buildings and vegetation) and irrelevant dynamic background. Only dynamic targets that may pose a fire risk (such as flames and smoke) are processed. On the one hand, the amount of data for subsequent feature extraction and mutation trend analysis is greatly reduced, thereby improving the real-time performance of the algorithm. On the other hand, the interference of static background noise on fire features is avoided, ensuring that the extracted fire features (such as flame color and smoke form) are more concentrated and accurate, providing high-quality input data for subsequent identification.

[0030] Step S140: extracting fire features in the foreground motion area.

[0031] Step S150 : identifying the fire area from the foreground motion area according to the fire characteristics of the foreground motion area.

[0032] Step S160 , calculating the characteristic mutation trend of each pixel in the fire area.

[0033] Unlike traditional methods that rely on fixed thresholds and machine learning methods that require large numbers of labeled samples, this embodiment directly exploits the dynamic patterns of real-world fire characteristics through pixel-level mutation trend analysis. This approach eliminates the need for manually set fixed thresholds and does not rely on training data from specific scenarios. Experimental validation demonstrates that this method maintains high fire area recognition accuracy in low-light environments such as rainy days, blurred distant scenes, and mixed light source scenarios (such as the presence of streetlights and vehicle lights in the monitored area), significantly improving its adaptability to complex scenarios compared to traditional methods.

[0034] Step S170 , identifying interference pixels in the fire area according to the characteristic mutation trend of each pixel in the fire area.

[0035] Step S180: Eliminate interfering pixels.

[0036] In this embodiment, only the identified interfering pixels are eliminated, retaining the original pixel values ​​and mutation trends of the real fire area, and completely preserving key dynamic information such as flame color and smoke diffusion, providing richer data support for subsequent advanced analysis such as fire level assessment and fire spread prediction.

[0037] According to the technical solution of this embodiment, through the progressive processing flow of "foreground motion area extraction-fire feature focusing-pixel-level mutation trend analysis-interference pixel elimination", significant progress has been made in reducing false alarm rates, improving adaptability to complex scenarios, and retaining key feature dynamic information, providing effective technical support for the reliable operation of the fire monitoring system.

[0038] like Figure 2 As shown, an embodiment of the present invention provides a method for removing data outliers for fire monitoring. Compared with the previous embodiment, the method for removing data outliers for fire monitoring in this embodiment includes step S130:

[0039] In step S210 , any pixel in the image frame is taken as a target pixel, a color histogram of the target pixel is obtained, and a dense optical flow field of the target pixel is calculated.

[0040] Step S220 , extracting the color feature of the target pixel from the color histogram of the target pixel.

[0041] Step S230 : calculating the optical flow direction feature of the target pixel according to the dense optical flow field of the target pixel.

[0042] In this embodiment, static background interference (such as fluttering curtains, flying birds, and other transient objects) may be mistaken for fire areas, while dynamic background interference (such as global motion caused by camera shake or flashing vehicle lights) may disrupt foreground segmentation. This embodiment constructs a comprehensive feature for each pixel through steps S210-S240: a color histogram captures the chromaticity distribution characteristics of the flame, such as orange-red hues and high-brightness areas, and a dense optical flow field reflects the dynamic motion patterns of the pixel (such as the upward spread of flames and the horizontal diffusion of smoke). The combination of these two forms a more comprehensive feature representation, effectively distinguishing between true fire motion (color and motion direction consistent with combustion patterns) and interference motion (abrupt color changes but random motion direction).

[0043] Step S240: The color feature and the optical flow direction feature of the target pixel are combined to form a comprehensive feature of the target pixel.

[0044] Step S250: generating a foreground mask according to the comprehensive features of each pixel in the image frame.

[0045] Step S260 : Process the image frame using the foreground mask to obtain a foreground motion region of the image frame.

[0046] Step S250 specifically includes:

[0047] (1) Set the pixel property function for the target pixel ,in, is the preset color weight, represents dot product, The function is a natural exponential function, represents the location of the target pixel, is the comprehensive feature of the target pixel, is the mean of the comprehensive features of all pixels in the image frame, is the variance of the comprehensive features of all pixels in the image frame, is the dense optical flow field of the target pixel, is the mean of the dense optical flow fields of all pixels in the image frame.

