Control room autonomous inspection operation device and method

Abstract

The invention relates to the technical field of infrared image processing, in particular to an autonomous inspection operation device and method for a control room, and solves the technical problem of low temperature detection precision in the prior art. The device comprises a data acquisition unit which is used for acquiring infrared image data of equipment in a control room; the infrared image data is used for representing an infrared time domain signal acquired based on infrared radiation reflection of the surface of the equipment in the control room; the signal processing unit is used for carrying out adaptive time-frequency transformation based on an infrared time-domain signal in the infrared image data to obtain a corresponding time-frequency matrix; the time-frequency matrix is used for representing energy distribution of infrared radiation on the surface of equipment in the control room at different times and different frequencies; and the inspection unit is used for separating the target heat radiation based on the time-frequency matrix and performing equipment temperature inspection according to the separated target heat radiation.

Description

Technical Field

[0001] This invention relates to the field of infrared image processing technology, specifically to a control room autonomous inspection device and method. Background Technology

[0002] In industrial automation and intelligent operation and maintenance systems, the control room serves as the core control hub for industrial production. The operating status of key equipment such as metal busbars and switch cabinets within the control room directly affects the safety and reliability of the entire production system. Among these, equipment temperature is a core indicator reflecting operating status. Abnormal overheating of equipment is often a precursor to faults such as short circuits and poor connections. Therefore, temperature monitoring of control room equipment is a crucial aspect of industrial operation and maintenance.

[0003] Currently, the mainstream inspection method is gradually shifting from manual inspection to automated inspection. Autonomous inspection devices that integrate infrared sensing and signal processing technologies can acquire infrared radiation information of equipment in real time, enabling automated temperature monitoring and anomaly warning. However, existing temperature inspection solutions are easily affected by environmental factors, resulting in insufficient temperature detection accuracy and failing to meet the needs of high-precision industrial-grade inspections. This may affect the timely detection of potential faults and reduce operational reliability. Summary of the Invention

[0004] To address the technical problem of low temperature detection accuracy in existing technologies, the present invention aims to provide a control room autonomous inspection device and method, the specific technical solution of which is as follows: This application provides a control room autonomous inspection device, including: The data acquisition unit is used to acquire infrared image data of the equipment in the control room; the infrared image data is used to characterize the infrared time-domain signal acquired based on the infrared radiation reflection from the surface of the equipment in the control room. The signal processing unit is used to perform adaptive time-frequency transformation based on the infrared time-domain signal in the infrared image data to obtain the corresponding time-frequency matrix; the time-frequency matrix is ​​used to characterize the energy distribution of infrared radiation on the surface of indoor equipment at different times and frequencies; The inspection unit is used to separate the target thermal radiation based on the time-frequency matrix and to perform equipment temperature inspection based on the separated target thermal radiation.

[0005] In one possible implementation, the signal processing unit is specifically used for: The window function width of the adaptive time-frequency transform is dynamically adjusted based on the frequency and duration of the infrared time-domain signal. The width of the dynamically adjusted window function is corrected based on the distance to the reflection source of the infrared time domain signal and the roughness of the corresponding device surface; The infrared time-domain signal is subjected to time-frequency transformation based on the corrected window function width to obtain the corresponding time-frequency matrix.

[0006] In one possible implementation, the signal processing unit is specifically used for: The distance influence factor is determined based on the distance to the reflection source, and the roughness influence factor is determined based on the surface roughness of the equipment; among them, the distance influence factor is negatively correlated with the distance to the reflection source, and the roughness influence factor is negatively correlated with the surface roughness of the equipment. The window function width is adjusted based on the distance influence factor and the roughness influence factor.

[0007] In one possible implementation, the inspection unit is specifically used for: Extract time-frequency eigenvectors from the time-frequency matrix; the time-frequency eigenvectors include at least one of frequency entropy, time-frequency energy moment, and phase consistency index; frequency entropy is used to characterize the uniformity of signal energy distribution in the frequency dimension; time-frequency energy moment is used to characterize the energy centroid of signal energy in the time-frequency domain; phase consistency index is used to characterize the degree of fluctuation of signal phase in the time-frequency domain; Target thermal radiation is separated based on time-frequency feature vectors.

[0008] In one possible implementation, the inspection unit is specifically used for: The time-frequency feature vector is input into the preset classification model to determine the signal category of the infrared time domain signal; the signal category includes heating signal and reflected light signal; the preset classification model adopts a random forest classifier, which is trained through historical sample data; Infrared time-domain signals classified as heat signals are separated from infrared image data and used as target thermal radiation.

[0009] In one possible implementation, the data acquisition unit includes a dual-band infrared-polarization composite sensor and a lidar; The dual-band infrared-polarization composite sensor is used to adjust the polarizer orientation in real time and obtain infrared time-domain signals through dual-band collaborative acquisition to obtain infrared image data; LiDAR is used to acquire spatial location data of indoor control equipment; the spatial location data includes three-dimensional point cloud data of the indoor control equipment, which is used to establish a mapping relationship between pixels in infrared image data and spatial locations of indoor equipment.

[0010] In one possible implementation, the device further includes a preprocessing unit for preprocessing the data acquired by the data acquisition unit. The preprocessing operations include: Non-uniformity correction, median filtering, and histogram equalization are performed on infrared image data; Outlier removal from spatial location data; Spatiotemporal registration is performed on infrared image data and spatial location data.

