Temperature early-warning method and apparatus for power station cabinet device, and computer device

Abstract

The present application relates to a temperature early-warning method and apparatus for a power station cabinet device, and a computer device, a computer-readable storage medium and a computer program product. The method comprises: according to a collection parameter corresponding to a current collection, collecting and acquiring a dual-light image of a power station cabinet device to be subjected to analysis, wherein the dual-light image comprises an infrared thermal image and a visible-light image; performing pre-processing, feature extraction processing, spatial registration processing and information fusion processing on the dual-light image, so as to generate a fused image; performing temperature field reconstruction processing and hotspot detection processing on the fused image, so as to obtain temperature field distribution information and hotspot area information; and on the basis of the temperature field distribution information and the hotspot area information, performing temperature analysis on the power station cabinet device, so as to obtain temperature early-warning information of the power station cabinet device.

Description

Temperature early warning methods, devices, and computer equipment for power plant cabinet equipment

[0001] Cross-reference to related applications

[0002] This application claims priority to Chinese patent application filed on December 9, 2024, application number 2024117955297, entitled "Temperature Early Warning Method, Device and Computer Equipment for Power Station Panel Equipment", the entire contents of which are incorporated herein by reference. Technical Field

[0003] This application relates to the field of computer technology, and in particular to a method, device, computer equipment, computer-readable storage medium, and computer program product for temperature early warning of power station cabinet equipment. Background Technology

[0004] Power plant cabinet equipment is a critical component of power systems, primarily used for control, protection, and monitoring. These devices generate significant heat during operation; excessively high temperatures can lead to equipment malfunctions, insulation aging, and even fires. Therefore, real-time temperature monitoring and early warning systems for power plant cabinet equipment are of paramount practical importance and are crucial for ensuring the safe and stable operation of power systems.

[0005] Traditional temperature monitoring methods often rely on a single temperature sensor or manual inspection. These methods suffer from problems such as slow response, insufficient detection accuracy, and difficulty in providing comprehensive coverage. In particular, during the operation of power plant cabinet equipment, local overheating or malfunctions may occur. Traditional monitoring methods often fail to effectively identify potential temperature anomalies, resulting in poor accuracy in temperature warnings for power plant cabinet equipment. Summary of the Invention

[0006] According to various embodiments of this application, a method, apparatus, computer equipment, computer-readable storage medium, and computer program product for temperature early warning of power plant cabinet equipment are provided.

[0007] Firstly, this application provides a method for early warning of temperature in power plant cabinet equipment, including:

[0008] If this is the first data collection, the data collection parameters corresponding to the lowest temperature anomaly level will be used as the data collection parameters for this data collection; if this is not the first data collection, the data collection parameters corresponding to the temperature anomaly level determined in the previous data collection will be used as the data collection parameters for this data collection.

[0009] According to the acquisition parameters corresponding to this acquisition, dual-light images of the power station cabinet equipment to be analyzed are acquired; the dual-light images include infrared thermal imaging images and visible light images;

[0010] The two-light image is preprocessed to obtain a preprocessed two-light image; the preprocessing includes at least quality assessment processing, image enhancement processing, and region segmentation processing; multi-scale local features and regional features are extracted from the preprocessed two-light image to construct feature description information; based on the region segmentation result of the two-light image, initial registration points are selected from the feature points corresponding to the feature description information; the region segmentation result is obtained through the region segmentation processing; based on the feature description information, candidate registration points are selected from the initial registration points; a random sampling consistency model is used to remove erroneous registration points from the candidate registration points to obtain the target registration point; based on the... The target registration points are defined, and the affine transformation matrix of the infrared thermal imaging image relative to the visible light image is calculated and determined. Based on the affine transformation matrix and the infrared thermal imaging image, a registered infrared thermal imaging image is obtained. Using edge information from the visible light image, an edge map of the visible light image is obtained, and a temperature gradient map of the registered infrared thermal imaging image is calculated. Based on the edge map and the temperature gradient map, the fusion weights of the two images are determined. Based on the fusion weights, the registered infrared thermal imaging image and the visible light image are fused to obtain a fused image.

[0011] The fused image is subjected to temperature field reconstruction processing and hotspot detection processing to obtain temperature field distribution information and hotspot region information; the temperature field distribution information is determined based on the initial temperature field distribution information, the temperature gradient map of the registered infrared thermal imaging image, the fused image, the preset reference temperature, and the preset importance weight of the power station cabinet equipment area; the initial temperature field information is obtained based on the registered infrared thermal imaging image.

[0012] Based on the temperature field distribution information and the hot spot area information, temperature analysis is performed on the power station cabinet equipment to obtain temperature early warning information for the power station cabinet equipment.

[0013] Secondly, this application also provides a temperature early warning device for power plant cabinet equipment, comprising:

[0014] The data acquisition parameter determination module is used to determine the acquisition parameters of the lowest temperature anomaly level as the acquisition parameters of the current acquisition when the current acquisition is the first acquisition; and to determine the acquisition parameters of the temperature anomaly level determined in the previous acquisition as the acquisition parameters of the current acquisition when the current acquisition is not the first acquisition.

[0015] The image acquisition module is used to acquire dual-light images of the power station cabinet equipment to be analyzed according to the acquisition parameters corresponding to this acquisition; the dual-light images include infrared thermal imaging images and visible light images.

