Electrical fire intelligent identification method and system based on visual image

Abstract

The invention relates to the technical field of electrical fire identification, in particular to an electrical fire intelligent identification method and system based on a visual image, and the system comprises a visual image fire identification platform, a fire information base, an infrared image analysis unit, a region identification unit, a feature matching unit, a dynamic management and control unit and an alarm response unit. According to the method, regional fire monitoring feedback analysis is performed on the infrared feature image preliminarily, so that the suspicious risk condition of each monitoring point in the target electrical equipment can be known, and data support is provided for subsequent analysis; whether the target electrical equipment has a fire or not is analyzed from the two aspects of infrared feature image processing and monitoring video image feature extraction, and through the mode of combining the two fire analysis methods, the fire recognition error rate is reduced, the target electrical equipment with a disaster is analyzed deeply, and the fire recognition efficiency is improved. Reasonable management and control can be carried out according to the dynamic change trend of the fire, and the fire management efficiency can be improved.

Description

Technical Field

[0001] The present invention relates to the technical field of electrical fire identification, and in particular to a method and system for intelligently identifying electrical fires based on visual images. Background Art

[0002] Electrical fires are usually manifested as an increase in line current, which leads to an increase in temperature, thus causing the cable to catch fire. The traditional method of monitoring electrical fires is to detect through cable temperature, residual current and smoke; however, research and practical application of the above methods and existing technologies have found that the above methods and existing technologies have at least the following defects: too many video monitoring devices are deployed, and it is impossible to conduct targeted and purposeful monitoring based on the relevant precursors of fire anomalies, which is easy to cause unnecessary waste of monitoring resources and affect the accuracy of recognition results, thereby resulting in a high false alarm rate of fire safety events, and it is impossible to reasonably control the dynamic change trend of the fire, resulting in an increase in the impact of the fire and a reduction in the efficiency of electrical fire management; In view of the above technical defects, a solution is now proposed. Summary of the invention

[0003] The purpose of the present invention is to provide an intelligent identification method and system for electrical fires based on visual images to solve the above-mentioned technical defects. The present invention helps to understand the suspicious risk situation of each monitoring point in the target electrical equipment through preliminary regional fire monitoring feedback analysis of infrared feature images, so as to provide data support for subsequent analysis. The infrared feature image processing and monitoring video image feature extraction are used to analyze whether a fire has occurred in the target electrical equipment. The combination of the two fire analysis methods helps to reduce the fire identification error rate, and helps to provide timely and accurate early warning in the early stage of the fire, as well as in-depth analysis of the target electrical equipment in disaster situation, so as to carry out reasonable management and control according to the dynamic change trend of the fire, so as to minimize the impact of the fire, and help to improve the efficiency of fire management.

[0004] The object of the present invention can be achieved by the following technical solutions: an intelligent electrical fire recognition system based on visual images, comprising a visual image fire recognition platform, a fire information library, an infrared image analysis unit, a region recognition unit, a feature matching unit, a dynamic control unit, and an alarm response unit; The visual image fire recognition platform is used to retrieve the infrared characteristic image of the target electrical equipment from the fire information database and send the infrared characteristic image to the infrared image analysis unit; After receiving the infrared characteristic image, the infrared image analysis unit immediately performs regional fire monitoring feedback analysis on the infrared characteristic image, compares and analyzes the obtained temperature fluctuation index, and obtains a feedback signal or a normal temperature signal; The area recognition unit is used to respond to the feedback signal, and further perform regional fire recognition and early warning analysis on the infrared feature image, and divide it into risk area and normal area, further analyze the risk area, perform discriminant analysis on the obtained gray deviation risk index, and obtain gray normal signal or alarm signal; The feature matching unit is used to respond to the feedback signal and collect the monitoring video image of the target electrical equipment, and perform feature matching recognition and early warning feedback operation on the monitoring video image to obtain a flame signal or a smoke self-detection signal. When the smoke self-detection signal is generated, smoke recognition analysis is performed to obtain a fire alarm signal; At the same time, the generated alarm signal, fire alarm signal or flame signal is processed, and if at least one of the alarm signal, fire alarm signal or flame signal is generated, a fire warning signal is obtained; The dynamic management and control unit is used to respond to fire warning signals and collect continuous frame infrared feature images of risk areas, conduct dynamic risk trend assessment and analysis on the continuous frame infrared feature images, and send the obtained fire dynamic out-of-control level DJ to the alarm response unit.