[0011] In one possible implementation, the inspection unit is specifically used for: The corrected emissivity parameter is determined based on the surface roughness of the equipment corresponding to the target thermal radiation, and the temperature value of the corresponding equipment surface is determined based on the target thermal radiation and the corrected emissivity parameter using the radiation formula. The corrected emissivity parameter is positively correlated with the surface roughness; the temperature value is positively correlated with the target thermal radiation and negatively correlated with the corrected emissivity parameter. Conduct equipment temperature inspections based on the temperature values ​​on the equipment surface.

[0012] In one possible implementation, the inspection unit is specifically used for: A temperature warning is triggered when the temperature on the surface of the equipment exceeds the rated temperature threshold.

[0013] This application provides a method for autonomous inspection of a control room, including: Infrared image data of equipment in the control room is acquired; the infrared image data is used to characterize the infrared time-domain signal acquired based on the infrared radiation reflection from the surface of the equipment in the control room. An adaptive time-frequency transformation is performed on the infrared time-domain signal in the infrared image data to obtain the corresponding time-frequency matrix; the time-frequency matrix is ​​used to characterize the energy distribution of infrared radiation on the surface of indoor equipment at different times and frequencies; The target thermal radiation is separated based on the time-frequency matrix, and the equipment temperature is inspected based on the separated target thermal radiation.

[0014] The present invention has the following beneficial effects: This application provides an autonomous inspection device and method for control rooms. It acquires infrared image data of equipment within the control room to obtain infrared time-domain signals based on the reflection of infrared radiation from the equipment surfaces. Then, through adaptive time-frequency transformation, the time-domain signals are converted into a multi-dimensional time-frequency matrix. By analyzing the energy distribution of infrared radiation at different times and frequencies, the reflected light and thermal radiation signals are separated, allowing for equipment temperature inspection based on the separated target thermal radiation. Compared to current temperature detection solutions, this application avoids environmental reflection interference during temperature inspection, thereby further improving temperature detection accuracy and inspection efficiency. It can be widely applied to control room inspection scenarios in industries such as power, petrochemicals, and energy. Attached Figure Description

[0015] To more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0016] Figure 1 This is a schematic diagram of the structure of a control room autonomous inspection device according to an embodiment of the present invention; Figure 2 This is a schematic diagram of another control room autonomous inspection device provided in one embodiment of the present invention; Figure 3 This is a flowchart illustrating a method for autonomous inspection of a control room, as provided in one embodiment of the present invention. Detailed Implementation

[0017] To further illustrate the technical means and effects adopted by the present invention to achieve its intended purpose, the following, in conjunction with the accompanying drawings and preferred embodiments, details the specific implementation, structure, features, and effects of a control room autonomous inspection operation device and method proposed according to the present invention. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0018] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0019] In industrial automation and intelligent operation and maintenance systems, the control room serves as the core control hub for industrial production. The operating status of key equipment such as metal busbars and switch cabinets within the control room directly affects the safety and reliability of the entire production system. Among these, equipment temperature is a core indicator reflecting operating status. Abnormal overheating of equipment is often a precursor to faults such as short circuits and poor connections. Therefore, temperature monitoring of control room equipment is a crucial aspect of industrial operation and maintenance.

[0020] Currently, the mainstream inspection method is gradually shifting from manual inspection to automated inspection. Autonomous inspection devices that integrate infrared sensing and signal processing technologies can acquire infrared radiation information of equipment in real time, enabling automated temperature monitoring and anomaly warning. However, existing temperature inspection solutions are easily affected by environmental factors.

[0021] For example, due to the presence of various light sources such as lighting fixtures and other heat-generating equipment in the control room environment, the high reflectivity of metal equipment surfaces to infrared radiation causes a large amount of ambient reflected radiation to be mixed into the collected infrared signals, interfering with the actual heat generation signals of the equipment. Therefore, the temperature detected by existing technologies is severely interfered with, resulting in insufficient temperature detection accuracy and failing to meet the requirements of high-precision industrial-grade inspections. This may, in turn, affect the timely detection of potential faults and reduce operational reliability.

[0022] In view of this, this application provides an autonomous inspection device and method for control rooms. It acquires infrared image data of equipment within the control room to obtain infrared time-domain signals based on the infrared radiation reflection from the equipment surface. Then, through adaptive time-frequency transformation, the time-domain signals are converted into a multi-dimensional time-frequency matrix. By analyzing the energy distribution of infrared radiation at different times and frequencies, the reflected light and thermal radiation signals are separated, allowing for equipment temperature inspection based on the separated target thermal radiation. Compared to current temperature detection schemes, this application avoids environmental reflection interference during temperature inspection, thereby further improving temperature detection accuracy and inspection efficiency. It can be widely applied to control room inspection scenarios in industries such as power, petrochemicals, and energy.

[0023] The following description, in conjunction with the accompanying drawings, details the specific scheme of the autonomous inspection device and method for control rooms provided by the present invention.

[0024] Please see Figure 1 The diagram shows a structural diagram of a control room autonomous inspection device according to an embodiment of the present invention. The device includes a data acquisition unit 101, a signal processing unit 102, and an inspection unit 103.

