A multi-mode multi-channel online monitoring device for switch cabinet

Abstract

The application relates to the technical field of online monitoring, in particular to a switch cabinet multi-mode multi-channel online monitoring device which comprises a video data acquisition module, a switch cabinet component identification module, a multi-channel image fusion module, a device state monitoring module and an abnormal picture early warning module. In the application, the imaging integrity is improved, the quality evaluation comprehensiveness is guaranteed, the component contour is extracted and the temperature distribution is analyzed, the structure is analyzed in combination with multi-source data, the identification precision is improved, the video fusion proportion is dynamically adjusted, the component definition is improved, the visual consistency in a complex environment is guaranteed, the image and temperature and shape data are fused, the abnormal heating and loosening are accurately detected, the suspicious area is marked in real time, the automatic marking and data analysis strengthen the abnormal visualization and evaluation, and timely early warning is realized. The multi-channel fusion combines the shape and temperature monitoring to form a closed loop optimization.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of online monitoring, in particular to a switch cabinet multi-mode multi-channel online monitoring device. BACKGROUND

[0002] The technical field of online monitoring includes real-time monitoring and abnormal detection of the operating state of power equipment, and its core content is to ensure the safe and stable operation of the power system through remote data acquisition and analysis means. This technical field involves dynamic monitoring of internal environmental parameters of switch cabinets, including key indicators such as displacement of mechanical components, contact temperature, arc activity, and insulation state, which requires the combination of video image acquisition and sensor data fusion technology to construct a complete visual information chain of equipment operating state, systematically covering data synchronization transmission, abnormal feature extraction, and risk warning functions.

[0003] Among them, a switch cabinet multi-mode multi-channel online monitoring device refers to a device based on a combination of fixed cameras and infrared thermal imagers. This device continuously captures the action trajectory of circuit breakers and the contact state inside the switch cabinet through video streaming, and synchronously superimposes real-time values collected by temperature sensors onto the video picture. The specific technical means include pre-setting multiple-angle camera units inside the cabinet to cover the operating mechanism and busbar area, using image grayscale value comparison to analyze the degree of contact oxidation, dynamically matching temperature data with video frames through timestamp alignment mechanism, and triggering local picture enlargement and data marking functions according to pre-set threshold values.

[0004] The existing technology has the problem of insufficient adaptability to changes in light intensity in terms of video data acquisition and processing, resulting in significant environmental light interference on imaging quality and difficulty in ensuring the stability of video signals in complex lighting environments. In the component recognition process, only a single edge detection method is relied upon, ignoring the role of thermal signal distribution patterns, which affects the accuracy of component recognition and makes it difficult to accurately match the internal structure of the equipment. The fusion of visible light and infrared video fails to dynamically adjust the fusion ratio based on component visibility, resulting in distortion of image information under different lighting environments, affecting the visibility of key components, and thus reducing the reliability of equipment state monitoring. State monitoring mainly relies on temperature sensor values, lacking comprehensive analysis of shape changes, which can easily cause missed detection or misjudgment of abnormal conditions. Abnormal picture warning is triggered based on set thresholds only, without in-depth analysis of the trend of abnormal areas, resulting in a lack of flexibility in the warning mechanism and difficulty in timely responding to potential faults. Overall, the existing technology has obvious shortcomings in terms of light adaptability, component recognition accuracy, image fusion strategy, comprehensive abnormal detection, and warning mechanism flexibility, affecting the accuracy and reliability of equipment monitoring. SUMMARY

[0005] The purpose of the present application is to solve the shortcomings of the prior art and to provide a switch cabinet multi-mode multi-channel online monitoring device.

[0006] To achieve the above object, the application adopts the following technical scheme: a switch cabinet multi-mode multi-channel online monitoring device, comprising:

[0007] The video data acquisition module acquires switch cabinet video, calls an ambient light sensor to measure light intensity, calculates the brightness uniformity of a visible light picture, analyzes the thermal signal coverage range of an infrared video, evaluates the imaging integrity of the two types of videos, and forms a video signal quality evaluation result;

[0008] The switch cabinet component recognition module calls the video signal quality evaluation result, detects edge features in the visible light video, extracts contour information of internal components of the switch cabinet, analyzes temperature distribution patterns in the infrared video, recognizes thermal source features of key components, and combines analysis of internal structures of the two types of videos to form a component recognition result;

[0009] The multi-channel image fusion module calls the component recognition result, dynamically fuses the visible light and infrared videos, adjusts the fusion proportion of difference areas based on the visibility of components, and generates a fused image data stream;

[0010] The device state monitoring module acquires the fused image data stream, analyzes the temperature change trend of infrared signals during device operation, combines component morphological changes in the visible light image, detects state abnormalities, and real-time highlights suspicious areas on a monitoring interface, and outputs a device state monitoring result.

[0011] As a further scheme of the application, the video signal quality evaluation result includes brightness uniformity, thermal signal coverage range, and imaging integrity, the component recognition result includes component contour information, temperature distribution pattern, thermal source feature, and device internal structure, the fused image data stream includes image fusion proportion and imaging effect, and the device state monitoring result includes temperature change trend, component morphological change, state abnormality, and suspicious area.

[0012] As a further scheme of the application, the video data acquisition module comprises:

[0013] The video signal acquisition submodule acquires visible light video and infrared video data of the switch cabinet, calls an ambient light sensor to measure ambient light intensity values, and generates visible light video data, infrared video data, and ambient light intensity values;

[0014] The light analysis submodule calculates the mean and variance of picture pixel brightness based on the visible light video data, infrared video data, and ambient light intensity values, analyzes the thermal signal intensity distribution range and continuity in the infrared video data, and generates brightness uniformity values and thermal signal coverage states;

[0015] The integrity assessment submodule calls the brightness equalization value and thermal signal coverage status, compares the preset brightness deviation standard with the thermal signal coverage standard, assesses the integrity of visible light and infrared video imaging, and generates video signal quality assessment results.

[0016] As a further aspect of the present invention, the specific calculation formula for the distribution range and continuity of thermal signal intensity in the analyzed infrared video data is as follows:

[0017] ;

[0018] Calculate the continuity characteristic value of the thermal signal distribution, and generate the brightness equalization value and thermal signal coverage status;

[0019] in, Representing the Continuity characteristics of thermal signal distribution in frame infrared images. Representing the The number of pixels in a frame identified as hot signal regions. Representing the The pixel in the thermal signal region is at the The heat intensity value in the frame, Representing the The average thermal intensity of all pixels in the thermal signal region of the frame. Representing the The average thermal intensity of all pixels in the thermal signal region of the frame. Representing the The standard deviation of the brightness of all pixels in the frame. Representing the The average brightness of all pixels in the frame. To prevent extremely small constants with a denominator of zero.

[0020] As a further aspect of the present invention, the switch cabinet component identification module includes:

[0021] The edge contour extraction submodule calls the visible light video frame brightness data and signal-to-noise ratio data in the video signal quality assessment results, detects the gradient change amplitude of the video frame edge, filters edge points according to the gradient direction and preset standards, connects adjacent points to form a closed boundary, extracts the geometric shape information of the switch cabinet components and generates a component contour feature set.

