Edge intelligence driven infrared image real-time analysis and abnormal early warning method and system

Abstract

The present application relates to the technical field of anomaly detection, in particular to an edge intelligent driven infrared image real-time analysis and anomaly early warning method and system, comprising the following steps: obtaining power distribution cabinet monitoring view field thermal image frame and generating temperature matrix, screening threshold exceeding pixels and determining connected domain, calculating area and centroid to generate feature set, determining centroid offset consistency and counting continuous frame and alarm interval, generating rollback direction state and calculating threshold adjustment amount, updating threshold according to the same and writing into alarm register and synchronizing sound and light threshold, in the present application, through the correspondence between pixel level temperature matrix and physical coordinates, accurate quantitative positioning of abnormal heat area is realized, the connected domain area, centroid and maximum temperature features are introduced to enhance the discrimination dimension, the continuous frame centroid consistency and continuous time constraint are combined to improve the stable heat source identification ability, and the alarm threshold is controlled and adjusted according to the heat area evolution state, forming an edge side adaptive closed loop early warning, improving the alarm accuracy and operation adaptability.

Description

Technical Field

[0001] This invention relates to the field of anomaly detection technology, and in particular to a method and system for real-time analysis and anomaly early warning of infrared images driven by edge intelligence. Background Technology

[0002] Anomaly detection technology refers to a field that continuously collects and analyzes equipment operating status or environmental information to identify abnormal states and trigger early warnings. Its core aspects include acquiring abnormal information, extracting abnormal features, setting abnormal judgment rules, and organizing early warning triggering methods. Generally, it employs a process of establishing benchmarks and comparative analysis of collected data. First, reference features or threshold ranges are established under normal operating conditions or target scenarios. Then, real-time data is compared frame-by-frame or time-by-time. Quantifiable indicators such as temperature difference, brightness distribution changes, regional morphological deviations, or continuous frame changes are used to complete anomaly identification and output early warning signals. Traditional infrared image real-time analysis and anomaly early warning refer to the use of infrared thermal imaging... This patent addresses a method for detecting the temperature field and its changes in an image as an object, and determining whether there are overheating, cooling failure, hot spot diffusion, or abnormal temperature regions. The technical issue addressed in this patent is to process infrared images in real time and provide anomaly alarms under edge-side deployment conditions. Traditional methods typically involve performing grayscale threshold segmentation on infrared images to extract high-temperature or low-temperature abnormal regions, and then combining connected component analysis to obtain the area, location, and morphological parameters of the abnormal regions. At the same time, differential calculations are performed on consecutive frames to determine hot spot growth or temperature abrupt changes. Anomaly determination is then made based on a pre-set temperature threshold range, region area threshold, and rate of change threshold. The determination result is then used as the basis for triggering early warning and outputting alarm information.

[0003] Current anomaly detection methods largely rely on frame-by-frame comparison using fixed temperature thresholds or empirically set ranges. These thresholds remain constant over long periods, and as equipment operating conditions, ambient temperature, or load levels change, the thresholds gradually deviate from the actual state. This can easily trigger invalid alarms frequently under high load conditions or miss early risk signals during slow temperature increases. Anomaly region determination typically focuses on single-frame grayscale or temperature segmentation results, paying insufficient attention to the positional stability of anomaly regions over continuous time. This leads to false hotspots caused by reflections, occlusions, or transient thermal disturbances being mistaken for real anomalies. Continuous frame differential analysis often focuses on the overall temperature change amplitude, lacking constraints on the spatial migration of local heat sources, making it difficult to distinguish hotspot diffusion from random noise. Furthermore, there is a lack of feedback between alarm triggering and threshold adjustment; even when alarms occur frequently, the original threshold is still used, failing to adjust the judgment strategy based on historical alarm behavior. In edge deployment scenarios, this static rule-based and single-criteria operating mode easily leads to alarm jitter, misjudgments by maintenance personnel, and delays in handling, ultimately affecting the operational safety and maintenance efficiency of power distribution equipment. Summary of the Invention

[0004] To address the technical problems existing in the prior art, embodiments of the present invention provide a method for real-time analysis and anomaly warning of infrared images driven by edge intelligence, comprising the following steps:

[0005] To achieve the above objectives, the present invention adopts the following technical solution: a method for real-time analysis and anomaly warning of infrared images driven by edge intelligence, comprising the following steps:

[0006] S1: Acquire thermal image frames of the monitoring field of view of the circuit breaker contacts and busbar cable joints of the distribution cabinet by the infrared thermal imaging camera, extract the pixel temperature values ​​and write them into the temperature matrix to generate a thermal image temperature matrix.

[0007] S2: Based on the thermal image temperature matrix, filter pixels that exceed the temperature threshold and determine the connected components, calculate and record the area and centroid coordinates of the connected components, and generate an abnormal hot area connected component feature set.

[0008] S3: Based on the abnormal hot zone connected component feature set, calculate the consistency of centroid coordinate offset of consecutive frames, count the number of continuous frames of hot zone and generate alarm timestamp sequence, compare alarm interval with interval threshold and continuous frame threshold, and generate backtracking direction determination status.

[0009] S4: Determine the state based on the back-off direction, compare the temperature threshold back-off step size with the upper limit of the maximum temperature change, and calculate and generate the threshold adjustment amount; S5: Based on the threshold adjustment amount, combined with the backtracking direction determination status update temperature threshold, the update result is written into the edge computing terminal alarm determination register and synchronously configured to the audible and visual alarm trigger threshold to form an abnormal warning trigger threshold configuration.

[0010] As a further aspect of the present invention, the thermal imaging temperature matrix includes pixel temperature value distribution, row and column coordinate index, and calibration coordinate mapping relationship; the abnormal hot zone connected domain feature set includes connected domain area, connected domain centroid coordinates, and connected domain maximum temperature value; the rollback direction determination state specifically refers to alarm interval comparison conclusion, continuous frame count comparison conclusion, and threshold rollback direction type; the threshold adjustment amount specifically includes threshold update step size, single change limit value, and threshold adjustment amount selection value; the abnormal warning trigger threshold configuration includes warning temperature threshold setting value, alarm determination register write value, and audible and visual alarm trigger threshold value.

[0011] As a further aspect of the present invention, the specific steps of S1 are as follows:

[0012] S101: Acquire thermal image frame data output by infrared thermal imaging camera monitoring the field of view of circuit breaker contacts and busbar cable joints in distribution cabinet, collect temperature values ​​corresponding to pixels in the frame, arrange pixel temperatures according to the original row and column coordinates of the thermal image frame, and obtain pixel temperature array values.

[0013] S102: Based on the pixel temperature array value, call the row and column coordinate information of the current frame of the infrared thermal imaging camera, the corresponding pixel temperature value and its row and column coordinates, perform coordinate alignment consistency check and position correction, and generate a pixel coordinate temperature correspondence table.

[0014] S103: Based on the pixel coordinate temperature correspondence table, call the row and column coordinate mapping relationship in the camera calibration table, perform calibration correction calculation and complete the row and column coordinate mapping replacement, and rewrite the corrected coordinates and corresponding temperature values ​​into the matrix structure to generate a thermal image temperature matrix.

[0015] As a further aspect of the present invention, the specific steps of S2 are as follows:

[0016] S201: Based on the thermal imaging temperature matrix, filter and compare the temperature values ​​in the matrix with the set temperature threshold, retain the row and column coordinates of pixels that exceed the temperature threshold, and generate a set of pixel coordinates exceeding the threshold.

[0017] S202: Based on the set of pixel coordinates exceeding the threshold, determine the connectivity of pixel row and column coordinates in the up, down, left, and right directions, perform adjacent pixel coordinate merging operation, divide continuous coordinates into groups and count the number of coordinates in each group, and obtain the connected region area set.