[0048] In this embodiment, the pixel property function innovatively combines the statistical characteristics of color distribution and motion pattern: through the Gaussian kernel function Penalize the color deviation from the mean and suppress background noise (such as local brightness fluctuations caused by lighting changes); Measures the consistency of pixel motion direction with the global motion trend, eliminating the interference of reverse motion (such as reverse displacement caused by camera shaking). Dynamically balance the weights of color and motion features. Compared with the segmentation method with fixed threshold, it can adapt to the differences in feature distribution in different scenes.

[0049] (2) Set the foreground indicator value for the target pixel , set the pixel function at the target pixel When the foreground indicator value of the target pixel Takes 1, otherwise takes 0, where is the preset segmentation threshold.

[0050] In this embodiment, the foreground indicator value The binarization processing, combined with the setting of the adaptive threshold T, is conducive to filtering various types of interference such as instantaneous interference and global interference.

[0051] (3) Generate a foreground mask based on the foreground indicator value of each pixel in the image frame.

[0052] According to the technical solution of this embodiment, through feature fusion, adaptive statistical modeling and efficient computing design, more accurate foreground segmentation and interference suppression are achieved in fire monitoring scenarios, laying a solid foundation for subsequent outlier elimination and fire determination.

[0053] like Figure 3 As shown, an embodiment of the present invention provides a method for removing abnormal values ​​from data for fire monitoring. Compared with the previous embodiment, the method for removing abnormal values ​​from data for fire monitoring in this embodiment includes: step S140:

[0054] Step S310: obtaining a historical flame region in a historical image frame at a time previous to the current time.

[0055] Step S320 : selecting a candidate flame region in the foreground motion region according to the historical flame region and a preset magnification ratio.

[0056] Step S330 , using a preset U-Net network to identify a flame area from the candidate flame areas.

[0057] In this embodiment, steps S310-S330 utilize the flame region from the previous frame as prior information and dynamically adjust the candidate region range based on a preset magnification ratio, effectively narrowing the search space and significantly reducing subsequent computational effort. A U-Net network is used to semantically segment the candidate regions. Its encoder-decoder architecture accurately captures the texture and morphological characteristics of flames (such as irregular edges and dynamic spread patterns).

[0058] Step S340 , calculating the area change rate and edge jaggedness of the flame area as fire characteristics of the foreground motion area.

[0059] The step S340 of calculating the edge jaggedness of the flame area includes: (1) Set the grid size according to the preset fire monitoring accuracy. (2) Divide the flame area into K grids of the same size according to the grid size. (3) The color gradient of at least one pixel in the statistical grid exceeds the preset color threshold The number of grids ,in, The function is a conditional function, which takes 1 when the condition is met, otherwise it takes 0. The function is used to get the maximum value. is the i-th grid among K grids, for Any pixel in for The color gradient, (4) Calculate the fractal dimension of the edge curve of the flame area , (5) Calculate the edge jaggedness of the flame area .

[0060] In this embodiment, the dynamic geometric characteristics of flames are incorporated into the outlier determination system for the first time through the area change rate and edge jaggedness indicators proposed in step S340. The area change rate can reflect the spread / contraction trend of the flame. The flame area of ​​a real fire typically shows a continuous growth or fluctuating upward trend, while interference from flying insects or flickering lights can cause sudden changes in area. This embodiment innovates the calculation of edge jaggedness, quantifying the complexity of the flame edge through grid division and color gradient statistics. The edges of real flames exhibit fractal characteristics due to combustion instability, while the edges of artificial light sources or reflective objects are typically regular geometric shapes. Therefore, fractal dimensions can be converted into quantifiable features. In scenes with strong lighting (such as direct headlights) or low visibility (such as rainy and foggy weather), fractal dimension analysis can effectively suppress edge noise caused by light reflection, significantly improving flame recognition accuracy.