[0022] The heat source correlation analysis submodule calls the geometric boundary range of the component contour feature set to limit the detection range of abnormal temperature difference areas. By analyzing the temperature distribution data of infrared video frames, it calculates the temperature difference between the target area and the surrounding area, filters abnormal temperature rise areas according to preset thresholds, records the coordinates and changing trends of the highest temperature point, and generates a heat source distribution map in combination with the component contour.

[0023] The structural analysis submodule calls the temperature distribution data of the heat source distribution map and the geometric boundary information of the component contour feature set, spatially aligns the center point of the component boundary with the highest temperature point of the heat source, calculates whether the positional deviation between the two is less than the preset spatial tolerance range, filters the component heat source association groups that meet the conditions, and generates the component recognition result based on the contour shape and heat source distribution pattern in the association group.

[0024] As a further aspect of the present invention, the multi-channel image fusion module includes:

[0025] The component recognition submodule calls the edge contour features and texture features in the component recognition results, obtains the input frame sequence of visible light image and infrared image, extracts the region coordinates of the same component in visible light and infrared image, compares the edge sharpness value and texture complexity value of the corresponding component in visible light and infrared image, and filters to obtain the component difference region coordinate set;

[0026] The difference allocation submodule calculates the mean and standard deviation of grayscale values ​​of the corresponding region in the visible light image based on the coordinate set of the difference region of the component, and simultaneously calculates the mean and contrast values ​​of thermal radiation intensity of the corresponding region in the infrared image. It performs a difference operation on the mean of visible light grayscale values ​​and the mean of infrared thermal radiation intensity, and determines the fusion ratio by combining the ratio of the standard deviation and the contrast value, thereby generating a set of difference region fusion ratios.

[0027] The dynamic fusion submodule calls the coordinate set of the component difference region and the fusion ratio set of the difference region, performs pixel-by-pixel superposition operation on the difference regions of the visible light image and the infrared image, and fuses the non-difference regions according to the preset ratio. It integrates all region results and stitches them into a continuous frame sequence to generate a fused image data stream.

[0028] As a further aspect of the present invention, the specific calculation formula for comparing the edge sharpness value and texture complexity value of corresponding components in visible light and infrared images is as follows:

[0029] ;

[0030] Calculate the composite index values ​​of the features, and filter to obtain the coordinate set of component difference areas; among them... Representative features are compared with composite index values. Representing the visible light image Edge sharpness value per pixel, Representing the infrared image Edge sharpness value per pixel, Representing the visible light image Texture complexity value per pixel, Representing the infrared image Texture complexity value per pixel, The total number of pixels representing edge features involved in the comparison. This represents the total number of pixels involved in the calculation of texture features.

[0031] As a further aspect of the present invention, the device status monitoring module includes:

[0032] The image fusion submodule monitors the infrared image temperature sequence and visible light image frame sequence during the operation of the monitoring device based on the fused image data stream. It aligns the temperature value of each pixel in the infrared image with the spatial coordinates of the corresponding area in the visible light image on the time axis, and superimposes the temperature data onto the visible light image coordinate system to generate multi-source image fusion data.

[0033] The temperature trend analysis submodule calls the temperature sequence in the multi-source image fusion data, extracts the temperature range value and the temperature change amplitude between adjacent frames within the continuous time window of the device area, compares the range value with the temperature fluctuation range corresponding to the preset device type, filters the range exceeding the limit area, calculates the statistical fluctuation range area where the change amplitude exceeds the benchmark change amplitude of the same type of device, and generates the temperature anomaly mark area.

[0034] The morphology association determination submodule extracts the component contour deformation of the temperature anomaly marked area, calculates the mean Euclidean distance of the contour point set between adjacent frames based on the visible light image frame sequence, matches the spatial coordinates of the area where the deformation exceeds the assembly tolerance range with the temperature anomaly area, marks the boundary of the area where both deformation exceeds the standard and temperature anomaly exists, and maps the coordinates of the temperature anomaly area to the monitoring interface to generate equipment status monitoring results.

[0035] As a further aspect of the present invention, the specific calculation formula for the mean Euclidean distance of the contour point sets between adjacent frames is as follows:

[0036] ;

[0037] in, The mean Euclidean distance between the sets of contour points in adjacent frames. Let Euclidean distance represent the Euclidean distance between the set of contour points between frame i and frame j. represents the average Euclidean distance between all contour point sets, and Q represents the total number of contour point sets calculated.

[0038] As a further aspect of the present invention, it also includes:

[0039] The abnormal screen warning module calls the device status monitoring results, automatically marks abnormal areas on the monitoring screen and adjusts the display mode. At the same time, it compares and analyzes the data of the abnormal areas, assesses the severity of the problem, and triggers abnormal screen warning information.

[0040] The abnormal picture early warning information includes abnormal area annotation, display mode adjustment, data comparison analysis, and evaluation of problem severity.

[0041] The abnormal picture early warning module includes:

[0042] The abnormal area annotation submodule obtains the device operation data set in the device state monitoring result, calls a preset reference comparison rule, performs item-by-item difference calculation on the partition values in the data set and the corresponding reference, filters the partitions whose differences exceed a set deviation threshold, maps the partitions to a monitoring picture based on partition coordinates, and generates abnormal area coordinates.

[0043] The display adjustment and comparison submodule calls the abnormal area coordinates, extracts real-time monitoring data streams in the abnormal area, adjusts the color saturation and flicker frequency in the picture display setting, synchronously obtains the normal data mean value in the same area operation period, performs segmented time sequence comparison on the real-time data stream and the data mean value, and generates an abnormal data comparison set.

[0044] The abnormal evaluation triggering submodule extracts the maximum deviation amplitude and the mean value of fluctuation frequency based on the abnormal data comparison set, combines the load degree and the operation time proportion in the device operation data set, calculates an abnormal deviation value, matches the deviation value with a preset early warning level threshold interval, triggers a corresponding level early warning instruction, and generates abnormal picture early warning information.

[0045] Compared with the prior art, the application has the following advantages and positive effects:

[0046] In the application, the ambient light sensor is combined with visible light and infrared video to analyze the brightness balance and the thermal signal coverage range, to improve the imaging integrity and ensure the comprehensiveness of quality evaluation. The component contour is extracted and the temperature distribution is analyzed, the structure is analyzed combined with multi-source data, the recognition accuracy is improved, the video fusion proportion is dynamically adjusted, the component clarity is improved, the visual consistency in complex environment is ensured, the image, temperature and shape data are fused, the abnormal heating and loosening are accurately detected, the suspicious area is labeled in real time, the automatic labeling and data analysis strengthen the abnormal visualization and evaluation, and timely early warning is realized. The multi-channel fusion combined with shape and temperature monitoring forms a closed loop optimization, and the monitoring accuracy and detection efficiency are improved. BRIEF DESCRIPTION OF DRAWINGS

[0047] Figure 1 The system flowchart of the application is shown in the figure.