[0018] S203: Based on the connected region area set, call the temperature value of the corresponding pixel coordinate position, calculate the average value of the row coordinate and column coordinate of the connected region as the centroid coordinate, and compare the temperature values ​​in the connected region to extract the maximum temperature value, thereby generating an abnormal hot zone connected region feature set.

[0019] As a further aspect of the present invention, the specific steps of S3 are as follows:

[0020] S301: Based on the abnormal hot zone connected domain feature set, obtain the row and column coordinate values ​​of the centroid of the connected domain in consecutive frames, match the connected domains of adjacent frames based on the principle of minimum centroid distance to determine the corresponding connected domain, calculate the centroid coordinate difference of the corresponding connected domain, and generate a centroid offset consistency judgment value.

[0021] S302: Based on the centroid offset consistency judgment value, collect the frame sequence number corresponding to the connected component that meets the consistency condition, count the number of consecutively occurring frames and record the corresponding time stamp, and generate an alarm timestamp sequence.

[0022] S303: Based on the alarm timestamp sequence, calculate and compare the adjacent timestamp interval value with the alarm interval threshold, and at the same time calculate and compare the number of consecutively occurring frames with the continuous frame threshold to generate a rollback direction determination state.

[0023] As a further aspect of the present invention, the specific steps of S4 are as follows:

[0024] S401: Based on the back-off direction determination state, read the temperature threshold back-off step size value and the maximum temperature change limit value in the current configuration, compare the size relationship of the two types of values, limit the back-off step size value according to the maximum temperature change limit, and generate a threshold back-off constraint value.

[0025] S402: Based on the threshold backoff constraint value, calculate the difference between the temperature threshold backoff step size and the maximum temperature change limit value to obtain the threshold adjustment calculation amount;

[0026] S403: Based on the threshold adjustment calculation, determine the direction identifier corresponding to the state according to the rollback direction, assign a positive or negative sign to the threshold adjustment calculation and complete the sign correction, and generate the threshold adjustment value.

[0027] As a further aspect of the present invention, the specific steps of S5 are as follows:

[0028] S501: Based on the threshold adjustment value, read the current temperature threshold of the edge computing terminal, and combine it with the direction identifier in the back-off direction determination state to determine the temperature threshold adjustment direction and obtain the threshold update direction identifier value;

[0029] S502: Based on the threshold update direction identifier value, calculate the addition and subtraction result of the current temperature threshold value and the threshold adjustment value, write the calculated temperature threshold into the edge computing terminal alarm judgment register, and generate the updated temperature threshold value.

[0030] S503: Based on the updated temperature threshold value, write the value into the corresponding trigger threshold storage unit of the audible and visual alarm, and read the current trigger threshold parameter to update the value, thereby generating an abnormal warning trigger threshold configuration.

[0031] As a further aspect of the present invention, the infrared thermal imaging camera is an imaging device used to collect infrared radiation from the surface of the monitored equipment and output thermal image frame data of temperature distribution, and its output serves as the source of thermal image frame data.

[0032] The circuit breaker contacts of the distribution cabinet are conductive contact parts located inside the distribution cabinet for circuit switching, and their surface temperature is one of the objects monitored by infrared.

[0033] The busbar cable connector is a conductive connection component used to realize the electrical connection between the busbar and the cable, and its surface temperature is one of the objects monitored by infrared.

[0034] The thermal image frame data is a frame of temperature distribution image data output by the infrared thermal imaging camera within a unit sampling period, which includes pixel coordinates and corresponding temperature values.

[0035] The thermal image temperature matrix is ​​a two-dimensional temperature data set formed by arranging the temperature values ​​of each pixel in the thermal image frame data according to the row and column coordinate order.

[0036] The temperature threshold is a temperature judgment benchmark used to distinguish between normal pixels and abnormal pixels, and its value comes from the local configuration parameters of the edge computing terminal.

[0037] The connectivity determination is a process of judging the spatial adjacency relationship between adjacent pixel coordinates, used to verify whether abnormal pixels belong to the same abnormal region.

[0038] The area of ​​the connected region is the number of pixels covered by the set of abnormal pixels formed by connectivity determination;

[0039] The centroid coordinates are the center position coordinates calculated based on the coordinates of all pixels in the connected domain, and are used to characterize the spatial location of the abnormal hot zone.

[0040] The abnormal hot zone connected domain feature set is a data set consisting of the connected domain area, centroid coordinates, and the maximum temperature value within the connected domain.

[0041] As a further aspect of the present invention, the continuous frame count of the hot zone refers to the frame count value of the connected domain that is determined to be the same abnormal hot zone repeatedly in consecutive thermal image frames.

[0042] The alarm timestamp sequence is a time-sequential data set formed by the edge computing terminal recording the trigger time of each abnormal warning;

[0043] The alarm interval value is the time difference calculated from two adjacent alarm timestamps in the alarm timestamp sequence;

[0044] The alarm interval threshold is a time-based benchmark for determining whether alarm triggering is in a dense state, and its value is obtained by the edge computing terminal configuration.

[0045] The continuous frame threshold is a frame count criterion used to determine whether an abnormal hot zone persists, and its value comes from the edge computing terminal configuration parameters.

[0046] The rollback direction determination state is a determination result generated based on the comparison between the alarm interval value and the alarm interval threshold, and between the hot zone continuous frame count and the continuous frame threshold.

[0047] The temperature threshold backoff step size is the amount of temperature change used to adjust the warning temperature threshold, and its value comes from the edge computing terminal configuration parameters.

[0048] The maximum temperature change limit is a temperature constraint value used to limit the adjustment range of a single temperature threshold, and its value is obtained by the edge computing terminal configuration.

[0049] The threshold adjustment value is the temperature change obtained after filtering based on the temperature threshold backoff step size and the upper limit of the maximum temperature change.

[0050] The abnormal warning trigger threshold configuration is a set of warning threshold parameters formed by applying the threshold adjustment value to the current temperature threshold, which is used for abnormal warning determination.

[0051] An edge-intelligent-driven real-time infrared image analysis and anomaly early warning system includes:

[0052] The infrared acquisition module is used to perform S1: acquire the thermal image frame data output by the infrared thermal imaging camera monitoring the circuit breaker contacts and busbar cable joints of the distribution cabinet, acquire the temperature value of the pixel in the thermal image frame and write it into the temperature matrix according to the row and column coordinates, and generate the thermal image temperature matrix by corresponding to the coordinates of the temperature matrix and the camera calibration table.

[0053] The hot zone extraction module is used to perform S2: based on the thermal image temperature matrix, filter the pixel coordinates whose temperature values ​​exceed the temperature threshold and perform connectivity determination, calculate the area of ​​the connected domain and the centroid coordinates and extract the maximum temperature value of the connected domain, and generate an abnormal hot zone connected domain feature set.

[0054] The timing determination module is used to execute S3: calculate and determine whether the centroid coordinate offset of the connected domain in consecutive frames is consistent based on the abnormal hot zone connected domain feature set, count the number of continuous frames of the hot zone and generate an alarm timestamp sequence, compare the alarm interval value with the alarm interval threshold, and at the same time compare the number of continuous frames of the hot zone with the continuous frame threshold to generate a rollback direction determination state.

[0055] The threshold backoff module is used to execute S4: based on the backoff direction determination state, read and compare the temperature threshold backoff step size with the upper limit of the maximum temperature change, calculate the threshold adjustment amount, and generate the threshold adjustment amount value;

[0056] The early warning configuration module is used to execute S5: adjust the value according to the threshold, read the current temperature threshold and select the threshold update direction according to the backtracking direction, calculate the updated temperature threshold and write it into the edge computing terminal alarm judgment register, synchronize the temperature threshold to the trigger threshold of the audible and visual alarm, and generate an abnormal early warning trigger threshold configuration.