[0025] The data acquisition unit 101 is used to acquire infrared image data of the control room equipment.

[0026] Infrared image data is used to characterize the infrared time-domain signal collected based on the infrared radiation reflection from the surface of equipment in the control room. This infrared time-domain signal can characterize the change in the intensity of infrared radiation on the equipment surface over time. For example, when the equipment heats up, the intensity of the infrared time-domain signal will increase with the increase of temperature; when reflected by ambient light, the signal intensity will fluctuate with the change of reflection angle.

[0027] For example, the data acquisition unit 101 may include a dual-band infrared-polarization composite sensor and a lidar. The dual-band infrared-polarization composite sensor is used to adjust the polarizer orientation in real time and acquire infrared time-domain signals through dual-band collaborative acquisition to obtain infrared image data. The lidar is used to acquire spatial position data of the control room equipment.

[0028] This application utilizes a dual-band infrared-polarization composite sensor deployed in conjunction with a lidar system to achieve multi-dimensional acquisition of infrared radiation signals and spatial information of reflection sources from metal surfaces. The dual-band infrared-polarization composite sensor includes a long-wave infrared module and a short-wave infrared module, responding to the device's own heating and reflected light from the sensitive environment, respectively. Specular reflection can be dynamically suppressed by adjusting the polarizer.

[0029] The lidar can scan and control the spatial location data of indoor devices in real time. This spatial location data can include three-dimensional point cloud data of the indoor devices, used to establish a mapping relationship between pixels in infrared image data and the spatial location of the indoor devices. For example, the spatial location data can also include spatial information such as the three-dimensional coordinates, distance, and reflection intensity of the indoor reflective source. Thus, this application can construct a multimodal dataset containing spectral characteristics, polarization attributes, and spatial location.

[0030] To address the signal interference problem caused by specular reflection from metal surfaces, this application utilizes dual-band collaborative acquisition. It leverages the characteristics of long-wave infrared radiation emphasizing the low reflectivity of the device itself and short-wave infrared radiation emphasizing the high reflectivity of the environment, marking suspicious reflective areas based on the intensity differences between the bands. Specifically, the device's own heating signal typically exhibits greater intensity in long-wavelength areas than in short-wavelength areas, while the opposite is true for environmental reflection signals.

[0031] Furthermore, this application can also adjust the polarizer direction in real time according to the reflection angle through polarization filtering. For example, it can adjust according to a polarization degree threshold. When the current polarization degree is detected to be greater than the polarization degree threshold (e.g., 0.7), the specular reflection signal is suppressed and the device's own unpolarized thermal radiation is preserved by rotating the polarizer to the orthogonal direction of the reflected light (e.g., vertical polarization filtering is used when the reflected light is horizontally polarized).

[0032] The signal processing unit 102 is used to perform adaptive time-frequency transformation based on the infrared time-domain signal in the infrared image data to obtain the corresponding time-frequency matrix.

[0033] The time-frequency matrix is ​​used to characterize the energy distribution of infrared radiation on the surface of equipment in the control room at different times and frequencies. For example, the rows of the time-frequency matrix can represent the time dimension, and the columns can represent the frequency dimension. The element values ​​in the time-frequency matrix characterize the energy distribution of infrared radiation on the surface of the equipment at the corresponding time and frequency. For example, the equipment heating signal usually exhibits low frequency and time continuity due to slow temperature changes, while ambient reflected light usually exhibits high frequency and short duration non-stationary high-frequency characteristics.

[0034] Adaptive time-frequency transformation refers to a signal processing method that dynamically adjusts transformation parameters based on the characteristics of the infrared time-domain signal (such as frequency and duration) during time-frequency conversion. For example, this application can dynamically adjust the window function width and perform time-frequency transformation on the infrared time-domain signal according to the adjusted window function width. Inspection unit 103 is used to separate target thermal radiation based on time-frequency matrix and to perform equipment temperature inspection based on the separated target thermal radiation.

[0035] The target thermal radiation refers to the infrared radiation generated by the equipment's own heating, which needs to be obtained by removing the energy component corresponding to the environmental reflected radiation from the time-frequency matrix. After separating the target thermal radiation, this application can further combine the texture parameters of the equipment surface to correct the final detected temperature value, thereby performing equipment temperature inspection based on the corrected temperature value.

[0036] In some embodiments, the inspection unit 103 is used to determine the corrected emissivity parameter based on the roughness of the surface of the device corresponding to the target thermal radiation, and to determine the temperature value of the surface of the device corresponding to the target thermal radiation and the corrected emissivity parameter using a radiation formula.

[0037] The corrected emissivity parameter is positively correlated with roughness. Temperature is positively correlated with target thermal radiation and negatively correlated with the corrected emissivity parameter. Emissivity refers to an object's ability to emit infrared radiation. The emissivity of a metal device surface changes with roughness; the lower the roughness (the smoother the surface), the lower the emissivity. In this embodiment, the corrected emissivity parameter refers to the emissivity dynamically adjusted based on the device surface roughness, which can be determined by calculating the roughness from visible light images in the control room.

[0038] For example, the corrected emissivity parameter satisfies the following formula: in, To correct the emissivity parameters. The baseline emissivity parameter is determined by the material of the equipment and ranges from 0 to 1. This is the roughness weighting factor, used to adjust the weighting effect of roughness. For roughness.