[0048] Figure 2 The flowchart of the application is shown in the figure. DETAILED DESCRIPTION

[0049] In order to make the purpose, technical scheme and advantages of the present application clearer, the present application will be further described in detail below with reference to the drawings and examples. It should be understood that the specific examples described herein are only used to explain the present application and not to limit the present application.

[0050] In the description of the present application, it should be understood that the terms "length", "width", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer" and the like indicate the orientation or positional relationship based on the orientation or positional relationship shown in the drawings, and are only for the convenience of describing the present application and simplifying the description, and do not indicate or imply that the devices or elements referred to must have a particular orientation, be constructed and operated in a particular orientation, and therefore cannot be understood as limiting the present application. In addition, in the description of the present application, the meaning of "a plurality of" is two or more, unless otherwise explicitly and specifically limited.

[0051] Please refer to Figure 1 A switch cabinet multi-mode multi-channel online monitoring device comprises:

[0052] The video data acquisition module acquires switch cabinet video, calls an ambient light sensor to measure light intensity, calculates the brightness uniformity of a visible light picture, analyzes the thermal signal coverage range of an infrared video, evaluates the imaging integrity of the two types of videos, and forms a video signal quality evaluation result;

[0053] The switch cabinet component identification module calls the video signal quality evaluation result, detects edge features in the visible light video, extracts contour information of internal components of the switch cabinet, simultaneously analyzes temperature distribution patterns in the infrared video, identifies thermal source features of key components, and combines analysis of internal structures of the two types of videos to form a component identification result;

[0054] The multi-channel image fusion module calls the component identification result, dynamically fuses the visible light and infrared videos, adjusts the fusion proportion of difference areas based on the visibility of components, optimizes the imaging effect in night or different light environments, and generates a fused image data stream;

[0055] The device state monitoring module acquires the fused image data stream, analyzes the temperature change trend of infrared signals in the device running process, combines component morphological changes in the visible light image, detects state abnormalities, and real-time highlights suspicious areas on the monitoring interface, and outputs a device state monitoring result;

[0056] The abnormal picture early warning module calls the device state monitoring result, automatically labels abnormal areas on the monitoring picture, adjusts the display mode, simultaneously compares and analyzes data of the abnormal areas, evaluates the severity of the problem, and triggers an abnormal picture early warning information.

[0057] The video signal quality evaluation result includes brightness uniformity, thermal signal coverage range, imaging integrity, the component recognition result includes component contour information, temperature distribution pattern, heat source feature, device internal structure, the fused image data stream includes image fusion proportion, imaging effect, the device state monitoring result includes temperature change trend, component morphology change, state anomaly, suspicious area, and the abnormal picture early warning information includes abnormal area labeling, display mode adjustment, data comparison analysis, and evaluation problem severity.

[0058] Please refer to Figure 2 The video data acquisition module includes:

[0059] The video signal acquisition sub-module acquires visible light video and infrared video data of the switch cabinet, calls an ambient light sensor to measure an ambient light intensity value, and generates visible light video data, infrared video data, and the ambient light intensity value.

[0060] First, the video acquisition terminal is started and positioned to the target area of the switch cabinet. The acquisition angle, acquisition frame rate, and focal length parameters are set through a control program to obtain clear visible light image data. Meanwhile, the infrared video acquisition channel is operated in parallel to record thermal imaging of the same area. The ambient light sensor module is started synchronously, and its installation position needs to be ensured to be unobstructed and as close to the surface of the photographed device as possible, for detecting the actual illumination level in the environment. In the environment with frequent light changes or complex light sources (such as strong light or backlight conditions in the factory), the sensor sampling frequency should be increased to more than 10 Hz to ensure continuous and effective measurement data. The collected data includes the light intensity value at the current time (unit: lux). If the detected illumination value is 350 lux at a certain time, this value is associated and marked in the video frame timestamp collected synchronously. Each set of visible light and infrared video data is labeled with the acquisition time and light value in the whole process, forming a complete data pair. For example, in the inspection of a certain switch cabinet, an infrared video and a visible light video are collected at 10:30 am, and the synchronous recorded ambient light intensity is 520 lux. The two videos will carry this light value for subsequent light analysis submodule processing. The final video data can be matched and marked with the ambient illumination for forming multi-dimensional visual data.

[0061] The light analysis submodule calculates the mean and variance of the picture pixel brightness based on the visible light video data, infrared video data, and ambient light intensity value, analyzes the thermal signal intensity distribution range and continuity in the infrared video data, and generates the brightness uniformity value and thermal signal coverage state.

[0062] The specific calculation formula for analyzing the thermal signal intensity distribution range and continuity in the infrared video data is as follows:

[0063] ;

[0064] The continuity characteristic value of the thermal signal distribution is calculated, and the brightness equalization value and the thermal signal coverage state are generated.

[0065] wherein, represents the number of pixels in the frame frame that are identified as thermal signal regions, represents the number of pixels in the frame frame that are identified as thermal signal regions, represents the number of pixels in the frame frame that are identified as thermal signal regions, represents the thermal intensity value of the pixel in the frame frame that is in the thermal signal region, represents the average thermal intensity value of all pixels in the frame frame that are in the thermal signal region, represents the average thermal intensity value of all pixels in the frame frame that are in the thermal signal region, represents the standard deviation of the brightness of all pixels in the frame represents the average brightness value of all pixels in the frame represents the average brightness value of all pixels in the frame is a very small constant to prevent the denominator from being zero;

[0066] Formula details and formula calculation derivation process:

[0067] This formula is used to calculate the continuity characteristic value of the thermal signal distribution in the frame frame infrared image . The following is the definition and derivation process of each parameter in the formula, including how to calculate each parameter:

[0068] Usually automatically identified by image processing algorithms from the thermal region in the infrared image and counted. Suppose 500 pixels are detected.

[0069] Data output from the infrared camera, the thermal intensity of each pixel is recorded. For example, select some specific pixel thermal intensity values as [200, 220, 230, 210, 205].

[0070] The calculation method is to sum the thermal intensity values of all thermal signal pixels and divide by . Using the previous pixel value, the average thermal intensity .

[0071] The calculation method is the same as . Suppose the average thermal intensity of the previous frame is 211.

[0072] and Obtained by analyzing the corresponding frame of visible light image. Set , .

[0073] The specific calculation process is as follows:

[0074] First, calculate the absolute value of the difference between each thermal signal pixel and the average thermal intensity, sum them, and then divide by... :

[0075] ;

[0076] Next, calculate the ratio of the standard deviation of luminance to the average luminance, and add a very small constant:

[0077] ;

[0078] Finally, calculate the square root of the average thermal intensity difference between the two frames:

[0079] ;

[0080] Substitute the values ​​of these three parts into In the formula:

[0081] ;

[0082] This result indicates that the first Compared to the first frame The frame exhibits significant stability in terms of the continuity of the thermal signal distribution. The calculation of this eigenvalue reflects the variability of the thermal signal and is directly related to the generation of the thermal signal coverage state.