[0057] Compared with the prior art, the advantages and positive effects of the present invention are as follows:

[0058] In this invention, the precise quantitative location of abnormal hot zones is achieved by corresponding pixel-level temperature matrices with physical coordinates. The features of connected domain area, centroid, and maximum temperature are introduced to enhance the discrimination dimension. The ability to identify stable heat sources is improved by combining the centroid consistency of continuous frames and duration constraints. The alarm threshold is adjusted in a controlled manner according to the evolution state of the hot zone to form an edge-side adaptive closed-loop early warning, thereby improving the accuracy of alarms and operational adaptability. Attached Figure Description

[0059] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0060] Figure 1 This is a schematic diagram of the steps of the present invention;

[0061] Figure 2 This is a detailed schematic diagram of S1 of the present invention;

[0062] Figure 3 This is a detailed schematic diagram of S2 of the present invention;

[0063] Figure 4 This is a detailed schematic diagram of S3 of the present invention;

[0064] Figure 5 This is a detailed schematic diagram of S4 of the present invention;

[0065] Figure 6 This is a detailed schematic diagram of S5 of the present invention. Detailed Implementation

[0066] The technical solution of the present invention will now be described with reference to the accompanying drawings.

[0067] In this embodiment of the invention, sometimes a subscript such as W1 may be written in a non-subscript form such as W1. When the difference is not emphasized, the meaning they express is the same.

[0068] To make the technical problems, technical solutions and advantages of the present invention clearer, a detailed description will be given below in conjunction with the accompanying drawings and specific embodiments.

[0069] Please see Figure 1 This invention provides a method for real-time analysis and anomaly warning of infrared images driven by edge intelligence, comprising the following steps:

[0070] S1: Acquire thermal image frame data from the infrared thermal imaging camera monitoring the circuit breaker contacts and busbar cable joints of the distribution cabinet, collect the temperature values ​​of the pixels in the thermal image frame and write them into the temperature matrix according to the row and column coordinates, and generate the thermal image temperature matrix by matching the coordinates of the temperature matrix with the camera calibration table.

[0071] An infrared thermal imaging camera is an imaging device used to collect infrared radiation from the surface of a monitored device and output thermal image frame data of temperature distribution. Its output serves as the source of thermal image frame data.

[0072] The circuit breaker contacts in the distribution cabinet are conductive contact parts located inside the distribution cabinet for switching circuits on and off, and their surface temperature is one of the objects monitored by infrared sensors.

[0073] Busbar cable joints are conductive connection components used to achieve electrical connection between busbars and cables, and their surface temperature is one of the objects monitored by infrared sensors.

[0074] Thermal image frame data is a frame of temperature distribution image data output by an infrared thermal imaging camera within a unit sampling period, which includes pixel coordinates and corresponding temperature values;

[0075] A thermal image temperature matrix is ​​a two-dimensional temperature data set formed by arranging the temperature values ​​of each pixel in a thermal image frame according to their row and column coordinates.

[0076] S2: Based on the thermal image temperature matrix, filter the pixel coordinates with temperature values ​​exceeding the temperature threshold and determine connectivity. Calculate the area of ​​the connected domain and the centroid coordinates, extract the maximum temperature value of the connected domain, and generate a connected domain feature set of abnormal hot areas.

[0077] The temperature threshold is a temperature criterion used to distinguish between normal and abnormal pixels, and its value comes from the local configuration parameters of the edge computing terminal.

[0078] Connectivity determination is a process of judging the spatial adjacency relationship between adjacent pixel coordinates, used to verify whether abnormal pixels belong to the same abnormal region;

[0079] The area of ​​a connected component is the number of pixels covered by the set of anomalous pixels formed through connectivity determination.

[0080] Centroid coordinates are the center position coordinates calculated based on the coordinates of all pixels in the connected domain, and are used to characterize the spatial location of abnormal hot areas;

[0081] The feature set of the connected domain of the anomalous hot zone is a data set consisting of the area of ​​the connected domain, the coordinates of the centroid, and the maximum temperature value within the connected domain.

[0082] S3: Based on the abnormal hot zone connected component feature set, calculate and determine whether the centroid coordinate offset of the connected components in consecutive frames is consistent, count the number of continuous frames of the hot zone and generate an alarm timestamp sequence, compare the alarm interval value with the alarm interval threshold, and at the same time compare the number of continuous frames of the hot zone with the continuous frame threshold to generate a backtracking direction determination status.

[0083] The hot zone persistence frame count refers to the frame count value of connected components that are identified as the same abnormal hot zone repeatedly in consecutive thermal image frames;

[0084] The alarm timestamp sequence is a time-sequential data set formed by the edge computing terminal recording the trigger time of each abnormal warning;

[0085] The alarm interval value is the time difference calculated from two adjacent alarm timestamps in the alarm timestamp sequence;

[0086] The alarm interval threshold is a time-based benchmark used to determine whether alarm triggering is in a dense state, and its value is obtained by the edge computing terminal configuration.

[0087] The persistent frame threshold is a frame count criterion used to determine whether an abnormal hot zone persists. Its value comes from the edge computing terminal configuration parameters.

[0088] The rollback direction determination status is generated based on the comparison between the alarm interval value and the alarm interval threshold, as well as the hot zone continuous frame count and the continuous frame threshold.

[0089] S4: Based on the back-off direction determination status, read and compare the temperature threshold back-off step size with the upper limit of the maximum temperature change, calculate the threshold adjustment amount, and generate the threshold adjustment amount value;

[0090] The temperature threshold backoff step size is the amount of temperature change used to adjust the warning temperature threshold, and its value comes from the edge computing terminal configuration parameters.

[0091] The maximum temperature change limit is a temperature constraint value used to limit the magnitude of a single temperature threshold adjustment, and its value is obtained by the edge computing terminal configuration.

[0092] The threshold adjustment value is the temperature change obtained after filtering based on the temperature threshold backoff step size and the upper limit of the maximum temperature change.

[0093] S5: Adjust the value according to the threshold, read the current temperature threshold and select the threshold update direction according to the backtracking direction, calculate the updated temperature threshold and write it to the edge computing terminal alarm judgment register, synchronize the temperature threshold to the audible and visual alarm trigger threshold, and generate the abnormal warning trigger threshold configuration.

[0094] The abnormal warning trigger threshold configuration is a set of warning threshold parameters formed by applying the threshold adjustment value to the current temperature threshold, which is used for abnormal warning determination.

[0095] Please see Figure 2 The specific steps of S1 are as follows:

[0096] S101: Acquire thermal image frame data output by infrared thermal imaging camera monitoring the field of view of circuit breaker contacts and busbar cable joints in distribution cabinet, collect temperature values ​​corresponding to pixels in the frame, arrange pixel temperatures according to the original row and column coordinates of the thermal image frame, and obtain pixel temperature array values.

[0097] The monitoring focuses on a single frame image formed by an infrared thermal imaging device observing the connection area between the circuit breaker contacts and busbar cables within a distribution cabinet. In practice, the infrared detection array first receives infrared radiation intensity signals from various surface locations within a single sampling cycle. These signals are output sequentially according to the physical arrangement of the detection units, with each unit corresponding to a specific spatial sampling point. The radiation energy at this point is converted into a voltage signal and recorded immediately. Then, based on the voltage-temperature correlation established at the factory, the voltage value is converted into a temperature value. For example, in a certain frame, the detection unit in row 120, column 85 outputs a voltage of 2.35 V, which is converted to a temperature of 53.75 °C; the detection unit in row 118, column 90 outputs a voltage of 2.65 V, which, using the same conversion, yields a temperature of 61.25 °C. The temperature conversion process for all pixels uses the same parameter source and calculation path. Then, the system arranges the obtained temperature data sequentially according to the row and column parameters of the current frame. Assuming the current thermal image resolution is 320×240, the system first assigns the first 320 temperature values ​​to the first row, then assigns the next 320 temperature values ​​to the second row, and so on until all 240 rows are filled. In this arrangement process, the row and column assignment of each temperature value is determined only by its position in the output sequence, without introducing other judgment conditions, thus forming a pixel temperature array with a clear row and column structure. For example, in this array, the temperature at the 120th row and 85th column can be clearly read as 53.75 ℃, and the temperature at the 118th row and 90th column is 61.25 ℃, and it is ensured that any temperature value in this array can be directly located by row and column index.