[0005] Preferably, the regional fire monitoring feedback analysis process of the infrared image analysis unit is as follows: The target electrical equipment operation time period is collected, and the target electrical equipment operation time period is set as the time threshold, g fire monitoring points are set inside the target electrical equipment, g is a natural number greater than zero, an infrared characteristic image corresponding to the initial operation time of the target electrical equipment within the time threshold is obtained, and the infrared characteristic image corresponding to the initial operation time of the target electrical equipment is set as the reference image; The infrared characteristic image of the target electrical equipment within the time threshold is acquired in real time, and then the temperature floating index of the same fire monitoring point of the infrared characteristic image of the target electrical equipment and the reference image is acquired, and the temperature floating index is compared and analyzed with the preset temperature floating index threshold to generate a feedback signal or a normal temperature signal.

[0006] Preferably, the regional fire identification and warning analysis process of the regional identification unit is as follows: A monitoring area of ​​a fire monitoring point in a target electrical device corresponding to a feedback signal is obtained, and an infrared characteristic image of the monitoring area of ​​the fire monitoring point in the target electrical device is obtained at the same time, and the infrared characteristic image of the monitoring area of ​​the fire monitoring point in the target electrical device is set as a suspicious characteristic image, and the suspicious characteristic image is divided into i sub-region blocks, where i is a natural number greater than zero, and each sub-region block is preprocessed, and the preprocessing includes scaling and cropping, and then pixel values ​​of each sub-region block are obtained, and the pixel values ​​of each sub-region block are set as pixel risk assessment values, and the pixel risk assessment values ​​are discriminated and analyzed to obtain risk areas and normal areas.

[0007] Preferably, the grayscale change amount of the risk area within the time threshold is obtained, and the grayscale change amount of the risk area is set as the grayscale deviation risk index, and the grayscale deviation risk index is discriminated and analyzed to generate a grayscale normal signal or an alarm signal.

[0008] Preferably, the feature matching recognition warning feedback operation process of the feature matching unit is as follows: A monitoring video image of a monitoring area corresponding to a fire monitoring point in a target electrical device whose feedback signal is within a time threshold is obtained, and flame features are extracted from the processed monitoring video image by image processing technology. The image processing technology includes noise reduction and contrast enhancement. The matching degree between the flame features of the processed monitoring video image and a preset flame feature template is set as the flame matching degree, and the flame matching degree is discriminated and processed to generate a flame signal or a smoke self-detection signal.

[0009] Preferably, when a smoke self-test signal is generated, multiple groups of historical smoke video images of electrical equipment are obtained, and image processing is performed on the smoke video images, the image processing includes denoising, enhancement, and color correction, smoke features are extracted from the smoke video images after image processing, and a smoke recognition model is constructed through a deep learning algorithm, and the constructed smoke recognition model is set as a target smoke recognition model; At the same time, the percentage value corresponding to the total number of wrong identifications in the total number of identifications of the historical target smoke identification model is obtained, and the percentage value corresponding to the total number of wrong identifications in the total number of identifications of the historical target smoke identification model is set as the model credibility value, and the model credibility value is discriminated. If the model credibility value is less than the preset model credibility value threshold, a standard signal is generated. When the standard signal is generated, the monitoring video image of the monitoring area of ​​the fire monitoring point in the target electrical equipment is substituted into the target smoke identification model for identification and comparison: If smoke appears in the surveillance video image, a fire alarm signal is generated; If no smoke is present in the surveillance video image, no signal is generated.