[0039] For example, the radiation formula can be the Stefan-Boltzmann formula, and the temperature value satisfies the following formula: in, This is a temperature value, and the unit is Kelvin (K). Thermal radiation is measured in watts per square meter, representing the total energy radiated per square meter per second from the surface of an object. To correct the emissivity parameters. The Stefan-Boltzmann constant has a value of [value missing]. .

[0040] The inspection unit 103 is also used to perform equipment temperature inspection based on the temperature value of the equipment surface.

[0041] In some embodiments, the inspection unit 103 may trigger a temperature warning if the temperature value on the surface of the equipment exceeds the rated temperature threshold.

[0042] For example, the above processing operations of the inspection unit 103 can be accelerated by a field-programmable gate array (FPGA) chip.

[0043] Based on the above technical solution, this application embodiment can acquire infrared image data of control room equipment through data acquisition unit 101 to obtain infrared time-domain signals collected based on infrared radiation reflection from the surface of the control room equipment. Then, signal processing unit 102 performs adaptive time-frequency transformation on the infrared time-domain signals in the infrared image data, converting the time-domain signals into a multi-dimensional time-frequency matrix. By analyzing the energy distribution of infrared radiation at different times and frequencies, the reflected light and thermal radiation signals are separated, allowing inspection unit 103 to perform equipment temperature inspection based on the separated target thermal radiation. Compared with current temperature detection solutions, this application can avoid environmental reflection interference during temperature inspection, thereby further improving temperature detection accuracy and inspection efficiency. It can be widely used in control room inspection scenarios in industries such as power, petrochemicals, and energy.

[0044] As one possible embodiment of this application, combined with Figure 1 ,like Figure 2 As shown, the autonomous inspection device 10 for the control room also includes a preprocessing unit 104. The preprocessing unit 104 is used to perform preprocessing operations on the data collected by the data acquisition unit 101.

[0045] The preprocessing operations include: Non-uniformity correction, median filtering, and histogram equalization are performed on infrared image data; Outlier removal from spatial location data; Spatiotemporal registration is performed on infrared image data and spatial location data.

[0046] For example, this application can sequentially perform non-uniformity correction on long-wave / short-wave infrared images to eliminate sensor noise, use median filtering with a 3×3 window to remove salt-and-pepper noise, and enhance details in low-contrast areas through histogram equalization. For spatial location data acquired by LiDAR, a density-based spatial clustering of applications with noise (DBSCAN) algorithm (neighborhood radius 0.1m, minimum number of points 5) is used to remove outliers and exclude abnormal data. In one example, this application can output a list of reflector coordinates, three-dimensional coordinates, distance, and data based on short-wave infrared intensity. Preliminary temperature estimate calculated This provides a reliable input for subsequent signal analysis.

[0047] It should be noted that, to address the issue of spatial inconsistency in multi-source data, this application employs spatiotemporal registration technology to spatially align the spatial location data acquired by the lidar with the infrared image data collected by the dual-band infrared-polarization composite sensor, thereby enabling cross-modal data fusion analysis. For example, this application can establish a transformation relationship from three-dimensional coordinates in the spatial location data to two-dimensional coordinates in the infrared image data by jointly calibrating the internal and external parameters of the dual-band infrared sensor and the lidar. This transformation relationship can map the three-dimensional position of each reflection source in the spatial location to the corresponding pixel position in the infrared image, eliminating the reference differences between the coordinate systems of different sensors. Furthermore, for each pixel in the infrared image, this application can search for the point in the spatial location closest to the projected position of that pixel (e.g., setting a distance threshold ≤ 5cm to ensure matching effectiveness), obtaining the spatial distance and reflection source identifier (if present) corresponding to that pixel, thereby adding a corresponding spatial location attribute to each infrared pixel and generating a tagged infrared image containing spectral information and spatial coordinates.

[0048] Based on the above technical solution, this application introduces a preprocessing unit 104 to provide high-quality raw data for subsequent signal processing and inspection. Through operations such as non-uniformity correction and median filtering, problems such as sensor noise and data misalignment are eliminated, avoiding subsequent processing errors caused by poor raw data quality. The spatiotemporal matching criterion realizes the fusion of multi-source data, ensuring the correspondence between infrared signals and spatial positions, laying the foundation for the application of cross-modal features (such as infrared signals and distance), significantly improving the reliability of raw data, and further ensuring the accuracy of the entire inspection process.

[0049] As one possible embodiment of this application, to further improve the time-frequency transformation accuracy of the signal processing unit 102 for infrared time-domain signals, this embodiment can dynamically adjust and correct the window function width according to the signal characteristics. Specifically, the signal processing unit 102 is used to: dynamically adjust the window function width of the adaptive time-frequency transformation according to the frequency and duration of the infrared time-domain signal, and correct the dynamically adjusted window function width according to the distance to the reflection source of the infrared time-domain signal and the roughness of the corresponding device surface. Then, based on the corrected window function width, it performs time-frequency transformation on the infrared time-domain signal to obtain the corresponding time-frequency matrix.