[0098] S102: Based on the pixel temperature array value, call the row and column coordinate information of the current frame of the infrared thermal imaging camera, the corresponding pixel temperature value and its row and column coordinates, perform coordinate alignment consistency check and position correction, and generate a pixel coordinate temperature correspondence table.

[0099] First, the row scan count and column scan count are read from the current working state of the camera, and these counts are matched with the index positions in the pixel temperature array. For example, when the pixel index number is 38485, the theoretical row number 120 is obtained by dividing the index by the column width 320, and the theoretical column number 85 is obtained by subtracting the product of the row number and the column width from the index. Then, the theoretical row and column numbers are compared with the actual scan row and column numbers recorded by the camera. If the comparison results show that there is a deviation between the two, the deviation values ​​Δr and Δc are calculated and processed according to the pre-set interval rules. When |Δr| or |Δc| is less than or equal to 2, the theoretical row and column numbers are directly used to replace the recorded row and column numbers. When the deviation value is greater than 2, the pixel is recorded as an anomaly and stored separately. In an actual monitoring, 5 pixels are randomly selected from a frame for verification. The verification data is sorted after horizontal and vertical replacement as follows.

[0100] Table 1: Pixel Row and Column Coordinate Consistency Check Table

[0101] project Pixel 1 Pixel 2 Pixel 3 Pixel 4 Pixel 5 Pixel index number 38485 38170 39010 37200 40125 Theoretical line number 120 119 121 116 125 Theoretical column number 85 210 50 80 125 Record line number 120 119 123 116 125 Record column number 86 210 50 83 126 Line deviation Δr 0 0 2 0 0 Column deviation Δc 1 0 0 3 1 Processing results Column number replacement remain unchanged line number replacement Log error Column number replacement

[0102] By reading each column of data in the table and performing difference calculations and interval judgments, it can be confirmed that coordinate replacement is performed when the column deviation is 1 or the row deviation is 2, while when the column deviation reaches 3, it is classified as an abnormal record. After completing the same processing for all pixels in the entire frame, the system organizes all the verified pixels into a set of correspondences containing the corrected row coordinates, corrected column coordinates, and corresponding temperature values. For example, the record items (120, 85, 53.75 ℃), (119, 210, 58.40 ℃), and (121, 50, 62.10 ℃) can be obtained, and it is ensured that each record is generated by a clear comparison and judgment process.

[0103] S103: Based on the pixel coordinate temperature correspondence table, call the row and column coordinate mapping relationship in the camera calibration table, perform calibration correction calculation and complete the row and column coordinate mapping replacement, rewrite the corrected coordinates and corresponding temperature values ​​into the matrix structure, and generate the thermal image temperature matrix.

[0104] First, the corresponding values ​​between the original and corrected row numbers are read from the calibration data. For example, in the calibration table, original row 118 corresponds to corrected row 117.8, and original row 121 corresponds to corrected row 120.5. Simultaneously, the column direction correspondence is read, such as original column 85 corresponding to corrected column 84.8, and original column 90 corresponding to corrected column 89.6. Then, for each pixel coordinate temperature record, its row and column numbers are extracted, and the corresponding mapping nodes above and below are found in the calibration table. The corrected coordinates are calculated through numerical interpolation. For example, for original row number 120, interpolation between points 118 and 121 yields a corrected row value of 119.6; for original column number 85, interpolation between the corresponding values ​​of 84 and 90 yields a corrected column value of 84.8. Next, the corrected coordinate values ​​are processed for precision. If the decimal part is greater than 0.5, one decimal place is retained; otherwise, the integer part is rounded down. In this example, 119.6 and 84.8 both meet the retention condition. Finally, the corrected coordinates are compared with the original temperature value of 53.75. The ℃ is rewritten to the corresponding position in the temperature matrix. The matrix index is determined based on the corrected rows and columns. For example, 53.75 ℃ is written to the storage unit corresponding to row 119.6 and column 84.8. After repeating the above reading, interpolation, accuracy judgment and writing process for all pixels in the busbar cable joint area, a set of thermal image temperature matrix data after calibration and mapping correction is formed.

[0105] Please see Figure 3 The specific steps of S2 are as follows:

[0106] S201: Based on the thermal image temperature matrix, filter and compare the temperature values ​​in the matrix with the set temperature threshold, retain the row and column coordinates of pixels that exceed the temperature threshold, and generate a set of pixel coordinates exceeding the threshold.

[0107] First, each matrix cell is read row by row and column by column according to the matrix storage order. The row and column coordinates of the corresponding pixel, along with the temperature value at that location, are read and combined into a comparable data record. Then, a temperature threshold is introduced as a filtering criterion. This threshold is derived from statistical results of the distribution cabinet under long-term stable operation. In actual operation, temperature matrix data was continuously collected for 72 hours from the circuit breaker contacts and busbar cable joint areas, with a sampling period of 5 minutes, resulting in 864 frames of data. After summarizing the pixel temperatures at the same physical location, the average and dispersion were calculated. The statistical results showed that the average temperature in this area under stable operation was 48.6 ℃, and the standard deviation was 4.2 ℃. Based on this, the threshold was set to the sum of the average value and three times the standard deviation, i.e., 61.2 ℃. Then, a numerical comparison operation is performed on the temperature value of each pixel in the matrix. When the read temperature value is greater than 61.2 ℃, the row and column coordinates of that pixel are immediately written to the over-threshold cache set. When the temperature value is less than or equal to 61.2 ℃, the threshold is set to 61.2 ℃. When the temperature exceeds a certain threshold, no write operation is performed. For example, in an actual monitoring frame, the matrix position (120, 85) corresponds to a temperature of 68.2 ℃, which meets the comparison condition and is recorded as a valid coordinate. However, the matrix position (118, 90) corresponds to a temperature of 59.8 ℃, which does not meet the condition and is not entered into the cache. Through the above pixel-by-pixel reading, value-by-value comparison and condition writing process, after traversing the entire matrix, only the pixel coordinate records of the temperature exceeding the threshold are retained, and finally a set of over-threshold pixel coordinates composed of several row and column coordinate pairs is formed.

[0108] S202: Based on the set of pixel coordinates exceeding the threshold, determine the connectivity of pixel row and column coordinates in the up, down, left, and right directions, perform adjacent pixel coordinate merging operation, divide continuous coordinates into groups and count the number of coordinates in each group, and obtain the connected region area set.

[0109] First, all coordinate pairs within the coordinate set are sorted in ascending order of row coordinates. If row coordinates are the same, they are then sorted in ascending order of column coordinates, ensuring a stable traversal order for the coordinate sequence. Then, starting from the first sorted coordinate, the current coordinate is retrieved sequentially and its adjacency is calculated with the remaining coordinates not yet merged. Specifically, the absolute values ​​of the differences between two coordinates in the row and column directions are calculated. If the row difference is no greater than 1 and the column difference is no greater than 1, it is determined that the two pixels have a direct connection in the vertical or horizontal direction, and they are grouped into the same coordinate group. For example, coordinates (120, 85) and (120, 85) are grouped together. The coordinates (120, 85) have a row difference of 0 and a column difference of 1, which meet the condition and are assigned to the same group. However, the coordinates (120, 85) and (122, 85) have a row difference of 2, which do not meet the condition and are not merged. The same adjacency judgment is performed on the newly added coordinates in the current group until no more mergeable coordinates appear in the group. Then, the number of coordinates contained in the group is counted and used as the area value of the group. In one actual monitoring data, the set of super-threshold pixel coordinates contains a total of 23 coordinates. After the above item-by-item judgment and merging process, 3 consecutive coordinate groups are formed. The merging results and the number statistics are shown in the table below.