[0010] Preferably, the dynamic risk trend assessment and analysis process of the dynamic management and control unit is as follows: The alarm time period corresponding to the alarm signal is obtained, and then the continuous frame infrared characteristic images of the risk area corresponding to the alarm signal in the alarm time period are obtained, and then the value obtained by subtracting the grayscale value of the previous frame of the infrared characteristic image from the grayscale value of the next frame of the infrared characteristic image is set as the grayscale dynamic value; The ratio between the number of grayscale dynamic values ​​less than the preset grayscale dynamic value threshold and the number of grayscale dynamic values ​​greater than or equal to the preset grayscale dynamic value threshold is set as the thermal runaway risk coefficient, and the thermal runaway risk coefficient is compared and analyzed with the stored preset thermal runaway risk coefficient range to obtain the first-level state, the second-level state and the third-level state, and the first-level state, the second-level state and the third-level state are set as the fire dynamic out-of-control level DJ, DJ=1, 2, 3.

[0011] A method for intelligently identifying electrical fires based on visual images is applied to an intelligently identifying electrical fires based on visual images, and includes the following steps: Step 1: Collect infrared characteristic images of target electrical equipment and conduct regional fire monitoring feedback analysis, compare and analyze the obtained temperature fluctuation index, if a feedback signal is obtained, proceed to step 2, if a normal temperature signal is obtained, feedback output; Step 2: Further perform regional fire identification and early warning analysis on the infrared feature image by means of information feedback to determine whether a fire has occurred in the target electrical equipment. If a normal grayscale signal is obtained, feedback output is performed; if an alarm signal is obtained, the process is substituted into step 5. Step 3: Collect monitoring video images of the target electrical equipment through information feedback and perform feature matching recognition and early warning feedback operations to determine whether the target electrical equipment has obvious fire characteristics. If a flame signal is obtained, proceed to step 5; if a smoke self-detection signal is obtained, proceed to step 4; Step 4: Perform smoke recognition analysis on the surveillance video image in a progressive manner. If a fire alarm signal is obtained, proceed to step 5; Step 5: Processing the generated alarm signal, fire alarm signal or flame signal, if at least one of the alarm signal, fire alarm signal or flame signal is generated, obtaining a fire warning signal and outputting it as feedback; Step 6: Based on the fire warning signal, the continuous frame infrared feature images of the risk area are analyzed to divide the thermal runaway risk coefficient, and then the divided fire dynamic runaway level DJ is fed back and output.

[0012] The beneficial effects of the present invention are as follows: The present invention performs preliminary regional fire monitoring feedback analysis on infrared feature images, which helps to understand the suspicious risk situation of each monitoring point in the target electrical equipment, so as to provide data support for subsequent analysis, and analyzes whether a fire occurs in the target electrical equipment from two perspectives: infrared feature image processing and monitoring video image feature extraction. The infrared feature image is further analyzed for regional fire identification and early warning by means of information feedback, so as to timely remind management personnel to manage the target electrical equipment, and at the same time, it helps to provide timely and accurate early warnings in the early stages of a fire. The present invention performs feature matching, identification, warning and feedback operations on monitoring video images so as to perform fire identification and alarm from the perspective of feature matching, thereby helping to improve the fire identification efficiency of target electrical equipment. The combination of the two fire analysis methods helps to reduce the fire identification error rate and conduct in-depth analysis of target electrical equipment in disaster situations so as to perform reasonable management and control based on the dynamic change trend of the fire, so as to minimize the impact of the fire and help improve the efficiency of fire management. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] The present invention will be further described below in conjunction with the accompanying drawings; Figure 1 It is a flowchart of the system of the present invention; Figure 2 It is a reference analysis diagram of the method of the present invention. DETAILED DESCRIPTION

[0014] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0015] Embodiment 1 See also Figure 1 to Figure 2 As shown, the present invention is an intelligent electrical fire recognition system based on visual images, including a visual image fire recognition platform, a fire information database, an infrared image analysis unit, a region recognition unit, a feature matching unit, a dynamic management and control unit and an alarm response unit. The fire information database is connected to the visual image fire recognition platform in a one-way communication manner, the visual image fire recognition platform is connected to the infrared image analysis unit in a one-way communication manner, the infrared image analysis unit is connected to the region recognition unit and the feature matching unit in a one-way communication manner, the region recognition unit and the feature matching unit are connected to the alarm response unit in a one-way communication manner, the region recognition unit is connected to the feature matching unit in a one-way communication manner, the feature matching unit is connected to the dynamic management and control unit in a one-way communication manner, and the dynamic management and control unit is connected to the alarm response unit in a one-way communication manner.