[0050] In this context, the window width refers to the length of the window used to extract the time-domain signal in time-frequency transformation. Wide window functions are suitable for analyzing low-frequency signals (improving frequency resolution), while narrow window functions are suitable for analyzing high-frequency signals (improving time resolution). Signal frequency refers to the frequency at which the signal intensity changes periodically (e.g., the frequency of ambient reflected light signals is typically higher than 10Hz, while the frequency of equipment heating signals is typically lower than 10Hz). Signal duration refers to the length of time the signal intensity exceeds the background noise threshold (e.g., the duration of reflected light signals is typically less than 0.2 seconds, while the duration of heating signals is typically longer than 0.2 seconds).

[0051] It should be noted that traditional Fourier transform algorithms typically use a fixed window length for analysis. This makes it difficult to balance time-frequency resolution, leading to distorted feature extraction and difficulty in accurately distinguishing reflected light from heat signals. Furthermore, the S-transform is a time-frequency transform scheme that combines short-time Fourier transform and continuous wavelet transform. It uses a Gaussian window function to achieve an inverse relationship between window width and frequency, thus resolving the time-frequency resolution conflict caused by the fixed window length of the traditional Fourier transform.

[0052] However, the window function width in this S-transform is driven solely by mathematical means (the window width is inversely proportional to the frequency), which is not suitable for temperature detection scenarios involving equipment temperature monitoring. For example, this scheme has significant drawbacks when dealing with the high-frequency transient fluctuations unique to metallic reflections: for reflected light signals with frequencies higher than 10Hz and durations shorter than 0.2 seconds, an excessively narrow window width can easily lead to insufficient periods within the window, causing energy leakage. When processing low-frequency heating signals, the default window width may cause spectral leakage due to period truncation, and since the duration constraint of the reflected light is not considered, misjudgments are easily made when the signal duration is abnormal.

[0053] Therefore, in the above embodiments of this application, the window function width can be dynamically adjusted by combining the frequency and duration of the infrared time-domain signal, thereby achieving effective extraction of different signals. For example, this application can achieve discretized window function width adjustment by setting frequency and duration thresholds.

[0054] For example, regarding frequency thresholds, using 10Hz as the dividing point, when the signal frequency is greater than 10Hz, this application can enable a narrow window (e.g., a window function width of 0.1 seconds) to ensure that the 10Hz signal contains one complete cycle, thereby accurately capturing the instantaneous characteristics of high-frequency reflected light; when the frequency is less than or equal to 10Hz, this application can switch to a wide window (e.g., a window function width of 1 second) to ensure that the 5Hz signal contains two complete cycles, effectively reducing the spectral leakage rate of low-frequency heating signals. This frequency threshold setting is based on the physical characteristics of metal equipment heating and environmental reflection, i.e., the equipment temperature rise typically changes slowly on a minute-by-minute basis (corresponding to a frequency <10Hz), while specular reflected light fluctuates on a millisecond-by-millisecond basis with changes in the light source angle (frequency >10Hz).

[0055] In addition, the reflected light from the metal surface may be affected by factors such as equipment vibration and light source movement, resulting in abnormal situations where the duration is less than 0.2 seconds but the frequency is less than 10Hz (such as an 8Hz signal lasting 0.15 seconds). If a wide window is used directly, a large amount of environmental noise may be introduced, leading to an increase in the misjudgment rate of the heating signal. Therefore, based on this, this application can also adjust the duration threshold.

[0056] For example, the duration threshold can be set to 0.2 seconds. Firstly, this application can calculate the duration of the signal by measuring the time interval between when the time-domain signal first exceeds the signal strength threshold and when it falls below that threshold. For example, the signal strength threshold can be set to twice the average amplitude of the background noise. Taking an average background noise of 10mV as an example, the corresponding signal strength threshold would be set to 20mV. This effectively filters out random interference with amplitudes not exceeding twice the noise average while ensuring the capture of the true time-domain start and end characteristics of the signal, thus distinguishing the time-domain characteristics of reflected light and heating signals through the time interval.

[0057] Subsequently, when the signal duration is less than 0.2 seconds, regardless of whether the frequency is higher than 10Hz, this application can adjust the window function width using a window width adjustment algorithm. For example, the window function width can be adjusted to 1.2 times the previous value (e.g., the window function width is adjusted from 0.15 seconds to 0.18 seconds) to ensure that the window contains at least one complete cycle while suppressing noise interference. When the signal duration is greater than or equal to 0.2 seconds, this application can perform time-frequency conversion of the signal according to the previously configured window function width to ensure the accuracy of time-frequency analysis of long-term signals.

[0058] It should be noted that the distance between the metal surface of the equipment in the control room and the autonomous inspection device 10 in the control room, as well as the surface roughness, will significantly affect the intensity of reflected light. For example, the reflected light energy of a smooth metal surface (roughness less than 0.3) at close range (less than 1m) is strong and the frequency fluctuates greatly. In this case, using a fixed window width can easily lead to feature saturation distortion. Therefore, this application can further correct the dynamically adjusted window function width according to the distance of the infrared time domain signal to the reflection source and the roughness of the corresponding equipment surface.

[0059] In some embodiments, the signal processing unit 102 is specifically configured to: determine a distance influence factor based on the distance to the reflection source, determine a roughness influence factor based on the roughness of the device surface, and correct the window function width according to the distance influence factor and the roughness influence factor.

[0060] Among them, the distance influence factor is negatively correlated with the distance to the reflection source, and the roughness influence factor is negatively correlated with the roughness of the equipment surface.