[0110] Table 2: Statistics on Connectivity Grouping and Area of ​​Pixels Exceeding Threshold

[0111] Connected component numbering Includes pixel coordinates example Number of coordinates Connected component A (120,85),(120,86),(121,85),(121,86),(121,87),(122,86) 6 Connected component B (124,90),(124,91),(124,92),(125,90),(125,91),(125,92),(126,91),(126,92),(126,93) 9 Connected domain C (130,78),(130,79),(131,78),(131,79),(132,78),(132,79),(133,79),(133,80) 8

[0112] The number of coordinates in each connected component in the table comes directly from the counting results of the coordinate entries in the corresponding group. After traversing, merging and counting all coordinates, the number of coordinates in each group is organized into an area value set {6, 9, 8}, thus obtaining the area set of the connected components.

[0113] S203: Based on the connected region area set, call the temperature value of the corresponding pixel coordinate position, calculate the average value of the row coordinate and column coordinate of the connected region as the centroid coordinate, and compare the temperature values ​​in the connected region to extract the maximum temperature value, generating an abnormal hot area connected region feature set.

[0114] First, for a specific connected component group, read the row and column coordinates of all pixels within that group, and then perform summation and averaging operations on the row and column coordinate sets respectively. The average row coordinate is obtained by adding all row coordinates within the group and dividing by the number of pixels. The average column coordinate is calculated in the same way. Together, they constitute the centroid coordinates of the connected component. For example, for connected component A, the row coordinates of its 6 pixels are 120, 120, 121, 121, 121, 122. Adding them together gives 725, which is then divided by 6 to obtain the average row coordinate. The mean is 120.83, and the column coordinates are 85, 86, 85, 86, 87, 86. Adding these together gives 515, which, divided by 6, yields the column average of 85.83. Then, the corresponding pixel temperature values ​​are extracted within the same connected component. A comparison operation is performed on each temperature value. Initially, the first pixel temperature is set to the current maximum value, and then compared with the temperatures of subsequent pixels. When a higher value is found, the maximum value is updated. In connected component A, the corresponding temperature values ​​are 65.4, 66.1, 68.2, 67.5, 66.8, and 64.9. After comparison, the maximum temperature value was found to be 68.2 ℃. The above centroid calculation and temperature comparison operations were repeated for connected domains B and C. For example, the centroid coordinates of connected domain B were calculated to be (125.0, 91.2), and the maximum temperature value was 71.3 ℃. The centroid coordinates of connected domain C were calculated to be (131.4, 78.9), and the maximum temperature value was 69.5 ℃. Finally, the area value, centroid coordinates, and maximum temperature value corresponding to each connected domain were organized into a structured record to form a feature set of connected domains of the abnormal hot zone.

[0115] Please see Figure 4 The specific steps of S3 are as follows:

[0116] S301: Based on the abnormal hot zone connected component feature set, obtain the row and column coordinate values ​​of the centroid of the connected component in consecutive frames, match the connected components of adjacent frames based on the principle of minimum centroid distance to determine the corresponding connected component, calculate the centroid coordinate difference of the corresponding connected component, and generate a centroid offset consistency judgment value.

[0117] First, the centroid row and column coordinates of each connected component in consecutive frames are read sequentially. Connected components of the same physical region in adjacent frames are then compared one by one as candidate matching objects. During execution, a connected component from the previous frame is fixed, and its centroid coordinates are extracted as the reference coordinates. For example, in frame 15, the centroid row coordinate of a connected component is read as 120.8, and the column coordinate as 85.9. Then, in frame 16, the centroid coordinates of all connected components are read as the set to be compared. For example, if there are two candidate connected components with centroid coordinates of (121.3, 86.4) and (130.2, 92.1), the Euclidean distance between the reference coordinates and each candidate coordinate is calculated. The difference in row coordinates is obtained by subtracting the difference in column coordinates. Finally, the square root of the sum of the squared differences is taken to obtain the distance value. In this example, the row difference of the first candidate connected component is 0.5, the column difference is 0.5, the sum of squares is 0.5, and the square root distance is 0.71. The row difference of the second candidate connected component is 9.4, the column difference is 6.2, the sum of squares is 126.4, and the square root distance is 11.2. Then, all the calculated distance values ​​are compared numerically, and the candidate connected component corresponding to the minimum distance is selected as the matching result. The minimum distance value is recorded as the centroid offset of the current frame pair. In the continuous frame calculation process, a consistency judgment threshold of 1.5 is introduced as a comparison benchmark. After repeatedly performing coordinate extraction, distance calculation, minimum value filtering and threshold comparison operations frame by frame, a set of centroid offset consistency judgment values ​​is formed. In order to make it easier to intuitively present the matching and distance changes between continuous frames, the original table is swapped to obtain the following data example.

[0118] Table 3: Centroid Matching and Offset Distance Calculation Table for Connected Components in Continuous Frames

[0119] project Record 1 Record 2 Record 3 Record 4 Previous frame number 15 15 16 17 Next frame number 16 16 17 18 previous frame centroid coordinates (120.8,85.9) (120.8,85.9) (121.3,86.4) (121.9,86.8) Candidate centroid coordinates of the next frame (121.3,86.4) (130.2,92.1) (121.9,86.8) (123.5,88.6) Calculate distance 0.71 11.2 0.72 2.30 Is it the smallest? yes no yes yes Judgment result Consistent Inconsistent Consistent Inconsistent

[0120] The swapped table allows for direct horizontal comparison of centroid distance values ​​and judgment status across different frame pairs, categorized by field dimensions.

[0121] S302: Based on the centroid offset consistency judgment value, collect the frame sequence number of the connected component that meets the consistency condition, count the number of consecutively occurring frames and record the corresponding timestamp, and generate an alarm timestamp sequence.

[0122] First, the system iterates through the set of judgment values ​​in frame sequence, retaining only adjacent frame pairs with consistent judgment results. The corresponding frame numbers are then extracted to form a consecutive frame candidate sequence. For example, in the judgment results of frames 15 to 18, frames 15-16 and 16-17 are marked as consistent, while frames 17-18 are marked as inconsistent. Therefore, the system merges frames 15, 16, and 17 into the same consecutive segment. The number of frames within this consecutive segment is then counted by calculating the total number of different frame numbers within the segment. In the example above, the consecutive segment contains 3 frames, so the number of consecutively occurring frames is recorded as 3. Then, according to the system... The frame sequence number is mapped to the time. The acquisition time of each frame is written into the record. For example, if the system acquisition frequency is 1 frame / minute, then the time of the 15th frame is 10:15, the time of the 16th frame is 10:16, and the time of the 17th frame is 10:17. Then, the start time of the consecutive segment, 10:15 and the end time, 10:17 are written into the time sequence as a set of time stamps. Then, the set of judgment values ​​is traversed again. When consecutive frames that meet the consistency condition are encountered again, the frame sequence number extraction, quantity statistics and time mapping operations are repeated. Finally, an alarm timestamp sequence consisting of multiple time intervals and the number of consecutive frames that appear in them is obtained.

[0123] S303: Based on the alarm timestamp sequence, calculate and compare the adjacent timestamp interval value with the alarm interval threshold, and at the same time calculate and compare the number of consecutively occurring frames with the continuous frame threshold to generate the rollback direction determination state;

[0124] First, extract two adjacent time interval records in chronological order. Read the end time of the previous record and the start time of the next record, convert the time to minutes, and then perform a subtraction operation to obtain the interval value. For example, if the end time of the previous interval is 10:17 and the start time of the next interval is 10:22, the time interval is 5 minutes. Then, compare this time interval with the alarm interval threshold of 5 minutes to obtain the result of whether the time is continuous or discontinuous. Next, for each time interval record, read its number of consecutive frames and compare it with the continuous frame threshold of 5 frames. In the actual data, the number of consecutive frames corresponding to the time interval 10:15 to 10:17 is 3, and the number of consecutive frames corresponding to the time interval 10:22 to 10:27 is 6, respectively, to obtain the comparison results of continuous insufficiency and continuous satisfaction. Finally, combine the time interval comparison result and the frame number comparison result to form the backtracking direction determination status of the corresponding abnormal segment.