[0016] The visual image fire identification platform is used to retrieve the infrared characteristic image of the target electrical equipment from the fire information database and send the infrared characteristic image to the infrared image analysis unit. After receiving the infrared characteristic image, the infrared image analysis unit immediately performs regional fire monitoring feedback analysis on the infrared characteristic image, which helps to understand the suspicious risk situation of each monitoring point in the target electrical equipment, so as to provide data support for subsequent analysis. The specific regional fire monitoring feedback analysis process is as follows: The target electrical equipment operation time period is collected, and the target electrical equipment operation time period is set as the time threshold, g fire monitoring points are set inside the target electrical equipment, g is a natural number greater than zero, an infrared characteristic image corresponding to the initial operation time of the target electrical equipment within the time threshold is obtained, and the infrared characteristic image corresponding to the initial operation time of the target electrical equipment is set as the reference image; Acquire the infrared characteristic image of the target electrical equipment within the time threshold in real time, and then obtain the temperature fluctuation index of the same fire monitoring point of the infrared characteristic image of the target electrical equipment and the reference image; In the embodiment of the present invention, the temperature fluctuation index represents the value obtained by subtracting the temperature value of the fire monitoring point in the reference image from the temperature value of the fire monitoring point in the infrared characteristic image. It should be noted that the larger the value of the temperature fluctuation index is, the greater the fire risk of the fire monitoring point in the target electrical equipment is. Compare and analyze the temperature fluctuation index with the preset temperature fluctuation index threshold: If the temperature fluctuation index is greater than or equal to a preset temperature fluctuation index threshold, a feedback signal is generated; If the temperature floating index is less than the preset temperature floating index threshold, a normal temperature signal is generated, and the normal temperature signal or feedback signal is sent to the alarm response unit via the area identification unit. After receiving the normal temperature signal or feedback signal, the alarm response unit immediately displays the preset warning text corresponding to the normal temperature signal or feedback signal, so as to intuitively understand the monitoring status of each monitoring point.