[0061] Regarding the distance influence factor, the closer the distance, the greater the intensity of the reflected signal, and the larger the distance influence factor becomes. The intensity of reflected light is directly related to the distance. The square of the value is inversely proportional to the value of the distance. To construct a parameter system with unified dimensions, this application may introduce a reference distance benchmark value. After standardizing the distance influence factor to 1m, the distance influence factor can be expressed as: This ensures that the distance influence factor matches the dimensionless roughness parameter R in numerical dimension.

[0062] Regarding the roughness influencing factor, roughness The smaller the value (the smoother the surface), the higher the reflectivity and the larger the roughness influence factor. The roughness influence factor can be set to... To characterize a smooth surface, the window width compression force needs to be enhanced.

[0063] For example, roughness can be calculated from visible light images. For instance, by obtaining the gray-level co-occurrence matrices in four directions (0°, 45°, 90°, and 135°) within the control room, extracting the contrast features of each gray-level co-occurrence matrix, taking the average value, and then normalizing it to the 0-1 interval through min-max normalization, the roughness at each location can be obtained.

[0064] In some embodiments, this application may determine correction coefficients based on distance influence factors and roughness influence factors, and correct the window function width based on the correction coefficients.

[0065] For example, the correction factor satisfies the following formula: in, For correction factor, Indicates the distance influence factor. This represents the roughness influence factor. The constant term 1 is the basic window width coefficient, ensuring the stability of the adjustment reference. The distance to the reflection source, measured by lidar, reflects the degree of attenuation of the reflected light energy; the closer the distance, the higher the distance influence factor. This represents the roughness of the metal surface; the smaller the value, the smoother the surface and the higher the reflectivity.

[0066] The aforementioned correction coefficients can be mapped one-to-one to each reflection source, meaning the window function width for the time-domain signal corresponding to each reflection source can be adjusted according to the corresponding correction coefficients. For example, the window function width satisfies the following formula: in, For the first The window function width of the time-domain signal corresponding to each reflection source. The window function width before correction can be determined using the scheme provided in the above embodiments of this application. For the first Correction coefficients corresponding to each reflection source.

[0067] Based on the above technical solution, this application can adjust the window width according to the signal frequency and duration to ensure that signals with different frequency characteristics can obtain suitable time-frequency resolution (narrow window for high-frequency signals and wide window for low-frequency signals), which solves the problem that traditional fixed window width cannot take into account both high and low frequency signals. Secondly, the window width is corrected by combining the distance of the reflection source and the surface roughness of the equipment, taking into account the influence of environmental factors on the reflected signal (such as the more violent fluctuation of the reflected signal from a smooth surface at close range), which further improves the matching degree between the window width and the signal characteristics. Finally, the time-frequency matrix output by the time-frequency transformation based on the corrected window width has significantly improved accuracy of energy distribution, providing higher quality input data for the subsequent inspection unit 103 to separate the target thermal radiation.

[0068] It should be noted that traditional temperature detection schemes struggle to quantify the frequency distribution complexity of infrared signals, locate areas of concentrated energy, and capture differences in phase stability, resulting in ambiguity in distinguishing between multi-frequency mixed signals and single-frequency signals. Therefore, this application utilizes the obtained time-frequency matrix for multi-dimensional feature extraction, thereby further improving the accuracy of the inspection unit 103 in separating target thermal radiation.

[0069] As one possible embodiment of this application, time-frequency feature vectors are extracted from the time-frequency matrix, and target thermal radiation is separated based on the time-frequency feature vectors.

[0070] The time-frequency feature vector includes at least one of the following: frequency entropy, time-frequency energy moment, and phase consistency index.

[0071] Frequency entropy is used to characterize the uniformity of signal energy distribution along the frequency dimension. This application can obtain a normalized parameter characterizing the proportion of energy at each frequency by normalizing the energy of each frequency component in the time-frequency matrix. For reflected light, due to environmental interference, it exhibits multi-frequency mixing, resulting in a relatively uniform distribution of its normalized energy across frequency components. The uniformity of this distribution is positively correlated with the entropy value, thus leading to a high information entropy value for reflected light. For heating signals, the frequency is typically more singular, with energy concentrated in the low-frequency band, exhibiting a highly concentrated energy distribution and correspondingly lower information entropy values. This application can convert the uniformity of frequency energy distribution into information entropy values, thereby achieving a quantitative distinction between the multi-frequency mixing characteristics of reflected light (e.g., entropy > 3) and the single-frequency characteristics of heating signals (e.g., entropy < 2), effectively solving the problem of quantitatively distinguishing frequency complexity.

[0072] For example, the frequency entropy satisfies the following formula: in, Frequency components The frequency entropy, The number of frequency components in the time-frequency matrix. Frequency components The normalized parameter of frequency energy.

[0073] The time-frequency energy moment is used to characterize the energy centroid of a signal in the time-frequency domain. Since traditional schemes typically only assess the energy characteristics of a signal based on its amplitude, they cannot locate the energy distribution in the time-frequency domain. Therefore, this application can utilize the time-frequency energy density... As a weighting factor, a two-dimensional weighted average operation is performed on the time and frequency dimensions to accurately calculate the centroid of the signal energy distribution. For example, the energy of a high-frequency reflective signal is usually concentrated in the high-frequency region of the time-frequency domain, and the value of the time-frequency energy moment after weighted calculation is relatively large (e.g., greater than 5); while the energy of a low-frequency heating signal is usually concentrated in the low-frequency band, and the corresponding value of the time-frequency energy moment is relatively small (e.g., less than 3).