[0125] Please see Figure 5 The specific steps of S4 are as follows:

[0126] S401: Based on the rollback direction determination status, read the temperature threshold rollback step size value and the maximum temperature change limit value in the current configuration, compare the size relationship between the two types of values, limit the rollback step size value according to the maximum temperature change limit, and generate a threshold rollback constraint value.

[0127] First, the temperature threshold fallback step size and the maximum temperature change limit are read sequentially from the current running configuration file. Both parameters are maintained in numerical form over a long period and dynamically updated during operation. The temperature threshold fallback step size is derived from the statistical results of multiple historical manual and automatic adjustment records. For example, in the scenario of infrared monitoring of distribution cabinets, a review of 48 threshold adjustment records over the past 30 days revealed that the adjustment range was mainly between 0.8℃ and 2.4℃, with an average of approximately 1.9℃. Therefore, the temperature threshold fallback step size is set to 2.0℃ in the current configuration. The maximum temperature change limit is derived from the statistical analysis of the maximum temperature change in abnormal hot zones across consecutive frames. For example, during a continuous anomaly... The maximum temperature changed from 68.2℃ to 70.1℃, corresponding to a change of 1.9℃. Based on the statistical results of nearly 20 similar abnormal events, the maximum change did not exceed 2.2℃. Therefore, the upper limit of the maximum temperature change currently read by the system is 2.2℃. Subsequently, the system performs a step-by-step comparison between the backoff step size and the upper limit of the maximum temperature change. When the backoff step size is less than or equal to the upper limit, the original backoff step size is directly retained. When the backoff step size is greater than the upper limit, a replacement operation is performed to correct the backoff step size to the upper limit. This comparison and replacement process is completed entirely based on the numerical relationship. To clearly show the limiting results under different configuration conditions, the multiple sets of parameters collected in actual operation are summarized in the following table.

[0128] Table 4: Comparison of Temperature Threshold Backstep Size and Limiting Results

[0129] Serial Number Original backtracking step size (°C) Maximum temperature change limit (°C) Comparison results Threshold backoff constraint value (°C) 1 2.0 2.2 2.0 ≤ 2.2 2.0 2 2.5 2.2 2.5 ＞ 2.2 2.2 3 1.5 1.8 1.5 ≤ 1.8 1.5 4 3.0 2.0 3.0 ＞ 2.0 2.0

[0130] By performing comparison and replacement operations on each row of the data in the table, a unique threshold backoff constraint value can be obtained for each set of configuration conditions. This constraint value serves as the direct input result for subsequent threshold adjustment calculations.

[0131] S402: Based on the threshold backoff constraint value, calculate the difference between the temperature threshold backoff step size and the maximum temperature change limit value to obtain the threshold adjustment calculation amount;

[0132] First, the current temperature threshold backoff constraint value and the corresponding maximum temperature change limit value are read synchronously. Then, a subtraction operation is performed on the two to obtain the difference result. This difference is directly calculated by using the threshold backoff constraint value as the minuend and the maximum temperature change limit value as the subtrahend. For example, in the first group of data in the previous section, the threshold backoff constraint value is 2.0℃ and the maximum temperature change limit is 2.2℃, so 2.0−2.2 is performed, resulting in a difference of −0.2℃. In the second group of data, both are 2.2℃, so 2.2−2.2 is performed, resulting in a difference of 0℃. In the third group of data, 1.5−1.8 is performed, resulting in a difference of −0.3℃. This difference value directly reflects the numerical relationship between the current backoff step size and the allowed change limit. The entire calculation process only involves parameter reading and subtraction operations, without introducing other correction factors. By independently performing the same difference calculation for each threshold adjustment scenario, the system obtains the corresponding threshold adjustment calculation amount. This calculation amount is stored separately in numerical form, providing clear basic data for subsequent directional correction steps.

[0133] S403: Based on the threshold adjustment calculation amount, determine the direction identifier corresponding to the state according to the rollback direction, assign positive or negative signs to the threshold adjustment calculation amount and complete the sign correction, and generate the threshold adjustment amount value;

[0134] First, the system reads the direction identifier from the rollback direction determination status. This identifier distinguishes whether the current threshold adjustment should change in an increasing or decreasing direction. For example, if the abnormal hot zone persists and the temperature change has not yet decreased, the direction identifier is recorded as a decreasing direction. Conversely, if the abnormality gradually subsides or the overall temperature decreases, the direction identifier is recorded as an increasing direction. The system then reads the corresponding threshold adjustment calculation value, such as -0.2℃, -0.3℃, or 0℃, and performs sign processing on this value according to the direction identifier. When the direction identifier is a decreasing direction, the negative sign of the calculated value is directly retained. When the direction indicator is increasing, the absolute value of the calculated value is taken first, and then a positive sign is assigned. In the actual calculation example, if the calculated value is -0.3℃ and the direction indicator is decreasing, the final threshold adjustment is -0.3℃. If the direction indicator corresponding to the same calculated value is increasing, the final threshold adjustment is +0.3℃. When the calculated value is 0℃, the result after sign correction remains 0 regardless of the direction indicator. By reading the direction indicator, judging the direction type, and performing the specific operations of taking the absolute value or keeping the sign, the system generates a unique threshold adjustment value for each threshold adjustment.

[0135] Please see Figure 6 The specific steps of S5 are as follows:

[0136] S501: Based on the threshold adjustment value, read the current temperature threshold of the edge computing terminal, and combine it with the direction identifier in the back-off direction determination state to determine the temperature threshold adjustment direction and obtain the threshold update direction identifier value;

[0137] First, the system reads the currently effective temperature threshold value from the local non-volatile configuration area. This threshold value originates from the alarm judgment register write result of the previous cycle. For example, in the scenario of monitoring circuit breaker contacts in a distribution cabinet, the current threshold is recorded as 62.0℃. Then, the system synchronously reads the threshold adjustment value, which is obtained from previous calculation steps and already has a clear positive or negative sign. For example, the threshold adjustment value read this time is −0.3℃. Next, the system reads the direction identifier stored in the rollback direction judgment status. This direction identifier represents the adjustment direction using a discrete encoding method. For example, a value of 0 indicates a downward threshold adjustment, and a value of 1 indicates an upward threshold adjustment. In this example, the direction identifier read is 0, corresponding to a downward adjustment direction for the threshold. The system then checks the logical relationship between the direction identifier and the threshold adjustment value. Specifically, it determines whether the threshold adjustment value is less than 0 when the direction identifier is 0, and whether the threshold adjustment value is greater than 0 when the direction identifier is 1. If the determination result is true, the direction is confirmed to be consistent. In this example, the direction identifier is 0 and the threshold adjustment value is −0.3℃, which meets the condition of consistent direction. Therefore, the system determines the current threshold update direction as the downward adjustment direction and writes the result into the internal direction state variable, forming a threshold update direction identifier value that can be directly called by subsequent numerical calculation steps.

[0138] S502: Based on the threshold update direction identifier value, calculate the addition and subtraction result of the current temperature threshold value and the threshold adjustment value, write the calculated temperature threshold into the edge computing terminal alarm judgment register, and generate the updated temperature threshold value.