[0017] Embodiment 2 When a feedback signal is generated, the regional identification unit is used to respond to the feedback signal and further perform regional fire identification and early warning analysis on the infrared feature image to determine whether a fire occurs in the target electrical equipment, so as to promptly remind the management personnel to manage the target electrical equipment, and at the same time help to provide timely and accurate early warning in the early stage of the fire. The specific regional fire identification and early warning analysis process is as follows: The monitoring area of ​​the fire monitoring point in the target electrical equipment corresponding to the feedback signal is obtained, and the infrared characteristic image of the monitoring area of ​​the fire monitoring point in the target electrical equipment is obtained at the same time, and the infrared characteristic image of the monitoring area of ​​the fire monitoring point in the target electrical equipment is set as a suspicious characteristic image, and the suspicious characteristic image is divided into i sub-region blocks, i is a natural number greater than zero, and each sub-region block is pre-processed, and the pre-processing includes scaling, cropping, etc., and then the pixel value of each sub-region block is obtained, and the pixel value of each sub-region block is set as a pixel risk assessment value, and the pixel risk assessment value is discriminated and analyzed: If the pixel risk assessment value is greater than or equal to the preset pixel risk assessment value threshold, the corresponding area is determined to be a risk area; If the pixel risk assessment value is less than the preset pixel risk assessment value threshold, the corresponding area is judged as a normal area; The grayscale change amount of the risk area within the time threshold is further obtained, and the grayscale change amount of the risk area is set as the grayscale deviation risk index, and the grayscale deviation risk index is discriminated and analyzed: If the grayscale deviation risk index is less than the preset grayscale deviation risk index threshold, a grayscale normal signal is generated. After receiving the grayscale normal signal, the alarm response unit immediately performs the preset warning operation corresponding to the grayscale normal signal to reduce the risk of misjudgment of fire identification of the target electrical equipment; If the grayscale deviation risk index is greater than or equal to the preset grayscale deviation risk index threshold, an alarm signal is generated; When a feedback signal is generated, the feature matching unit is used to respond to the feedback signal and collect the monitoring video image of the target electrical equipment, and perform feature matching recognition warning feedback operation on the monitoring video image to determine whether the target electrical equipment has obvious fire features, so as to perform fire recognition alarm from the perspective of feature matching, thereby helping to improve the fire recognition efficiency of the target electrical equipment. The specific feature matching recognition warning feedback operation process is as follows: The monitoring video image of the monitoring area corresponding to the fire monitoring point in the target electrical equipment in the feedback signal within the time threshold is obtained, and the flame feature is extracted from the processed monitoring video image by image processing technology. The image processing technology includes noise reduction, contrast enhancement, etc., and the matching degree between the flame feature of the processed monitoring video image and the preset flame feature template is obtained. The matching degree between the flame feature of the processed monitoring video image and the preset flame feature template is set as the flame matching degree, and the flame matching degree is judged: If the flame matching degree is greater than or equal to a preset flame matching degree threshold, a flame signal is generated; If the flame matching degree is less than the preset flame matching degree threshold, a smoke self-test signal is generated; When a smoke self-test signal is generated, multiple groups of historical smoke video images of electrical equipment are obtained, and image processing is performed on the smoke video images. The image processing includes denoising, enhancement, color correction, etc. Smoke features are extracted from the smoke video images after image processing, and a smoke recognition model is constructed through a deep learning algorithm, and the constructed smoke recognition model is set as the target smoke recognition model; In the embodiment of the present invention, the smoke characteristics include texture, shape, etc.; In the embodiment of the present invention, the deep learning algorithm includes a convolutional neural network (CNN), etc. At the same time, the percentage value corresponding to the total number of wrong identifications in the total number of identifications of the historical target smoke identification model is obtained, and the percentage value corresponding to the total number of wrong identifications in the total number of identifications of the historical target smoke identification model is set as the model credibility value, and the model credibility value is discriminated. If the model credibility value is less than the preset model credibility value threshold, a standard signal is generated. When the standard signal is generated, the monitoring video image of the monitoring area of ​​the fire monitoring point in the target electrical equipment is substituted into the target smoke identification model for identification and comparison: If smoke appears in the surveillance video image, a fire alarm signal is generated; If no smoke appears in the surveillance video image, no signal is generated; The generated alarm signal or fire alarm signal or flame signal is processed. If at least one of the alarm signal or fire alarm signal or flame signal is generated, a fire warning signal is obtained, and the fire warning signal is sent to the alarm response unit. After receiving the fire warning signal, the alarm response unit immediately performs a preset warning operation corresponding to the fire warning signal, that is, the preset warning operation corresponding to the fire warning signal is: displaying the number and coordinates of the target electrical equipment, so as to promptly remind the management personnel to manage the target electrical equipment, and at the same time help to provide timely and accurate warning in the early stage of a fire.