[0061] According to the technical solution of this embodiment, the extracted area change rate and edge jaggedness features provide high-confidence input for subsequent pixel-level mutation trend analysis: by filtering out background pixels in the flame area, mutation trend analysis focuses on the actual flame area; fractal dimensions assist in determining the dynamic stability of the flame edge, providing new dimensional features for fire level assessment. In summary, this embodiment achieves more accurate fire feature extraction and interference suppression in complex scenarios through a spatiotemporal flame recognition mechanism, fractal dimension feature quantification, and multi-feature fusion decision-making, providing key technical support for improving the reliability of fire monitoring systems.

[0062] like Figure 4 As shown, an embodiment of the present invention provides a method for removing abnormal values ​​from data for fire monitoring. Compared with the previous embodiment, the method for removing abnormal values ​​from data for fire monitoring in this embodiment further includes, in step S140:

[0063] Step S410: extracting color features and texture features of the foreground motion area.

[0064] Step S420 : combining the color features and texture features of the foreground motion region into a fusion feature of the foreground motion region.

[0065] In step S430, the fusion features of the foreground motion area are input into a preset neural network, and the neural network identifies the smoke area from the foreground motion area.

[0066] In this embodiment, color features reflect the chromaticity distribution of captured smoke (e.g., grayish-white tones and low saturation characteristics), while texture features reflect the diffusion pattern of smoke (e.g., irregular edges and fuzzy gradients). Combining these two features creates a more comprehensive feature representation. Combined with the semantic segmentation capabilities of neural networks, this significantly improves the accuracy of smoke region recognition compared to traditional methods.

[0067] Step S440: extracting smoke diffusion features from the smoke area as fire features of the foreground motion area.

[0068] Wherein, step S440 includes:

[0069] (1) Get the texture histogram of the smoke area at the current time And the historical texture histogram of the previous time at the current time .

[0070] (2) Calculate texture histogram The probability distribution of , historical texture histogram The probability distribution of and texture histogram and historical texture histograms The joint probability distribution of .

[0071] (3) Calculate the texture continuity of smoke features, , In this embodiment, the texture continuity index proposed By calculating the mutual information between the texture distribution of the current frame and the historical frames, it is helpful to quantify the spatiotemporal continuity of smoke diffusion: due to the continuous combustion process, the texture distribution of real smoke has time correlation; interference noise, such as flying insects, can easily cause texture mutations in a single frame. This indicator can effectively filter out instantaneous interference while preserving the real smoke area to the greatest extent.

[0072] (4) Calculate the energy standard deviation of the texture histogram .

[0073] (5) Calculate the optical flow vector field of the smoke area at the current time .

[0074] (6) Calculate the optical flow vector field Speed ​​variance , speed standard deviation and velocity gradient .

[0075] (7) Calculate the smoothness of the motion in the smoke area ,in, Represents dot product.

[0076] (8) Calculate the physical continuity of the smoke area , as a smoke diffusion characteristic.

[0077] In this embodiment, the physical continuity index For the first time, the physical laws of smoke diffusion and motion characteristics are combined: the characteristics of real smoke diffusion at low speed due to airflow disturbance are reflected in The smaller the value, the smoke edge forms a smooth gradient due to density changes. The smaller the value, the discrete degree of the energy distribution of the smoke texture can be To reflect. The combination of the three forms The indicator can effectively distinguish between smoke and artificial light sources (such as flashing car lights).

[0078] According to the technical solution of this embodiment, by introducing the physical continuity feature analysis and multimodal feature fusion mechanism of the smoke area, the accuracy and anti-interference ability of smoke recognition in fire monitoring are significantly improved, and high-precision smoke recognition and interference suppression are achieved in complex scenarios, providing a key technical breakthrough for improving the reliability of the fire monitoring system.

[0079] In one embodiment of the present invention, a method for removing outliers from data for fire monitoring is provided. Compared with the above-mentioned embodiments, the method for removing outliers from data for fire monitoring in this embodiment includes, in step S170:

[0080] (1) Generate an anomaly probability matrix based on the characteristic mutation trend of each pixel in the fire area.

[0081] (2) Input the anomaly probability matrix into the trained neural network and output the environmental noise mask corresponding to the fire area.