[0139] First, the current temperature threshold value is read from the alarm judgment register of the edge computing terminal, and the threshold adjustment value is read simultaneously. Then, the numerical calculation method is determined according to the threshold update direction indicator value. When the direction indicator is downward adjustment, an algebraic addition operation is performed on the current threshold value and the threshold adjustment value. When the direction indicator is upward adjustment, the addition operation is also performed, but the adjustment value is positive. In actual operation, the calculation results under different initial thresholds and different adjustment combinations are different. To clearly illustrate this process, the rows and columns of the aforementioned example table are swapped and reorganized as follows, so that each set of parameters can be compared in the horizontal dimension.

[0140] Table 5: Example of Temperature Threshold Update Calculation Process

[0141] project Example 1 Example 2 Example 3 Example 4 Current temperature threshold (°C) 62.0 61.5 63.0 60.8 Threshold adjustment value (°C) −0.3 +0.2 −0.5 +0.4 Update direction down up down up Operational expressions 62.0+(−0.3) 61.5+0.2 63.0+(−0.5) 60.8+0.4 Updated threshold (°C) 61.7 61.7 62.5 61.2

[0142] By reading each data item in the table column by column and performing the corresponding addition operation, the updated temperature threshold value can be uniquely determined for each set of conditions. Then, the system writes the calculation result to the specified address of the alarm judgment register inside the edge computing terminal, overwriting the original threshold record and forming the updated temperature threshold value.

[0143] S503: Based on the updated temperature threshold value, write the value to the corresponding trigger threshold storage unit of the audible and visual alarm, and read the current trigger threshold parameter to update the value, generating an abnormal warning trigger threshold configuration.

[0144] First, the system reads the latest effective temperature threshold value from the alarm determination register through the internal communication interface. For example, if the read value is 61.7℃, then the system locates the address of the storage unit inside the audible and visual alarm that stores the temperature trigger threshold according to the address mapping relationship established during the system initialization phase. Next, the value of 61.7℃ is written into this storage unit, overwriting the previously stored old threshold value, for example, the original threshold was 62.0℃. After the write operation is completed, the system immediately performs a readback operation on this storage unit, rereads the current trigger threshold parameter and compares it with the written value. In this example, the readback result is still 61.7℃. The system then records this readback value as the currently valid audible and visual alarm trigger threshold configuration value. Through the above continuous execution steps of reading and updating the threshold, writing to the alarm trigger threshold storage unit, and readback verification, the synchronous update of the abnormal warning trigger threshold configuration is completed.

[0145] An edge-intelligent-driven real-time infrared image analysis and anomaly early warning system includes:

[0146] The infrared acquisition module is used to perform S1: acquire the thermal image frame data output by the infrared thermal imaging camera monitoring the circuit breaker contacts and busbar cable joints of the distribution cabinet, collect the temperature values ​​of the pixels in the thermal image frame and write them into the temperature matrix according to the row and column coordinates, and generate the thermal image temperature matrix by corresponding the coordinates of the temperature matrix and the camera calibration table.

[0147] The hot zone extraction module is used to perform S2: based on the thermal image temperature matrix, it filters the pixel coordinates with temperature values ​​exceeding the temperature threshold and performs connectivity determination, calculates the area and centroid coordinates of the connected region and extracts the maximum temperature value of the connected region, and generates an abnormal hot zone connected region feature set.

[0148] The timing determination module is used to execute S3: based on the abnormal hot zone connected component feature set, calculate and determine whether the centroid coordinate offset of the connected components in consecutive frames is consistent, count the number of continuous frames of the hot zone and generate an alarm timestamp sequence, compare the alarm interval value with the alarm interval threshold, and at the same time compare the number of continuous frames of the hot zone with the continuous frame threshold to generate a backtracking direction determination status.

[0149] The threshold backoff module is used to execute S4: determine the state based on the backoff direction, read and compare the temperature threshold backoff step size with the upper limit of the maximum temperature change, calculate the threshold adjustment amount, and generate the threshold adjustment amount value;

[0150] The early warning configuration module is used to execute S5: adjust the value according to the threshold, read the current temperature threshold and select the threshold update direction according to the backtracking direction, calculate the updated temperature threshold and write it to the edge computing terminal alarm judgment register, synchronize the temperature threshold to the trigger threshold of the audible and visual alarm, and generate the abnormal early warning trigger threshold configuration.

[0151] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for real-time infrared image analysis and anomaly early warning driven by edge intelligence, characterized in that, Includes the following steps: S1: Acquire thermal image frames of the monitoring field of view of the circuit breaker contacts and busbar cable joints of the distribution cabinet by the infrared thermal imaging camera, extract the pixel temperature values ​​and write them into the temperature matrix to generate a thermal image temperature matrix. S2: Based on the thermal image temperature matrix, filter pixels that exceed the temperature threshold and determine the connected components, calculate and record the area and centroid coordinates of the connected components, and generate an abnormal hot area connected component feature set. S3: Based on the abnormal hot zone connected component feature set, calculate the consistency of centroid coordinate offset of consecutive frames, count the number of continuous frames of hot zone and generate alarm timestamp sequence, compare alarm interval with interval threshold and continuous frame threshold, and generate backtracking direction determination status. S4: Determine the state based on the back-off direction, compare the temperature threshold back-off step size with the upper limit of the maximum temperature change, and calculate and generate the threshold adjustment amount; S5: Based on the threshold adjustment amount, combined with the backtracking direction determination status update temperature threshold, the update result is written into the edge computing terminal alarm determination register and synchronously configured to the audible and visual alarm trigger threshold to form an abnormal warning trigger threshold configuration.

2. The edge-driven real-time infrared image analysis and anomaly early warning method according to claim 1, characterized in that, The thermal imaging temperature matrix includes pixel temperature value distribution, row and column coordinate indexes, and calibration coordinate mapping relationships; the abnormal hot zone connected component feature set includes connected component area, connected component centroid coordinates, and maximum connected component temperature value; the rollback direction determination status specifically refers to alarm interval comparison conclusion, continuous frame count comparison conclusion, and threshold rollback direction type; the threshold adjustment amount specifically includes threshold update step size, single change limit, and threshold adjustment amount selection value; the abnormal warning trigger threshold configuration includes warning temperature threshold setting value, alarm determination register write value, and audible and visual alarm trigger threshold value.

3. The edge-driven real-time infrared image analysis and anomaly early warning method according to claim 1, characterized in that, The specific steps of S1 are as follows: S101: Acquire thermal image frame data output by infrared thermal imaging camera monitoring the field of view of circuit breaker contacts and busbar cable joints in distribution cabinet, collect temperature values ​​corresponding to pixels in the frame, arrange pixel temperatures according to the original row and column coordinates of the thermal image frame, and obtain pixel temperature array values. S102: Based on the pixel temperature array value, call the row and column coordinate information of the current frame of the infrared thermal imaging camera, the corresponding pixel temperature value and its row and column coordinates, perform coordinate alignment consistency check and position correction, and generate a pixel coordinate temperature correspondence table. S103: Based on the pixel coordinate temperature correspondence table, call the row and column coordinate mapping relationship in the camera calibration table, perform calibration correction calculation and complete the row and column coordinate mapping replacement, and rewrite the corrected coordinates and corresponding temperature values ​​into the matrix structure to generate a thermal image temperature matrix.

4. The edge-intelligent driven infrared image real-time analysis and anomaly early warning method according to claim 3, characterized in that, The specific steps of S2 are as follows: S201: Based on the thermal imaging temperature matrix, filter and compare the temperature values ​​in the matrix with the set temperature threshold, retain the row and column coordinates of pixels that exceed the temperature threshold, and generate a set of pixel coordinates exceeding the threshold. S202: Based on the set of pixel coordinates exceeding the threshold, determine the connectivity of pixel row and column coordinates in the up, down, left, and right directions, perform adjacent pixel coordinate merging operation, divide continuous coordinates into groups and count the number of coordinates in each group, and obtain the connected region area set. S203: Based on the connected region area set, call the temperature value of the corresponding pixel coordinate position, calculate the average value of the row coordinate and column coordinate of the connected region as the centroid coordinate, and compare the temperature values ​​in the connected region to extract the maximum temperature value, thereby generating an abnormal hot zone connected region feature set.