[0083] The integrity assessment submodule calls the brightness equalization value and thermal signal coverage status, compares the preset brightness deviation standard with the thermal signal coverage standard, assesses the integrity of visible light and infrared video imaging, and generates video signal quality assessment results.

[0084] Firstly, the preset brightness deviation allowance range in the system is called, for example, an error of ±20% is allowed, the brightness average and deviation ratio in the visible light analysis result are read during processing, and the set standard value is compared. If the deviation ratio is within 15%, it is marked as qualified brightness, otherwise it is marked as abnormal brightness. The judgment of the thermal signal coverage state is based on the preset coverage continuity standard, which is usually set to more than 70% of the area in the picture being covered by the thermal signal, and the temperature difference between adjacent areas should not be greater than 5°C. If the system detection result shows that the thermal signal area only accounts for 45% of the picture, and there is a clear temperature difference fault, then its thermal coverage state is judged to be substandard. In an actual scene, for example, the brightness average deviates from the benchmark by 17% in the video analysis of the device daily inspection, the thermal signal coverage rate is 82%, but there are multiple temperature mutation areas. Then the video signal is judged to be poor in thermal continuity. The final video signal quality evaluation result generated by the system is "infrared incomplete, visible light edge brightness uneven", and is recorded in the inspection log for manual review and system optimization. The values involved in all comparison processes, such as "deviation standard" and "coverage standard", can be configured by operation and maintenance personnel according to different device scenes. For example, the lower limit of the thermal signal coverage allowed by some high-voltage switch cabinets can be increased to 60% to adapt to the thermal reflection interference situation under complex wiring structure.

[0085] Please refer to Figure 2 The switch cabinet component recognition module comprises:

[0086] The edge contour extraction submodule calls the visible light video frame brightness data and signal-to-noise ratio data in the video signal quality evaluation result, detects the edge gradient change amplitude of the video frame, filters the edge points according to the gradient direction and the preset standard, connects the adjacent points to form a closed boundary, extracts the geometric shape information of the switch cabinet component, and generates a component contour feature set;

[0087] First, the overall brightness level of each frame of image needs to be analyzed. In actual operation, the average method of the gray value of each pixel point of the image can be used to extract the frame brightness. For example, processing an image with a resolution of 640x480, by traversing the pixel gray value row by row and column by column, dividing the total pixel number after counting the total sum can obtain the current frame brightness value. If the average gray value of the frame is 125, it is recorded as the current frame brightness. Then the signal-to-noise ratio data corresponding to the frame is extracted. The difference between the frame image and the previous and next frames can be used for image denoising processing, and then the ratio of image noise and signal strength is calculated. Combined with the historical setting, such as the frame with a signal-to-noise ratio lower than 20 dB will be marked as a frame with poor quality. After completing the image quality evaluation, enter the edge detection stage. The commonly used algorithm such as Canny edge detection is adopted. First, the image is blurred with Gaussian filter, and then the gradient change value of the image in X and Y directions is extracted, and then the edge strength of each pixel point is judged. By setting the edge strength threshold value such as 80, the edge points with large changes are retained, and the gradient direction of these points is extracted. The direction here is the direction angle of the gray value change in the image. For example, if the direction angle of a certain point is 90 degrees, and the standard direction set by the system is horizontal or vertical, the point can be retained as an edge candidate. Continue to process all the candidate edge points, judge the pixel adjacency relationship between the points, connect the adjacent and direction continuous edge points into closed curves, and finally form several closed region boundaries. The geometric features such as the circumscribed rectangle of the contour, the contour area, the center coordinates, etc. are calculated. For example, a contour is identified, the width is 120 pixels, the height is 220 pixels, and the center coordinates are (210, 210). The feature is recorded and sorted as the geometric contour feature information of the circuit breaker component, and stored in the feature set for subsequent structure analysis and comparison operation.

[0088] The heat source correlation analysis submodule calls the geometric boundary range of the component contour feature set to limit the detection range of the temperature difference abnormal area. By analyzing the temperature distribution data of the infrared video frame, the temperature difference value between the target area and the surrounding area is calculated, the abnormal temperature rise area is selected according to the preset threshold, the highest temperature point coordinates and the change trend are recorded, and the heat source distribution map is generated combined with the component contour;

[0089] The boundary coordinate values of each target component are first read from the contour feature set, for example, the upper left corner coordinate of a component contour is (100, 150), and the lower right corner coordinate is (320, 270), and the rectangular area is the detection area range of the component. The system usually expands the detection range to prevent edge heat sources from being missed, and sets the expansion range to be 10 pixels per edge, and then the analysis area becomes (90, 140)-(330, 280). The system extracts the temperature value of each pixel point in the above-mentioned area from the infrared video frame, and calculates the average temperature of the target area pixels, assuming that the average temperature of the area is 65°C. At the same time, the average temperature of the background area is extracted by sampling 20 pixels around the periphery of the area, and is assumed to be 46°C, and then the temperature difference between the two is calculated to be 19°C. The system presets the temperature difference threshold reference switch cabinet component material, load level setting, for example, 15°C is set as the abnormal threshold for high-power circuit breakers, and this time 19°C has exceeded the threshold. Further confirm that the area is an abnormal temperature rise area, and then locate the highest temperature point in the target area, such as coordinate (210, 190), temperature 72.4°C. The system continues to track the temperature change of the point in the continuous video frame, and extracts the average temperature change trend in 10 frames, such as from 68°C in the first frame to 72.4°C in the tenth frame, the system determines that the temperature rise continues to occur, and integrates the point coordinate, maximum temperature and change trend information into the heat source feature information of the current component, and generates a component heat source distribution map in combination with the geometric contour position relationship.

[0090] The structure analysis submodule calls the temperature distribution data of the heat source distribution map and the geometric boundary information of the component contour feature set, aligns the center point of the component boundary with the highest temperature point of the heat source in space, calculates whether the position deviation is less than the preset spatial tolerance range, selects the component heat source association group that meets the condition, and generates a component recognition result according to the contour shape and heat source distribution law in the association group;

[0091] First, the center point of the component outline region is determined by calculating the midpoint of the left, right, up, and down coordinates of the bounding box. For example, if the upper left of the component boundary is (100, 150) and the lower right is (320, 270), then the center point is (210, 210). The system then reads the coordinates of the highest temperature point recorded in the heat source map corresponding to this component region, for example, (215, 205), and determines whether the spatial distance between these two points is within the allowable tolerance range. The tolerance range is set to within 15 pixels. The distance is calculated using the coordinates of the center point and the heat source point. If the result is 7 pixels, the matching condition is met, and the heat source point is bound to the outline. Subsequently, the system performs structural matching analysis on the bound heat source and the component boundary to check whether the heat source is located within a specific structural region of the outline. For example, if the component is a rectangular structure and the heat source point is located in the central region of its long side, it can be classified as "edge heating". By statistically analyzing the spatial distribution characteristics of heat sources, such as multiple heat source points being concentrated within 10% of the upper edge of the boundary, the system further determines that its heat source characteristic type is "upper edge heat accumulation". Finally, the system integrates the boundary information of each heat source and component that meets the spatial tolerance requirements to generate component identification results, including information such as associated heat source location, number of heat source points, maximum temperature, and boundary matching deviation, forming the final heat source-structure matching identification table.