[0082] (3) Using the ambient noise mask, interfering pixels are identified from the fire area.

[0083] In this example, a pixel-level anomaly probability distribution is generated based on feature mutation trends (such as brightness gradients and color diffusion speeds), reflecting the confidence level of anomalies for each pixel. Anomaly probabilities in real fire areas typically exhibit spatial continuity (e.g., the gradual distribution of flame edges), while interfering noise (such as flying insects and light flickering) appears as localized, discrete high-probability points. A lightweight convolutional neural network is used to perform semantic segmentation on the anomaly probability matrix, and its spatial attention mechanism effectively suppresses isolated noise points.

[0084] Before step S180, the method further includes:

[0085] (1) Obtain the abnormal confidence of the interference pixel fed back by the preset sensor.

[0086] (2) Detect whether the abnormal confidence of the interference pixel is higher than the preset confidence threshold.

[0087] (3) When the judgment result is yes, proceed to step S180.

[0088] The technical solution of this embodiment combines data from multiple sources, including temperature, smoke concentration, and gas sensors, to calculate the confidence level of physical anomalies in interfering pixels. For example, insect interference might only trigger an anomaly in a visual sensor, while a real fire would be accompanied by a simultaneous rise in temperature and smoke concentration. The preset confidence threshold is automatically adjusted based on the ambient background noise level (such as light scattering caused by rain or fog). For example, in high-interference scenarios, the threshold can be dynamically lowered to ensure that real fire signals are not missed; in stable environments, the threshold can be raised to reduce false alarms.

[0089] like Figure 5 As shown, in one embodiment of the present invention, a data outlier removal system for fire monitoring is provided, comprising:

[0090] The video stream acquisition module 510 acquires the original video stream of the fire monitoring area.

[0091] The image frame extraction module 520 extracts the image frame at the current time from the original video stream.

[0092] The motion region extraction module 530 extracts the foreground motion region of the image frame from the image frame.

[0093] In this embodiment, the foreground motion area of ​​the image frame is extracted through step S130, which effectively eliminates interference from static background (such as buildings and vegetation) and irrelevant dynamic background. Only dynamic targets that may pose a fire risk (such as flames and smoke) are processed. On the one hand, the amount of data for subsequent feature extraction and mutation trend analysis is greatly reduced, thereby improving the real-time performance of the algorithm. On the other hand, the interference of static background noise on fire features is avoided, ensuring that the extracted fire features (such as flame color and smoke form) are more concentrated and accurate, providing high-quality input data for subsequent identification.

[0094] The fire feature extraction module 540 extracts fire features from the foreground motion area.

[0095] The fire area extraction module 550 identifies the fire area from the foreground motion area according to the fire characteristics of the foreground motion area.

[0096] The feature mutation calculation module 560 calculates the feature mutation trend of each pixel in the fire area.

[0097] Unlike traditional methods that rely on fixed thresholds and machine learning methods that require large numbers of labeled samples, this embodiment directly exploits the dynamic patterns of real-world fire characteristics through pixel-level mutation trend analysis. This approach eliminates the need for manually set fixed thresholds and does not rely on training data from specific scenarios. Experimental validation demonstrates that this method maintains high fire area recognition accuracy in low-light environments such as rainy days, blurred distant scenes, and mixed light source scenarios (such as the presence of streetlights and vehicle lights in the monitored area), significantly improving its adaptability to complex scenarios compared to traditional methods.

[0098] The interference pixel identification module 570 identifies interference pixels in the fire area according to the characteristic mutation trend of each pixel in the fire area.

[0099] The elimination processing module 580 eliminates interference pixels.

[0100] In this embodiment, only the identified interfering pixels are eliminated, retaining the original pixel values ​​and mutation trends of the real fire area, and completely preserving key dynamic information such as flame color and smoke diffusion, providing richer data support for subsequent advanced analysis such as fire level assessment and fire spread prediction.