[0018] Embodiment 3 When a fire warning signal is generated, the dynamic control unit is used to respond to the fire warning signal and collect continuous frame infrared feature images of the risk area, and conduct disaster dynamic risk trend assessment and analysis on the continuous frame infrared feature images, so as to carry out reasonable control and control scheme matching according to the dynamic change trend of the fire, so as to minimize the impact of the fire and help improve the efficiency of fire management. The specific disaster dynamic risk trend assessment and analysis process is as follows: The alarm time period corresponding to the alarm signal is obtained, and then the continuous frame infrared feature images of the risk area corresponding to the alarm signal within the alarm time period are obtained, and the gray value of each frame of the infrared feature image is obtained, and then the gray value of the next frame of infrared feature image minus the gray value of the previous frame of infrared feature image is obtained, and the value obtained by subtracting the previous frame of infrared feature image from the next frame of infrared feature image is set as the gray dynamic value. It should be noted that the larger the gray dynamic value, the greater the regional fire tendency and the higher the fire control demand; The number of grayscale dynamic values ​​greater than or equal to the preset grayscale dynamic value threshold and the number of grayscale dynamic values ​​less than the preset grayscale dynamic value threshold are obtained, and the ratio between the number of grayscale dynamic values ​​less than the preset grayscale dynamic value threshold and the number of grayscale dynamic values ​​greater than or equal to the preset grayscale dynamic value threshold is set as the thermal runaway risk coefficient, and the thermal runaway risk coefficient is compared and analyzed with the stored preset thermal runaway risk coefficient range: If the thermal runaway risk coefficient is greater than the maximum value in the preset thermal runaway risk coefficient range, it is judged as a level 1 state; If the thermal runaway risk coefficient falls within the preset thermal runaway risk coefficient range, it is judged as the second level state; If the thermal runaway risk coefficient is less than the minimum value in the preset thermal runaway risk coefficient range, it is judged as the third-level state, among which the fire out of control risks corresponding to the first-level state, the second-level state and the third-level state are abnormally increased, and the first-level state, the second-level state and the third-level state are set as the fire dynamic out of control level DJ, DJ=1, 2, 3, that is, when the fire dynamic out of control level DJ=1, it indicates the first-level state, when the fire dynamic out of control level DJ=2, it indicates the second-level state, and when the fire dynamic out of control level DJ=3, it indicates the third-level state. The fire dynamic out of control level DJ is sent to the alarm response unit. After receiving the fire dynamic out of control level DJ, the alarm response unit immediately matches the fire control plan corresponding to the current fire dynamic out of control level DJ, so as to perform reasonable control according to the dynamic change trend of the fire, so as to minimize the impact of the fire, and at the same time help to improve the efficiency of fire management.

[0019] Embodiment 4 An intelligent identification method for electrical fires based on visual images, such as Figure 2 As shown, the following steps are included: Step 1: Collect infrared characteristic images of target electrical equipment and conduct regional fire monitoring feedback analysis, compare and analyze the obtained temperature fluctuation index, if a feedback signal is obtained, proceed to step 2, if a normal temperature signal is obtained, feedback output; Step 2: Further perform regional fire identification and early warning analysis on the infrared feature image by means of information feedback to determine whether a fire has occurred in the target electrical equipment. If a normal grayscale signal is obtained, feedback output is performed; if an alarm signal is obtained, the process is substituted into step 5. Step 3: Collect monitoring video images of the target electrical equipment through information feedback and perform feature matching recognition and early warning feedback operations to determine whether the target electrical equipment has obvious fire characteristics. If a flame signal is obtained, proceed to step 5; if a smoke self-detection signal is obtained, proceed to step 4; Step 4: Perform smoke recognition analysis on the surveillance video image in a progressive manner. If a fire alarm signal is obtained, proceed to step 5; Step 5: Processing the generated alarm signal, fire alarm signal or flame signal, if at least one of the alarm signal, fire alarm signal or flame signal is generated, obtaining a fire warning signal and outputting it as feedback; Step 6: Analyze the continuous frame infrared characteristic images of the risk area under the fire warning signal to divide the thermal runaway risk coefficient, and then feedback and output the divided fire dynamic runaway level DJ; In summary, the present invention helps to understand the suspicious risk situation of each monitoring point in the target electrical equipment through preliminary regional fire monitoring feedback analysis of infrared feature images, so as to provide data support for subsequent analysis, and analyzes whether a fire has occurred in the target electrical equipment from two perspectives: infrared feature image processing and monitoring video image feature extraction. Regional fire identification and early warning analysis is further performed on the infrared feature image by means of information feedback, so as to timely remind management personnel to manage the target electrical equipment, and at the same time help to provide timely and accurate early warnings in the early stages of a fire. Feature matching and identification early warning feedback operations are performed on monitoring video images, so as to perform fire identification and alarm from the perspective of feature matching, thereby helping to improve the fire identification efficiency of the target electrical equipment. The combination of the two fire analysis methods helps to reduce the fire identification error rate, and in-depth analysis is performed on the target electrical equipment in disaster situation, so as to carry out reasonable management and control based on the dynamic change trend of the fire, so as to minimize the impact of the fire, and at the same time help to improve the efficiency of fire management.