[0092] Please see Figure 2 The multi-channel image fusion module includes:

[0093] The component recognition submodule calls the edge contour features and texture features in the component recognition results, obtains the input frame sequence of visible light image and infrared image, extracts the region coordinates of the same component in visible light and infrared image, compares the edge sharpness value and texture complexity value of the corresponding component in visible light and infrared image, and filters to obtain the coordinate set of component difference region.

[0094] The specific formulas for calculating the edge sharpness and texture complexity values ​​of corresponding parts in visible light and infrared images are as follows:

[0095] ;

[0096] Calculate the composite index values ​​of the features, and filter to obtain the coordinate set of component difference areas; among them... Representative features are compared with composite index values. Representing the visible light image Edge sharpness value per pixel, Representing the infrared image Edge sharpness value per pixel, Representing the visible light image Texture complexity value per pixel, Representing the infrared image Texture complexity value per pixel, Total number of pixels involved in the contrast of edge features, Total number of pixels involved in the computation of texture features;

[0097] This formula is used to evaluate the feature difference of components between visible and infrared images. The formula consists of two parts: one part calculates the total difference of edge sharpness, and the other part calculates the average difference of texture complexity.

[0098] Edge sharpness difference calculation: ;

[0099] Texture complexity difference calculation: ;

[0100] Parameter setting and data acquisition:

[0101] and Extracted from actual image data by image processing algorithms. Generally, edge sharpness can be calculated by methods such as Sobel operator or Canny edge detection.

[0102] and Texture complexity can be calculated by texture analysis methods such as gray level co-occurrence matrix (GLCM).

[0103] and Determined according to image resolution and selected area.

[0104] Numerical example and calculation process:

[0105] Assume a specific scene, pixels, pixels. Measured by image processing software, and The average values are 15 and 10 respectively, and The average values are 5 and 3 respectively.

[0106] Edge sharpness difference calculation:

[0107] ;

[0108] Texture complexity difference calculation:

[0109] ;

[0110] Result interpretation:

[0111] The calculated edge sharpness difference value is about 180.28, and the texture complexity difference value is 8. These two values combined indicate that there is a significant difference in the characteristics of the analyzed component between the visible light and infrared images. The composite feature comparison index value is 180.28 + 8 = 188.28, indicating that there is a significant difference in the characteristics of the images in the visible and infrared spectra. By analyzing this value, it can be determined which areas are inconsistent in the two image modes, effectively guiding subsequent image processing and analysis work.

[0112] Through actual monitoring and image processing algorithm extraction data, the practicality and accuracy of the parameters are ensured. This result provides a scientific basis for subsequent steps such as component difference area selection and analysis.

[0113] The difference allocation sub-module calculates the mean and standard deviation of the gray level of the corresponding area of the visible light image based on the component difference area coordinate set, and simultaneously calculates the mean and contrast value of the thermal radiation intensity of the corresponding area of the infrared image. The difference between the mean of the visible light gray level and the mean of the infrared thermal radiation intensity is calculated, and the fusion ratio is determined by combining the ratio of the standard deviation and the contrast value to generate a difference area fusion ratio set.

[0114] The corresponding area is extracted from the visible light image, and the pixel gray level of the area is statistically processed. First, the average level of the gray level is calculated to reflect the brightness reference of the area, and then the dispersion degree of the gray level distribution of the area is calculated as an indicator of the uniformity of the gray level. For example, in a 100x100 pixel area, the system will read and aggregate all pixel gray levels. If the average gray level is 123, it indicates that the overall brightness is moderate, and the standard deviation is 26, indicating that the brightness variation in the area is relatively obvious. Then, the thermal radiation intensity values of the corresponding area of the infrared image are read synchronously. Since the pixel values of the infrared image are essentially encoded as thermal intensity, the system directly inputs the thermal values for statistics, and calculates the average thermal value of the area, for example, 143, indicating that there is a certain heat source in the area. At the same time, the thermal intensity contrast is calculated based on the difference between the maximum and minimum values of the pixels, for example, the maximum thermal value is 180 and the minimum thermal value is 100, so the difference is 80, and the overall thermal contrast is moderate. By comparing the brightness mean of the visible light area and the thermal value mean of the infrared, the numerical difference is calculated, such as a difference of 20 units. At the same time, the fusion ratio is calculated by combining the ratio of the gray level distribution standard deviation and the infrared contrast, for example, the standard deviation of the area is 26 and the infrared contrast is 92 after conversion, so the fusion ratio is set to 0.16. This ratio means that the infrared image has a higher weight in this area. The system records the ratio for each difference area to generate a difference area fusion ratio set. The fusion ratios of all areas are stored in association with the coordinates to ensure that the weight settings of each area can be accurately extracted in subsequent operations.

[0115] The dynamic fusion sub-module calls the difference region coordinate set and the difference region fusion ratio set, and performs pixel-by-pixel superposition operation on the difference region of the visible light image and the infrared image, and fuses the non-difference region according to a preset ratio. The results of all regions are integrated and spliced into a continuous frame sequence to generate a fusion image data stream.

[0116] The dynamic fusion sub-module calls the difference region coordinate set and the difference region fusion ratio set, and performs pixel-by-pixel superposition operation on the difference region of the visible light image and the infrared image, and fuses the non-difference region according to a preset ratio. The results of all regions are integrated and spliced into a continuous frame sequence to generate a fusion image data stream.

[0117] Please refer to Figure 2 The device state monitoring module comprises:

[0118] The image fusion sub-module generates a multi-source image fusion data according to the fusion image data stream, the infrared image temperature sequence and the visible light image frame sequence synchronously collected in the device running process, aligns the temperature value of each pixel point of the infrared image and the space coordinates of the corresponding region of the visible light image on the time axis, and superimposes the temperature data to the visible light image coordinate system.