[0101] According to the technical solution of this embodiment, through the progressive processing flow of "foreground motion area extraction-fire feature focusing-pixel-level mutation trend analysis-interference pixel elimination", significant progress has been made in reducing false alarm rates, improving adaptability to complex scenarios, and retaining key feature dynamic information, providing effective technical support for the reliable operation of the fire monitoring system.

[0102] The basic principles of the present application have been described above in conjunction with specific embodiments. However, it should be noted that the advantages, strengths, and effects mentioned in this application are merely illustrative and not restrictive, and it should not be assumed that these advantages, strengths, and effects are required of each embodiment of this application. In addition, the specific details disclosed above are merely illustrative and facilitating understanding, and are not restrictive. The above details do not limit this application to necessarily being implemented using the above specific details.

[0103] The block diagrams of the devices, devices, equipment, and systems involved in this application are merely illustrative examples and are not intended to require or imply that they must be connected, arranged, or configured in the manner shown in the block diagrams. As will be appreciated by those skilled in the art, these devices, devices, equipment, and systems can be connected, arranged, or configured in any manner. Words such as "include," "comprise," "have," and the like are open-ended words, meaning "including but not limited to," and can be used interchangeably therewith. The words "or" and "and" used herein refer to the words "and / or" and can be used interchangeably therewith, unless the context clearly indicates otherwise. The word "such as" used herein refers to the phrase "such as but not limited to," and can be used interchangeably therewith.

[0104] It should also be noted that in the apparatus, device, and method of the present application, each component or each step can be decomposed and / or recombined, and such decomposition and / or recombination should be regarded as equivalent solutions of the present application.

[0105] The above description of the disclosed aspects is provided to enable any person skilled in the art to make or use the present application. Various modifications to these aspects will be readily apparent to those skilled in the art, and the general principles defined herein may be applied to other aspects without departing from the scope of the present application. Therefore, the present application is not intended to be limited to the aspects shown herein, but rather to be accorded the widest scope consistent with the principles and novel features disclosed herein.

[0106] The above description has been provided for the purpose of illustration and description. Furthermore, this description is not intended to limit the embodiments of the present application to the forms disclosed herein. Although a number of example aspects and embodiments have been discussed above, those skilled in the art will recognize certain variations, modifications, alterations, additions, and sub-combinations thereof.

Claims

1. Methods for eliminating outliers in fire monitoring data, including: Obtain the original video stream of the fire monitoring area; Extracting the image frame at the current time from the original video stream; extracting a foreground motion region of the image frame from the image frame; extracting fire features in the foreground motion area; identifying a fire area from the foreground motion area according to fire characteristics of the foreground motion area; Calculate the characteristic mutation trend of each pixel in the fire area; identifying interference pixels from the fire area according to a characteristic mutation trend of each pixel in the fire area; The interference pixels are eliminated.

2. The method for eliminating abnormal values ​​in fire monitoring data according to claim 1, wherein: Extracting a foreground motion region of the image frame from the image frame includes: Taking any pixel in the image frame as a target pixel, obtaining a color histogram of the target pixel, and calculating a dense optical flow field of the target pixel; Extracting a color feature of the target pixel from a color histogram of the target pixel; Calculating an optical flow direction feature of the target pixel according to the dense optical flow field of the target pixel; Splicing the color feature and the optical flow direction feature of the target pixel to form a comprehensive feature of the target pixel; generating a foreground mask based on the comprehensive features of each pixel in the image frame; The image frame is processed using the foreground mask to obtain a foreground motion region of the image frame.

3. The method for eliminating abnormal values ​​in fire monitoring data according to claim 2, wherein: Generating a foreground mask according to the comprehensive features of each pixel in the image frame, comprising: Set the pixel property function for the target pixel ,in, is the preset color weight, represents dot product, The function is a natural exponential function, represents the position of the target pixel, is the comprehensive feature of the target pixel, is the mean value of the comprehensive features of all pixels in the image frame, is the variance of the comprehensive features of all pixels in the image frame, is the dense optical flow field of the target pixel, is the mean value of the dense optical flow field of all pixels in the image frame; Set the foreground indicator value for the target pixel , set the pixel property function at the target pixel When the foreground indicator value of the target pixel Take 1, otherwise take 0, where is the preset segmentation threshold; The foreground mask is generated according to the foreground indication value of each pixel in the image frame.