[0020] The threshold is set to facilitate comparison. The size of the threshold depends on the amount of sample data and the number of bases set by technicians in this field for each group of sample data; as long as it does not affect the proportional relationship between the parameter and the quantized value.

[0021] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can make equivalent replacements or changes according to the technical scheme and inventive concept of the present invention within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. An intelligent electrical fire identification system based on visual images, characterized in that: It includes a visual image fire recognition platform, a fire information database, an infrared image analysis unit, a region recognition unit, a feature matching unit, a dynamic control unit, and an alarm response unit; The visual image fire recognition platform is used to retrieve the infrared characteristic image of the target electrical equipment from the fire information database and send the infrared characteristic image to the infrared image analysis unit; After receiving the infrared characteristic image, the infrared image analysis unit immediately performs regional fire monitoring feedback analysis on the infrared characteristic image, compares and analyzes the obtained temperature fluctuation index, and obtains a feedback signal or a normal temperature signal; The area recognition unit is used to respond to the feedback signal, and further perform regional fire recognition and early warning analysis on the infrared feature image, and divide it into risk area and normal area, further analyze the risk area, perform discriminant analysis on the obtained gray deviation risk index, and obtain gray normal signal or alarm signal; The feature matching unit is used to respond to the feedback signal and collect the monitoring video image of the target electrical equipment, and perform feature matching recognition and early warning feedback operation on the monitoring video image to obtain a flame signal or a smoke self-detection signal. When the smoke self-detection signal is generated, smoke recognition analysis is performed to obtain a fire alarm signal; At the same time, the generated alarm signal, fire alarm signal or flame signal is processed, and if at least one of the alarm signal, fire alarm signal or flame signal is generated, a fire warning signal is obtained; The dynamic management and control unit is used to respond to fire warning signals and collect continuous frame infrared feature images of risk areas, conduct dynamic risk trend assessment and analysis on the continuous frame infrared feature images, and send the obtained fire dynamic out-of-control level DJ to the alarm response unit.

2. The electrical fire intelligent identification system based on visual images according to claim 1 is characterized in that: The regional fire monitoring feedback analysis process of the infrared image analysis unit is as follows: The target electrical equipment operation time period is collected, and the target electrical equipment operation time period is set as the time threshold, g fire monitoring points are set inside the target electrical equipment, g is a natural number greater than zero, an infrared characteristic image corresponding to the initial operation time of the target electrical equipment within the time threshold is obtained, and the infrared characteristic image corresponding to the initial operation time of the target electrical equipment is set as the reference image; The infrared characteristic image of the target electrical equipment within the time threshold is acquired in real time, and then the temperature floating index of the same fire monitoring point of the infrared characteristic image of the target electrical equipment and the reference image is acquired, and the temperature floating index is compared and analyzed with the preset temperature floating index threshold to generate a feedback signal or a normal temperature signal.

3. The electrical fire intelligent identification system based on visual images according to claim 2 is characterized in that: The regional fire identification and warning analysis process of the regional identification unit is as follows: A monitoring area of ​​a fire monitoring point in a target electrical device corresponding to a feedback signal is obtained, and an infrared characteristic image of the monitoring area of ​​the fire monitoring point in the target electrical device is obtained at the same time, and the infrared characteristic image of the monitoring area of ​​the fire monitoring point in the target electrical device is set as a suspicious characteristic image, and the suspicious characteristic image is divided into i sub-region blocks, where i is a natural number greater than zero, and each sub-region block is preprocessed, and the preprocessing includes scaling and cropping, and then pixel values ​​of each sub-region block are obtained, and the pixel values ​​of each sub-region block are set as pixel risk assessment values, and the pixel risk assessment values ​​are discriminated and analyzed to obtain risk areas and normal areas.

4. The electrical fire intelligent identification system based on visual images according to claim 3 is characterized in that: The grayscale change amount of the risk area within the time threshold is obtained, and the grayscale change amount of the risk area is set as the grayscale deviation risk index, and the grayscale deviation risk index is discriminated and analyzed to generate a grayscale normal signal or an alarm signal.