[0119] After receiving the fusion image data stream, the image fusion sub-module needs to perform time axis synchronization processing on the collected infrared image and visible light image. For example, if the infrared image has 30 frames per second and the visible light image has 25 frames per second, the infrared image can be time-aligned according to the visible light frame rate by key frame interpolation, that is, the corresponding infrared temperature frame is matched in the visible light frame with a similar time stamp. If the time points do not completely coincide, the infrared image can be time-interpolated and reconstructed to fill in the missing frames or frame interference, and synchronization acquisition is completed. In terms of spatial alignment, feature point matching is used to extract edges or corners with high stability in the image, such as using the Harris method to identify the device outline edge, and then converting and aligning the infrared image to the visible light image coordinate system through affine transformation. In actual device scenarios, such as infrared images taken from the front of electrical equipment, which need to be superimposed on 1080P high-definition images, the infrared image can be interpolated and enlarged and then projected and mapped according to the positioning mark. After mapping, for each pixel point in the visible light image, its corresponding area in the infrared image is found, the temperature data of the area is extracted and assigned to the coordinate position. For example, the temperature of the infrared image area corresponding to the center position of the device heat source concentration area is 52.1 degrees Celsius, that is, the same temperature is assigned to the coordinate. Then, the temperature value is converted into a displayable layer, for example, different colors are assigned to different temperature ranges in a pseudo-color way, and the original visible light image is overlaid pixel by pixel to complete image layer fusion and generate a fusion image sequence containing infrared temperature, visible light color and spatial coordinate information, which is convenient for subsequent processing and analysis.

[0120] The temperature trend analysis submodule calls the temperature sequence in the multi-source image fusion data, extracts the temperature range and the temperature change amplitude between adjacent frames in the device area within a continuous time window, compares the range with the temperature fluctuation range corresponding to the preset device type, filters the range with an out-of-limit range, calculates the statistical fluctuation range of the range with a change amplitude exceeding the reference change amplitude of the same type of device, and generates a temperature anomaly degree marked area.

[0121] After the temperature trend analysis submodule obtains the fused image sequence, it needs to continuously extract temperature data in several frames of images for a specific device area. A certain number of continuous frames are usually selected in a time window manner, for example, 10 frames of data are extracted with a 0.4 second window length, and a temperature change sequence is formed for each pixel point. In actual operation, the temperature sensitive parts inside the device are selected, such as the radiator area of the power module, and the maximum and minimum temperature difference of the temperature sequence in this period is recorded to calculate the temperature range, and then compared with the pre-set allowed fluctuation range of the device. For example, the temperature fluctuation of a certain type of device should not exceed 0.7°C when it is working normally. If the actual fluctuation exceeds this threshold, the area is marked as an extreme difference anomaly. Then, the temperature variation between adjacent frames is further extracted, and the average fluctuation amplitude is calculated and compared with the historical average temperature difference benchmark of the same type of device in operation. The historical benchmark value is usually obtained through long-term operation data of a large sample. For example, the normal change range of the fan module is between 0.1°C and 0.2°C. If the current observation data is significantly higher than this interval, it is marked as a fluctuation anomaly area. Finally, the distribution of these abnormal areas is calculated, the abnormal degree of the whole area is evaluated through the area occupied by the abnormal area and the abnormal pixel density, and the temperature abnormality degree result is output, which provides a basis for subsequent analysis.

[0122] The shape correlation determination submodule extracts the component contour deformation variable of the temperature abnormality degree marked area, calculates the Euclidean distance mean value of the contour point set between adjacent frames based on the visible light image frame sequence, matches the area with temperature abnormality and the area with deformation variable exceeding the assembly tolerance range in space coordinates, marks the boundary of the area with both deformation variable exceeding the standard and temperature abnormality, and maps the temperature abnormality area coordinates to the monitoring interface to generate the device state monitoring result.

[0123] The specific calculation formula for calculating the Euclidean distance mean value of the contour point set between adjacent frames is as follows:

[0124] ;

[0125] Among them, represents the Euclidean distance mean value of the contour point set between adjacent frames, represents the Euclidean distance of the contour point set between the i-th frame and the j-th frame, represents the average Euclidean distance between all contour point sets, and Q represents the total number of contour point sets calculated.

[0126] Euclidean distance of contour point set

[0127] The Euclidean distance of the contour point set refers to the straight line distance between two contour points in the image. In order to calculate , the contour point set in the image needs to be extracted first. For example, assuming that the contour point coordinates of the i-th frame in the image are , the contour point coordinates of the jth frame are According to the three-dimensional coordinates of the contour points, the Euclidean distance can be calculated by the following formula:

[0128] ;

[0129] Suppose the contour point coordinates between the ith frame and the jth frame are and , then:

[0130] ;

[0131] The Euclidean distance between the two frames of contour points is calculated.

[0132] The average Euclidean distance

[0133] of all contour point sets is calculated. The average Euclidean distance of all contour point sets is calculated by averaging all the calculated values. Suppose there are 5 pairs of contour points in a calculation, and the calculated Euclidean distance of each contour point set is:

[0134] ;

[0135] Then the average Euclidean distance is:

[0136] ;

[0137] Therefore, .

[0138] The average Euclidean distance is calculated as:

[0139] After calculating the Euclidean distance between each pair of contour points, in order to measure the fluctuation of the Euclidean distance of all contour point sets, the formula needs to take the absolute value of the difference between each pair of distance and the average , and then calculate the average of all absolute differences. In the calculation process, all need to be traversed, and the difference between each pair of contour points and the average Euclidean distance is calculated. Suppose N is 5, then:

[0140] ;

[0141] The calculation result is:

[0142] ;

[0143] Therefore, .

[0144] The weight parameter and the adjustment coefficient setting are based on:

[0145] In the calculation of the Euclidean distance mean, the selection and setting of each parameter are related to the acquisition method of the actual image data and the image quality. Through the monitoring of the actual image, the influence of light changes, camera position errors, etc. may need to be considered, and then different weight parameters or adjustment coefficients are set to optimize the calculation result. In this formula, there is no explicit weight parameter, but in actual application, different weights may be assigned to different distance values according to the reliability of different inter-frame contour points, especially in the case of possible noise influence during image acquisition.

[0146] Please refer to Figure 2 , the abnormal picture early warning module includes:

[0147] The abnormal area labeling submodule acquires the device running data set in the device state monitoring result, calls the preset reference comparison rule, calculates the difference value between the partition value in the data set and the corresponding reference item by item, filters the partitions whose difference value exceeds the set deviation threshold, maps the partition coordinates to the monitoring picture, and generates abnormal area coordinates;

[0148] First, the reference rule needs to be preset according to the device type, application scenario and historical stable running period data, for example, in a certain power inspection system, a standard model of running data is established for 10kV outdoor switch cabinet equipment, covering temperature, voltage, current, vibration and other key indicators. Each indicator has a corresponding reference value in the normal state, such as switch cabinet contact temperature reference of 75℃, equipment vibration amplitude of 0.3g, and current of 180A. After collecting the current period monitoring data, the real-time values of each indicator are compared in turn, such as current temperature 78℃, vibration 0.45g, and current 195A. Compare with the reference value in turn, and get the offset value of each item. Filter through the preset deviation judgment threshold, for example, the offset allowed threshold is set to 2℃, 0.1g and 10A respectively. Then all three data exceed the standard range, record as abnormal item, the system further binds the monitoring point and the partition coordinates, and maps the abnormal index corresponding partition to the image monitoring interface quickly. In the mapping process, the partition-coordinate conversion table is referenced, for example, the temperature anomaly is located in the main switch contact area (block #A3, corresponding coordinates 12,8). The system aggregates these abnormal items to form an abnormal coordinate set, which is used as the basis for subsequent picture highlighting and alarm.