[0074] For example, the time-frequency energy moment satisfies the following formula: in, Time component Frequency components The time-frequency energy moment, This represents the number of time components in the time-frequency matrix. The number of frequency components in the time-frequency matrix. Time component Frequency components The time-frequency energy density is determined based on the time-frequency matrix (i.e., the square of the modulus of energy in the time-frequency matrix), reflecting the signal energy intensity at that time-frequency point. The weighted energy sum in the time-frequency dimension represents the weighted center of gravity of the energy distribution in the time-frequency domain. This represents the total energy in the entire time-frequency domain, used for normalization.

[0075] Phase consistency indices are used to characterize the degree of phase fluctuation in the time-frequency domain. Since reflected light and heat signals exhibit fundamental differences in phase stability, this application constructs a phase consistency index based on the physical characteristics of these two types of signals. Heat signals are affected by gradual temperature changes, resulting in smaller phase fluctuations; while reflected signals experience abrupt phase changes due to frequent changes in the reflection path. This application can quantitatively characterize phase stability by calculating the standard deviation of the phase angle of the time-frequency matrix and normalizing it. Specifically, the stable phase of a heat signal is typically converted into a higher consistency value (e.g., greater than 0.8), while the unstable phase of a reflected signal is typically converted into a lower value (e.g., less than 0.5).

[0076] For example, the phase consistency index satisfies the following formula: in, This is an indicator of the phase consistency of the signal. The phase angle of the time-frequency matrix. The standard deviation of the phase angle quantifies the degree of phase fluctuation in the time-frequency domain.

[0077] It should be noted that, regarding frequency entropy, reflected light exhibits multi-frequency mixing characteristics due to environmental interference, therefore the corresponding... The frequency entropy is relatively uniformly distributed across all frequency bands. Typically greater than 3; the heating signal has a single frequency, with energy concentrated in the low-frequency range. The frequency entropy obtained from the concentrated distribution Typically less than 2. This application can convert the uniformity of frequency distribution into a comparable value through frequency entropy, thus solving the problem of insufficient ability of traditional spectrum analysis to distinguish complex frequency components.

[0078] For the time-frequency energy moment, this application locates the energy centroid using a weighted average method. The energy of high-frequency reflected signals is typically concentrated in the high-frequency band, resulting in the time-frequency energy moment... Typically above 5, the energy of low-frequency heating signals is usually concentrated in the low-frequency band, resulting in the time-frequency energy moment. It is usually below 3.

[0079] For phase consistency indices, the heating signal is affected by gradual temperature changes, resulting in small phase fluctuations, thus yielding a phase consistency index. Typically greater than 0.8; the phase consistency index obtained is due to the abrupt phase change in the reflected signal caused by the change in the reflection path. It is usually less than 0.5.

[0080] Thus, based on the above three types of characteristic parameters, this application can more comprehensively analyze the characteristic differences between reflected light and heating signal, thereby achieving effective separation of the thermal radiation of heating signal.

[0081] In some embodiments, this application can classify the heating signals based on the above-mentioned feature parameters using a classification model, thereby processing the heating signals. The inspection unit 103 is specifically used to: input the time-frequency feature vector into a preset classification model, determine the signal category of the infrared time-domain signal, and separate the infrared time-domain signal with the signal category of heating signal from the infrared image data as the target thermal radiation.

[0082] The signal categories include heat signals and reflected light signals. The preset classification model uses a random forest classifier, which is trained using historical sample data.

[0083] For example, historical samples include time-frequency feature vectors labeled "heating signal" or "reflected light signal," with reflected light and heating signal marked as "0" and "1" respectively as category labels. The preset classification model consists of a random forest classifier containing 500 decision trees. During training, 80% of the samples are sampled using a bootstrap sampling method to train a single decision tree, and the remaining 20% ​​of the samples are used to evaluate the model performance. Each tree selects the optimal splitting feature based on the Gini coefficient and grows until it terminates naturally. Finally, the outputs of the decision trees are integrated through a majority voting mechanism to determine the signal category ("0" or "1").

[0084] Based on the above technical solution, the random forest classifier in this embodiment of the application has strong anti-interference ability and generalization ability through the ensemble learning of multiple decision trees. It can handle signal classification tasks in different control room environments, avoid the problem that traditional threshold judgment is easily affected by environmental changes, significantly improve the separation accuracy of target thermal radiation, and provide key guarantee for the accuracy of subsequent temperature calculation.

[0085] Please see Figure 3 The diagram illustrates a flowchart of a control room autonomous inspection method according to an embodiment of the present invention, which includes the following steps: Step 301: Collect infrared image data of the equipment in the control room.

[0086] Infrared image data is used to characterize infrared time-domain signals collected based on infrared radiation reflection from the surface of equipment in the control room.

[0087] Step 302: Perform adaptive time-frequency transformation based on the infrared time-domain signal in the infrared image data to obtain the corresponding time-frequency matrix.

[0088] The time-frequency matrix is ​​used to characterize the energy distribution of infrared radiation on the surface of indoor equipment at different times and frequencies.