5. The electrical fire intelligent identification system based on visual images according to claim 4 is characterized in that: The feature matching recognition warning feedback operation process of the feature matching unit is as follows: A monitoring video image of a monitoring area corresponding to a fire monitoring point in a target electrical device whose feedback signal is within a time threshold is obtained, and flame features are extracted from the processed monitoring video image by image processing technology. The image processing technology includes noise reduction and contrast enhancement. The matching degree between the flame features of the processed monitoring video image and a preset flame feature template is set as the flame matching degree, and the flame matching degree is discriminated and processed to generate a flame signal or a smoke self-detection signal.

6. The electrical fire intelligent identification system based on visual images according to claim 5 is characterized in that: When a smoke self-test signal is generated, multiple groups of historical smoke video images of electrical equipment are obtained, and image processing is performed on the smoke video images. The image processing includes denoising, enhancement, and color correction. Smoke features are extracted from the smoke video images after image processing, and a smoke recognition model is constructed through a deep learning algorithm. The constructed smoke recognition model is set as a target smoke recognition model. At the same time, the percentage value corresponding to the total number of wrong identifications in the total number of identifications of the historical target smoke identification model is obtained, and the percentage value corresponding to the total number of wrong identifications in the total number of identifications of the historical target smoke identification model is set as the model credibility value, and the model credibility value is discriminated. If the model credibility value is less than the preset model credibility value threshold, a standard signal is generated. When the standard signal is generated, the monitoring video image of the monitoring area of ​​the fire monitoring point in the target electrical equipment is substituted into the target smoke identification model for identification and comparison: If smoke appears in the surveillance video image, a fire alarm signal is generated; If no smoke is present in the surveillance video image, no signal is generated.

7. The electrical fire intelligent identification system based on visual images according to claim 6 is characterized in that: The dynamic risk trend assessment and analysis process of the dynamic control unit is as follows: The alarm time period corresponding to the alarm signal is obtained, and then the continuous frame infrared characteristic images of the risk area corresponding to the alarm signal in the alarm time period are obtained, and then the value obtained by subtracting the grayscale value of the previous frame of the infrared characteristic image from the grayscale value of the next frame of the infrared characteristic image is set as the grayscale dynamic value; The ratio between the number of grayscale dynamic values ​​less than the preset grayscale dynamic value threshold and the number of grayscale dynamic values ​​greater than or equal to the preset grayscale dynamic value threshold is set as the thermal runaway risk coefficient, and the thermal runaway risk coefficient is compared and analyzed with the stored preset thermal runaway risk coefficient range to obtain the first-level state, the second-level state and the third-level state, and the first-level state, the second-level state and the third-level state are set as the fire dynamic out-of-control level DJ, DJ=1, 2, 3.

8. A method for intelligent identification of electrical fires based on visual images, applied to the intelligent identification system for electrical fires based on visual images as claimed in any one of claims 1 to 7, characterized in that: The following steps are involved: Step 1: Collect infrared characteristic images of target electrical equipment and conduct regional fire monitoring feedback analysis, compare and analyze the obtained temperature fluctuation index, if a feedback signal is obtained, proceed to step 2, if a normal temperature signal is obtained, feedback output; Step 2: Further perform regional fire identification and early warning analysis on the infrared feature image by means of information feedback to determine whether a fire has occurred in the target electrical equipment. If a normal grayscale signal is obtained, feedback output is performed; if an alarm signal is obtained, the process is substituted into step 5. Step 3: Collect monitoring video images of the target electrical equipment through information feedback and perform feature matching recognition and early warning feedback operations to determine whether the target electrical equipment has obvious fire characteristics. If a flame signal is obtained, proceed to step 5; if a smoke self-detection signal is obtained, proceed to step 4; Step 4: Perform smoke recognition analysis on the surveillance video image in a progressive manner. If a fire alarm signal is obtained, proceed to step 5; Step 5: Processing the generated alarm signal, fire alarm signal or flame signal, if at least one of the alarm signal, fire alarm signal or flame signal is generated, obtaining a fire warning signal and outputting it as feedback; Step 6: Based on the fire warning signal, the continuous frame infrared feature images of the risk area are analyzed to divide the thermal runaway risk coefficient, and then the divided fire dynamic runaway level DJ is fed back and output.

Images