[0149] The display adjustment and comparison submodule calls the abnormal area coordinates, extracts the real-time monitoring data stream in the abnormal area, adjusts the color saturation and flicker frequency in the picture display setting, synchronously obtains the normal data mean value in the same area running period, compares the real-time data stream with the data mean value in the segmented time sequence, and generates an abnormal data comparison set;

[0150] After receiving the aforementioned abnormal area coordinates, the display adjustment and comparison submodule immediately calls the real-time monitoring data stream in the area for processing. For example, the system receives temperature abnormal coordinates (12, 8), and calls out the temperature data of the area in the last 10 seconds from the database or cache: [77.5, 78.0, 78.2, 78.4, 78.7]℃. At the same time, the visualization display parameters of the area are changed, such as increasing the original layer saturation to 180% and setting the flicker frequency of the area to once every 2 seconds to enhance its recognition. On this basis, the system synchronously calls the normal mean value data in the historical stable running period corresponding to the coordinate point, such as the historical temperature mean value of the area being 75.3℃. The system processes the current temperature stream in segments (such as every 10 seconds as a segment), calculates the average temperature in each segment, such as the current segment mean value being 78.16℃, calculates the deviation from the normal value, and through the processing of multiple time segments, the system forms an abnormal comparison data set containing time markers and deviation values, such as marker T1 segment deviation 2.86℃, T2 segment deviation 3.12℃, etc., providing basic data support for subsequent fluctuation trend and alarm level.

[0151] The abnormal evaluation trigger submodule extracts the maximum deviation amplitude and the mean value of the fluctuation frequency based on the abnormal data comparison set, combines the load degree and the running time proportion in the device running data set, calculates the abnormal deviation value, matches the deviation value with the preset warning level threshold interval, triggers the corresponding level of warning instruction, and generates abnormal picture warning information;

[0152] First, the maximum deviation amplitude in each time period is extracted, for example, the maximum temperature deviation value in multiple time periods is 3.20℃, and the system records this as the extreme value of this period deviation, and analyzes the overall fluctuation frequency, for example, if the sampling frequency is once per second, and the frequency of temperature change exceeding ±0.2℃ in 30 seconds of continuous data is detected to be 10 times, then the fluctuation frequency can be defined as 0.33Hz, then the system calls the running data of the equipment where the abnormal point is located, extracts the load degree of the current period, for example, if the device current is 195A and the rated current is 180A, then the load degree is calculated to be 1.08 times, which is classified as a "high load" area (defined as high when >1.00 times), combined with the running time of the device, for example, the device has been running for 16 hours, and the total plan is 24 hours, then the running ratio is 0.667, which is classified as "high" (according to the set interval, <0.4 is low, 0.4-0.7 is medium, >0.7 is high), the system combines the deviation extreme value, load intensity and running ratio for weighted combination processing to generate a comprehensive deviation value, for example, with a weight ratio of 4:3:3, the calculated deviation comprehensive value is 2.30℃, then the system matches the set warning level interval to determine, for example, the first level warning is 2.5℃ and above, the second level is 1.5℃ to 2.5℃, and the third level is 1.5℃ or below, the current result falls into the second level warning interval, the system generates a "Level-2" warning instruction, and based on the abnormal coordinate information (such as coordinates 12,8), it is superimposed and displayed on the monitoring main interface to form a corresponding warning identifier and detailed information for subsequent disposal personnel to identify and confirm.

[0153] The above is only a preferred embodiment of the present application, and does not limit the form of the present application in other forms. Any skilled person in the art can use the disclosed technical content to make changes or modifications to equivalent embodiments applied to other fields, but any simple modification, equivalent change and modification made according to the technical essence of the present application to the above embodiments without departing from the technical solution content of the present application, still belongs to the protection scope of the technical solution of the present application.

Claims

1. A switchgear multi-mode multi-channel online monitoring device, characterized in that: The method comprises the following steps: The video data acquisition module acquires the switch cabinet video, calls the ambient light sensor to measure the light intensity, calculates the brightness uniformity of the visible light picture, analyzes the thermal signal coverage range of the infrared video, evaluates the imaging integrity of the two types of videos, and forms a video signal quality evaluation result; The switch cabinet component identification module calls the video signal quality evaluation result, detects the edge features in the visible light video, extracts the contour information of the internal components of the switch cabinet, analyzes the temperature distribution pattern in the infrared video, identifies the heat source features of the key components, and combines the analysis of the internal structure of the equipment based on the two types of videos to form a component identification result; The multi-channel image fusion module calls the component identification result, dynamically fuses the visible light and infrared videos, adjusts the fusion proportion of the difference areas based on the visibility of the components, and generates a fused image data stream; The equipment state monitoring module acquires the fused image data stream, analyzes the temperature change trend of the infrared signal during the equipment operation, combines the component shape change in the visible light image, detects the state abnormality, and displays the suspicious area in real time on the monitoring interface, and outputs an equipment state monitoring result; The multi-channel image fusion module comprises: The component identification submodule calls the edge contour features and texture features in the component identification result, acquires the input frame sequence of the visible light image and the infrared image, extracts the region coordinates of the same component in the visible light and infrared images, compares the edge sharpness value and the texture complexity value of the corresponding components in the visible light and infrared images, and selects the component difference region coordinate set; The difference assignment submodule calculates the gray mean value and the standard deviation of the corresponding region of the visible light image based on the component difference region coordinate set, synchronously calculates the thermal radiation intensity mean value and the contrast value of the corresponding region of the infrared image, performs difference value operation on the visible light gray mean value and the infrared thermal radiation intensity mean value, determines the fusion proportion based on the ratio of the standard deviation and the contrast value, and generates a difference region fusion proportion set; The dynamic fusion submodule calls the component difference region coordinate set and the difference region fusion proportion set, performs pixel-by-pixel superposition operation on the difference regions of the visible light image and the infrared image, performs fusion on the non-difference regions according to the preset proportion, integrates all region results and splices them into a continuous frame sequence, and generates a fused image data stream; The corresponding region is extracted from the visible light image, the pixel gray value of the region is statistically processed, the average level of the gray value is calculated, the brightness reference of the region is reflected, the dispersion degree of the region gray value distribution is counted, and the index of the region gray uniformity is obtained; The difference region coordinate set and the fusion proportion set are called, and pixel-by-pixel fusion operation is performed on the difference regions of the input visible light image and infrared image, the coordinate range of the identified difference region in the image is extracted, the visible light pixel value and the infrared pixel value in the region are read according to the pixel, and the fusion value of each pixel is calculated according to the fusion proportion corresponding to the region in the proportion set; The video data acquisition module comprises: The video signal acquisition submodule acquires the visible light video and infrared video data of the switch cabinet, calls the ambient light sensor to measure the ambient light intensity value, generates visible light video data, infrared video data, and ambient light intensity value; The light analysis submodule calculates the mean and variance of the pixel brightness of a picture based on the visible light video data, the infrared video data and the ambient light intensity value, analyzes the intensity distribution range and continuity of the thermal signal in the infrared video data, and generates a brightness balance value and a thermal signal coverage state; The integrity evaluation submodule calls the brightness balance value and the thermal signal coverage state, compares a preset brightness deviation standard and a thermal signal coverage standard, evaluates the imaging integrity of the visible light and infrared video, and generates a video signal quality evaluation result; The video acquisition terminal is started and positioned to the target area of the switch cabinet, the acquisition angle, the acquisition frame rate and the focal length parameters are set through the control program, and the infrared video acquisition channel is operated in parallel, so that the thermal imaging of the same area is recorded, and the ambient light sensor module is started synchronously; The switch cabinet component recognition module comprises: The edge contour extraction submodule calls the visible light video frame brightness data and the signal-to-noise ratio data in the video signal quality evaluation result, detects the edge gradient change amplitude of the video frame, filters the edge points according to the gradient direction and the preset standard, connects the adjacent points to form a closed boundary, extracts the geometric shape information of the switch cabinet component and generates a component contour feature set; The heat source correlation analysis submodule calls the geometric boundary range of the component contour feature set, limits the detection range of the temperature difference abnormal area, calculates the temperature difference value of the target area and the periphery by analyzing the temperature distribution data of the infrared video frame, filters the abnormal temperature rise area according to the preset threshold, records the highest temperature point coordinates and the change trend, and generates a heat source distribution map in combination with the component contour; The structure analysis submodule calls the temperature distribution data of the heat source distribution map and the geometric boundary information of the component contour feature set, aligns the component boundary center point and the heat source highest temperature point in space, calculates whether the position deviation of the two is less than a preset spatial tolerance range, filters the component heat source correlation group meeting the condition, and generates a component recognition result according to the contour form and heat source distribution law in the correlation group.