4. The method for eliminating abnormal values ​​in fire monitoring data according to claim 1, wherein: Extracting fire features from the foreground motion area includes: Acquire a historical flame area in a historical image frame at a time previous to the current time; Selecting a candidate flame area in the foreground motion area according to the historical flame area and a preset magnification ratio; Using a preset U-Net network to identify the flame area from the candidate flame areas; The area change rate and edge jaggedness of the flame area are calculated as fire features of the foreground motion area.

5. The method for eliminating abnormal values ​​in fire monitoring data according to claim 4, wherein: Calculating the edge jaggedness of the flame area includes: Set the grid size according to the preset fire monitoring accuracy; Dividing the flame area into K grids of the same size according to the grid size; The color gradient of at least one pixel in the statistical grid exceeds the preset color threshold The number of grids ,in, The function is a conditional function, which takes 1 when the condition is met, otherwise it takes 0. The function is used to get the maximum value. is the i-th grid among K grids, for Any pixel in for Color gradient; Calculate the fractal dimension of the edge curve of the flame area ; Calculate the edge jaggedness of the flame area .

6. The method for eliminating abnormal values ​​in fire monitoring data according to claim 4, wherein: Extracting fire features from the foreground motion area also includes: Extracting color features and texture features of the foreground motion area; splicing the color features and texture features of the foreground motion region into a fusion feature of the foreground motion region; Inputting the fusion features of the foreground motion area into a preset neural network, and allowing the neural network to identify the smoke area from the foreground motion area; Smoke diffusion features are extracted from the smoke area as fire features of the foreground motion area.

7. The method for eliminating abnormal values ​​in fire monitoring data according to claim 6, wherein: Extracting smoke diffusion features from the smoke area includes: Get the texture histogram of the smoke area at the current time And the historical texture histogram of the previous time of the current time ; Calculate the texture histogram The probability distribution of , the historical texture histogram The probability distribution of And the texture histogram With the historical texture histogram The joint probability distribution of ; Calculate the texture continuity of smoke features ; Calculate the energy standard deviation of the texture histogram ; Calculate the optical flow vector field of the smoke area at the current time ; Calculate the optical flow vector field Speed ​​variance , speed standard deviation and velocity gradient ; Calculate the motion smoothness of the smoke area ,in, represents dot product; Calculate the physical continuity of the smoke area , as the smoke diffusion characteristic.

8. The method for eliminating abnormal values ​​in fire monitoring data according to claim 1, wherein: Identifying interference pixels in the fire area according to a characteristic mutation trend of each pixel in the fire area includes: generating an abnormal probability matrix according to a characteristic mutation trend of each pixel in the fire area; Inputting the abnormal probability matrix into a trained neural network and outputting an environmental noise mask corresponding to the fire area; Using the ambient noise mask, interfering pixels are identified from the fire area.

9. The method for eliminating abnormal values ​​in fire monitoring data according to claim 1, wherein: Before removing the interfering pixels, the method for removing abnormal values ​​from fire monitoring data further includes: Obtaining anomaly confidence of the interference pixel fed back by a preset sensor; Detecting whether the abnormality confidence of the interfering pixel is higher than a preset confidence threshold; When the judgment result is yes, the step of removing the interfering pixels is entered.

10. Data outlier removal system for fire monitoring, including: Video stream acquisition module, which obtains the original video stream of the fire monitoring area; An image frame extraction module extracts the image frame of the current time from the original video stream; A motion region extraction module extracts a foreground motion region of the image frame from the image frame; a fire feature extraction module for extracting fire features from the foreground motion area; a fire area extraction module, which identifies a fire area from the foreground motion area according to fire characteristics of the foreground motion area; a characteristic mutation calculation module for calculating the characteristic mutation trend of each pixel in the fire area; an interference pixel identification module, identifying interference pixels from the fire area according to a characteristic mutation trend of each pixel in the fire area; The elimination processing module performs elimination processing on the interference pixels.

Images