5. The edge-driven real-time infrared image analysis and anomaly early warning method according to claim 4, characterized in that, The specific steps for S3 are as follows: S301: Based on the abnormal hot zone connected domain feature set, obtain the row and column coordinate values ​​of the centroid of the connected domain in consecutive frames, match the connected domains of adjacent frames based on the principle of minimum centroid distance to determine the corresponding connected domain, calculate the centroid coordinate difference of the corresponding connected domain, and generate a centroid offset consistency judgment value. S302: Based on the centroid offset consistency judgment value, collect the frame sequence number corresponding to the connected component that meets the consistency condition, count the number of consecutively occurring frames and record the corresponding time stamp, and generate an alarm timestamp sequence. S303: Based on the alarm timestamp sequence, calculate and compare the adjacent timestamp interval value with the alarm interval threshold, and at the same time calculate and compare the number of consecutively occurring frames with the continuous frame threshold to generate a rollback direction determination state.

6. The edge-driven real-time infrared image analysis and anomaly early warning method according to claim 5, characterized in that, The specific steps of S4 are as follows: S401: Based on the back-off direction determination state, read the temperature threshold back-off step size value and the maximum temperature change limit value in the current configuration, compare the size relationship of the two types of values, limit the back-off step size value according to the maximum temperature change limit, and generate a threshold back-off constraint value. S402: Based on the threshold backoff constraint value, calculate the difference between the temperature threshold backoff step size and the maximum temperature change limit value to obtain the threshold adjustment calculation amount; S403: Based on the threshold adjustment calculation, determine the direction identifier corresponding to the state according to the rollback direction, assign a positive or negative sign to the threshold adjustment calculation and complete the sign correction, and generate the threshold adjustment value.

7. The edge-driven infrared image real-time analysis and anomaly early warning method according to claim 6, characterized in that, The specific steps of S5 are as follows: S501: Based on the threshold adjustment value, read the current temperature threshold of the edge computing terminal, and combine it with the direction identifier in the back-off direction determination state to determine the temperature threshold adjustment direction and obtain the threshold update direction identifier value; S502: Based on the threshold update direction identifier value, calculate the addition and subtraction result of the current temperature threshold value and the threshold adjustment value, write the calculated temperature threshold into the edge computing terminal alarm judgment register, and generate the updated temperature threshold value. S503: Based on the updated temperature threshold value, write the value into the corresponding trigger threshold storage unit of the audible and visual alarm, and read the current trigger threshold parameter to update the value, thereby generating an abnormal warning trigger threshold configuration.

8. The edge-driven real-time infrared image analysis and anomaly early warning method according to claim 1, characterized in that, The infrared thermal imaging camera is an imaging device used to collect infrared radiation from the surface of the monitored equipment and output thermal image frame data of temperature distribution, and its output serves as the source of thermal image frame data. The circuit breaker contacts of the distribution cabinet are conductive contact parts located inside the distribution cabinet for circuit switching, and their surface temperature is one of the objects monitored by infrared. The busbar cable connector is a conductive connection component used to realize the electrical connection between the busbar and the cable, and its surface temperature is one of the objects monitored by infrared. The thermal image frame data is a frame of temperature distribution image data output by the infrared thermal imaging camera within a unit sampling period, which includes pixel coordinates and corresponding temperature values. The thermal image temperature matrix is ​​a two-dimensional temperature data set formed by arranging the temperature values ​​of each pixel in the thermal image frame data according to the row and column coordinate order. The temperature threshold is a temperature judgment benchmark used to distinguish between normal pixels and abnormal pixels, and its value comes from the local configuration parameters of the edge computing terminal. The connectivity determination is a process of judging the spatial adjacency relationship between adjacent pixel coordinates, used to verify whether abnormal pixels belong to the same abnormal region. The area of ​​the connected region is the number of pixels covered by the set of abnormal pixels formed by connectivity determination; The centroid coordinates are the center position coordinates calculated based on the coordinates of all pixels in the connected domain, and are used to characterize the spatial location of the abnormal hot zone. The abnormal hot zone connected domain feature set is a data set consisting of the connected domain area, centroid coordinates, and the maximum temperature value within the connected domain.

9. The edge-intelligent driven infrared image real-time analysis and anomaly early warning method according to claim 1, characterized in that, The continuous frame count of the hot zone refers to the frame count value of the connected domain that is identified as the same abnormal hot zone repeatedly in consecutive thermal image frames. The alarm timestamp sequence is a time-sequential data set formed by the edge computing terminal recording the trigger time of each abnormal warning; The alarm interval value is the time difference calculated from two adjacent alarm timestamps in the alarm timestamp sequence; The alarm interval threshold is a time-based benchmark for determining whether alarm triggering is in a dense state, and its value is obtained by the edge computing terminal configuration. The continuous frame threshold is a frame count criterion used to determine whether an abnormal hot zone persists, and its value comes from the edge computing terminal configuration parameters. The rollback direction determination state is a determination result generated based on the comparison between the alarm interval value and the alarm interval threshold, and between the hot zone continuous frame count and the continuous frame threshold. The temperature threshold backoff step size is the amount of temperature change used to adjust the warning temperature threshold, and its value comes from the edge computing terminal configuration parameters. The maximum temperature change limit is a temperature constraint value used to limit the adjustment range of a single temperature threshold, and its value is obtained by the edge computing terminal configuration. The threshold adjustment value is the temperature change obtained after filtering based on the temperature threshold backoff step size and the upper limit of the maximum temperature change. The abnormal warning trigger threshold configuration is a set of warning threshold parameters formed by applying the threshold adjustment value to the current temperature threshold, which is used for abnormal warning determination.

10. An edge-intelligent driven real-time infrared image analysis and anomaly early warning system, characterized in that, The system is used to implement the edge-intelligent driven infrared image real-time analysis and anomaly early warning method according to any one of claims 1-9, the system comprising: The infrared acquisition module is used to perform S1: acquire the thermal image frame data output by the infrared thermal imaging camera monitoring the circuit breaker contacts and busbar cable joints of the distribution cabinet, acquire the temperature value of the pixel in the thermal image frame and write it into the temperature matrix according to the row and column coordinates, and generate the thermal image temperature matrix by corresponding to the coordinates of the temperature matrix and the camera calibration table. The hot zone extraction module is used to perform S2: based on the thermal image temperature matrix, filter the pixel coordinates whose temperature values ​​exceed the temperature threshold and perform connectivity determination, calculate the area of ​​the connected domain and the centroid coordinates and extract the maximum temperature value of the connected domain, and generate an abnormal hot zone connected domain feature set. The timing determination module is used to execute S3: calculate and determine whether the centroid coordinate offset of the connected domain in consecutive frames is consistent based on the abnormal hot zone connected domain feature set, count the number of continuous frames of the hot zone and generate an alarm timestamp sequence, compare the alarm interval value with the alarm interval threshold, and at the same time compare the number of continuous frames of the hot zone with the continuous frame threshold to generate a rollback direction determination state. The threshold backoff module is used to execute S4: based on the backoff direction determination state, read and compare the temperature threshold backoff step size with the upper limit of the maximum temperature change, calculate the threshold adjustment amount, and generate the threshold adjustment amount value; The early warning configuration module is used to execute S5: adjust the value according to the threshold, read the current temperature threshold and select the threshold update direction according to the backtracking direction, calculate the updated temperature threshold and write it into the edge computing terminal alarm judgment register, synchronize the temperature threshold to the trigger threshold of the audible and visual alarm, and generate an abnormal early warning trigger threshold configuration.

Images