2. The switchgear multi-mode multi-channel online monitoring device according to claim 1, characterized in that: The video signal quality evaluation result includes brightness balance, thermal signal coverage range and imaging integrity, the component recognition result includes component contour information, temperature distribution mode, heat source feature and internal structure of the equipment, the fusion image data stream includes image fusion proportion and imaging effect, and the equipment state monitoring result includes temperature change trend, component form change, state abnormality and suspicious area.

3. The switchgear multi-mode multi-channel online monitoring device according to claim 1, characterized in that: The specific calculation formula of the analysis of the thermal signal intensity distribution range and continuity in the infrared video data is as follows: ； The thermal signal distribution continuity characteristic value is calculated, the brightness balance value and the thermal signal coverage state are generated; wherein, represents the first frame infrared image, represents the number of pixel points in the first frame identified as a thermal signal region, represents the number of pixel points in the first thermal signal region in the first frame, represents the average value of thermal intensity of all thermal signal region pixels in the first frame, represents the average value of thermal intensity of all thermal signal region pixels in the first frame, represents the standard deviation of all pixel brightness in the first frame, represents the average value of all pixel brightness in the first frame, is a very small constant to prevent the denominator from being zero.

4. The switchgear multi-mode multi-channel online monitoring device according to claim 1, characterized in that: The specific calculation formula of the comparison of the edge sharpness value and the texture complexity value of the corresponding components of the visible light and infrared images is as follows: ； Calculate the composite index values ​​of the features, and filter to obtain the coordinate set of component difference areas; among them... Representative features are compared with composite index values. Representing the visible light image Edge sharpness value per pixel, Representing the infrared image Edge sharpness value per pixel, Representing the visible light image Texture complexity value per pixel, Representing the infrared image Texture complexity value per pixel, The total number of pixels representing edge features involved in the comparison. This represents the total number of pixels involved in the calculation of texture features.

5. The switchgear multi-mode multi-channel online monitoring device according to claim 1, characterized in that: The equipment state monitoring module comprises: The image fusion submodule monitors the infrared image temperature sequence and the visible light image frame sequence synchronously collected in the equipment running process according to the fusion image data stream, aligns the temperature value of each pixel point of the infrared image and the spatial coordinates of the corresponding area of the visible light image on the time axis, superimposes the temperature data on the visible light image coordinate system, and generates multi-source image fusion data. The temperature trend analysis submodule calls the temperature sequence in the multi-source image fusion data, extracts the temperature range of the equipment area in a continuous time window, compares the range with a preset temperature fluctuation range corresponding to the equipment type, screens the range exceeding the limit, calculates the statistical fluctuation range of the range exceeding the reference change amplitude of the same type of equipment, and generates a temperature anomaly degree marked area; The shape correlation determination submodule extracts the component contour deformation variable of the temperature anomaly degree marked area, calculates the Euclidean distance mean of the contour point set between adjacent frames based on the visible light image frame sequence, matches the area with the deformation variable exceeding the assembly tolerance range with the temperature abnormal area in space coordinates, marks the boundary of the area with both deformation variable exceeding the standard and temperature anomaly, maps the temperature abnormal area coordinates to the monitoring interface, and generates the equipment state monitoring result.

6. The switchgear multi-mode multi-channel online monitoring device according to claim 5, characterized in that: The specific calculation formula of the Euclidean distance mean of the contour point set between adjacent frames is: ； wherein, a mean of Euclidean distances between adjacent contour point sets, a Euclidean distance between contour point sets of the i-th frame and the j-th frame, a mean of Euclidean distances between all contour point sets, Q represents a total number of contour point sets.

7. The switchgear multi-mode multi-channel online monitoring device according to claim 1, characterized in that: Further comprising: The abnormal picture early warning module calls the equipment state monitoring result, automatically labels the abnormal area on the monitoring picture and adjusts the display mode, compares and analyzes the data of the abnormal area, evaluates the severity of the problem, and triggers the abnormal picture early warning information; The abnormal picture early warning information includes abnormal area labeling, display mode adjustment, data comparison and analysis, and problem severity evaluation; The abnormal picture early warning module comprises: The abnormal area labeling submodule obtains the equipment operation data set in the equipment state monitoring result, calls a preset reference comparison rule, calculates the difference between the partition values in the data set and the corresponding reference, screens the partition with a difference exceeding a set deviation threshold, maps the partition coordinates to the monitoring picture, and generates abnormal area coordinates; The display adjustment and comparison submodule calls the abnormal area coordinates, extracts the real-time monitoring data stream in the abnormal area, adjusts the color saturation and flicker frequency in the picture display setting, synchronously obtains the normal data mean in the same area operation period, compares the real-time data stream with the data mean in a segmented time sequence, and generates an abnormal data comparison set; The abnormal evaluation triggering submodule extracts the maximum deviation amplitude and the mean fluctuation frequency based on the abnormal data comparison set, combines the load degree and the operation time proportion in the equipment operation data set, calculates the abnormal deviation value, matches the deviation value with a preset early warning level threshold interval, triggers the early warning instruction of the corresponding level, and generates the abnormal picture early warning information.

Images