[0089] Step 303: Separate the target thermal radiation based on the time-frequency matrix, and perform equipment temperature inspection based on the separated target thermal radiation.

[0090] For related explanations, please refer to the above embodiments, which will not be repeated here.

[0091] It should be noted that the order of the above embodiments of the present invention is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. The processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0092] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

Claims

1. A control room autonomous inspection device, characterized in that, include: The data acquisition unit is used to acquire infrared image data of the control room equipment; The infrared image data is used to characterize the infrared time-domain signal collected based on the infrared radiation reflection from the surface of the equipment in the control room; The signal processing unit is used to perform adaptive time-frequency transformation based on the infrared time-domain signal in the infrared image data to obtain the corresponding time-frequency matrix; the time-frequency matrix is ​​used to characterize the energy distribution of infrared radiation on the surface of indoor equipment at different times and frequencies. The inspection unit is used to separate the target thermal radiation based on the time-frequency matrix and to perform equipment temperature inspection based on the separated target thermal radiation.

2. The autonomous inspection device for the control room according to claim 1, characterized in that, The signal processing unit is specifically used for: The window function width of the adaptive time-frequency transform is dynamically adjusted based on the frequency and duration of the infrared time-domain signal. The width of the dynamically adjusted window function is corrected based on the distance to the reflection source of the infrared time-domain signal and the roughness of the corresponding device surface; The infrared time-domain signal is subjected to time-frequency transformation based on the corrected window function width to obtain the corresponding time-frequency matrix.

3. The autonomous inspection device for the control room according to claim 2, characterized in that, The signal processing unit is specifically used for: A distance influence factor is determined based on the distance to the reflection source, and a roughness influence factor is determined based on the roughness of the device surface; wherein, the distance influence factor is negatively correlated with the distance to the reflection source; and the roughness influence factor is negatively correlated with the roughness of the device surface. The window function width is corrected based on the distance influence factor and the roughness influence factor.

4. The autonomous inspection device for the control room according to claim 1, characterized in that, The inspection unit is specifically used for: Extract time-frequency feature vectors from the time-frequency matrix; the time-frequency feature vectors include at least one of frequency entropy, time-frequency energy moment, and phase consistency index; the frequency entropy is used to characterize the uniformity of signal energy distribution in the frequency dimension; The time-frequency energy moment is used to characterize the energy centroid of the signal energy in the time-frequency domain; the phase consistency index is used to characterize the degree of fluctuation of the signal phase in the time-frequency domain. The target thermal radiation is separated based on the time-frequency feature vector.

5. The autonomous inspection device for the control room according to claim 4, characterized in that, The inspection unit is specifically used for: The time-frequency feature vector is input into a preset classification model to determine the signal category of the infrared time-domain signal; the signal category includes heating signals and reflected light signals; the preset classification model uses a random forest classifier, which is trained using historical sample data. The infrared time-domain signal classified as a heat signal is separated from the infrared image data and used as the target thermal radiation.

6. The autonomous inspection device for the control room according to claim 1, characterized in that, The data acquisition unit includes a dual-band infrared-polarization composite sensor and a lidar; The dual-band infrared-polarization composite sensor is used to adjust the polarizer direction in real time and obtain infrared time-domain signals through dual-band collaborative acquisition to obtain infrared image data; The lidar is used to acquire spatial location data of indoor control equipment; the spatial location data includes three-dimensional point cloud data of the indoor control equipment, which is used to establish a mapping relationship between pixels in the infrared image data and spatial locations of the indoor equipment.

7. The autonomous inspection device for the control room according to claim 6, characterized in that, The device further includes a preprocessing unit for preprocessing the data acquired by the data acquisition unit. The preprocessing operation includes: The infrared image data is subjected to non-uniformity correction, median filtering, and histogram equalization. Outlier removal is performed on the spatial location data; Spatiotemporal registration is performed on the infrared image data and the spatial location data.

8. The autonomous inspection device for the control room according to claim 1, characterized in that, The inspection unit is specifically used for: The corrected emissivity parameter is determined based on the roughness of the device surface corresponding to the target thermal radiation, and the temperature value of the corresponding device surface is determined based on the target thermal radiation and the corrected emissivity parameter using the radiation formula. The corrected emissivity parameter is positively correlated with the roughness; The temperature value is positively correlated with the target thermal radiation and negatively correlated with the corrected emissivity parameter; Conduct equipment temperature inspections based on the temperature values ​​on the equipment surface.

9. The autonomous inspection device for the control room according to claim 8, characterized in that, The inspection unit is specifically used for: A temperature warning is triggered when the temperature on the surface of the device exceeds the rated temperature threshold.

10. A method for autonomous inspection of a control room, characterized in that, include: Collect infrared image data from equipment in the control room; The infrared image data is used to characterize the infrared time-domain signal collected based on the infrared radiation reflection from the surface of the equipment in the control room; An adaptive time-frequency transformation is performed on the infrared time-domain signal in the infrared image data to obtain the corresponding time-frequency matrix; the time-frequency matrix is ​​used to characterize the energy distribution of infrared radiation on the surface of indoor equipment at different times and frequencies; Based on the time-frequency matrix, the target thermal radiation is separated, and the equipment temperature is inspected according to the separated target thermal radiation.

Images