[0016] An image fusion module is used to preprocess the two-light image to obtain a preprocessed two-light image. The preprocessing includes at least quality assessment processing, image enhancement processing, and region segmentation processing. Multi-scale local features and regional features are extracted from the preprocessed two-light image to construct feature description information. Based on the region segmentation result of the two-light image, initial registration points are selected from the feature points corresponding to the feature description information. The region segmentation result is obtained through the region segmentation processing. Based on the feature description information, candidate registration points are selected from the initial registration points. A random sampling consistency model is used to eliminate erroneous registration points among the candidate registration points to obtain the target registration. Based on the target registration points, the affine transformation matrix of the infrared thermal imaging image relative to the visible light image is calculated and determined; based on the affine transformation matrix and the infrared thermal imaging image, a registered infrared thermal imaging image is obtained; using the edge information in the visible light image, an edge map of the visible light image is obtained, and a temperature gradient map of the registered infrared thermal imaging image is obtained by calculating the temperature gradient of the registered infrared thermal imaging image; based on the edge map and the temperature gradient map, the fusion weight of the two-light image is determined; based on the fusion weight, the registered infrared thermal imaging image and the visible light image are fused to obtain a fused image;

[0017] The temperature detection module is used to perform temperature field reconstruction processing and hotspot detection processing on the fused image to obtain temperature field distribution information and hotspot region information. The temperature field distribution information is determined based on the initial temperature field distribution information, the temperature gradient map of the registered infrared thermal imaging image, the fused image, the preset reference temperature, and the preset importance weight of the power station cabinet equipment area. The initial temperature field distribution information is obtained based on the registered infrared thermal imaging image.

[0018] The temperature analysis module is used to perform temperature analysis on the power station cabinet equipment based on the temperature field distribution information and the hot spot area information, and to obtain temperature warning information for the power station cabinet equipment.

[0019] Thirdly, this application also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the power station cabinet equipment temperature early warning method described in any embodiment of this application.

[0020] Fourthly, this application also provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the power station cabinet equipment temperature early warning method described in any embodiment of this application.

[0021] Fifthly, this application also provides a computer program product, including a computer program that, when executed by a processor, implements the power station cabinet equipment temperature early warning method described in any embodiment of this application.

[0022] Details of one or more embodiments of this application are set forth in the following drawings and description. Other features, objects, and advantages of this application will become apparent from the specification, drawings, and claims. Attached Figure Description

[0023] To more clearly illustrate the technical solutions in the embodiments of this application or the conventional technology, the drawings used in the description of the embodiments or the conventional technology will be briefly introduced below. Obviously, the drawings described below are only embodiments of this application. For those skilled in the art, other drawings can be obtained based on the disclosed drawings without creative effort. The additional details or examples used to describe the drawings should not be considered as a limitation on the scope of any of the disclosed invention, the currently described embodiments and / or examples, and the best mode of these inventions as currently understood.

[0024] Figure 1 is a flowchart illustrating a power plant cabinet equipment temperature early warning method provided according to one or more embodiments.

[0025] Figure 2 is a flowchart illustrating the steps for generating a fused image according to one or more embodiments.

[0026] Figure 3 is a schematic diagram of a power plant cabinet equipment temperature early warning system provided according to one or more embodiments.

[0027] Figure 4 is a structural block diagram of a power plant cabinet equipment temperature early warning device provided according to one or more embodiments.

[0028] Figure 5 is an internal structure diagram of a computer device provided according to one or more embodiments. Detailed Implementation

[0029] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0030] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0031] In one embodiment, as shown in Figure 1, a method for temperature early warning of power plant cabinet equipment is provided. This embodiment illustrates the application of this method to a terminal. It is understood that this method can also be applied to a server, or to a system including both a terminal and a server, and is implemented through interaction between the terminal and the server. The terminal can be, but is not limited to, various personal computers, laptops, smartphones, and tablets. The server can be a standalone physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server providing cloud computing services. In this embodiment, the method includes the following steps:

[0032] Operate S101 to acquire dual-light images of the power station cabinet equipment to be analyzed, according to the acquisition parameters corresponding to this acquisition.

[0033] The dual-light image includes an infrared thermal imaging image and a visible light image.

[0034] Among them, the acquisition parameters refer to the camera parameters set by the terminal during the acquisition process, including the temperature range and emissivity of the infrared camera and the exposure time and white balance parameters of the visible light camera. These parameters are adjusted and configured according to the operating environment and monitoring needs of the power station cabinet equipment.

[0035] For example, the terminal simultaneously activates an infrared thermal imaging camera and a visible light camera via its built-in dual-light camera module, imaging the target device according to pre-defined acquisition parameters. The terminal adjusts the acquisition parameters based on the device's real-time operating status; for instance, it dynamically adjusts the emissivity of the infrared thermal imaging camera based on the highest temperature range of the device's surface to ensure accurate temperature measurement. Simultaneously, the terminal adjusts the exposure time and white balance parameters of the visible light camera in real-time based on ambient lighting conditions to obtain a clear visible light image. Finally, the terminal completes the synchronous acquisition of the dual-light images and transmits them to the subsequent processing module.

[0036] Operation S102 performs preprocessing, feature extraction, spatial registration, and information fusion on the dual-light image to generate a fused image.

[0037] Preprocessing involves performing operations such as noise removal, contrast enhancement, and brightness adjustment on the acquired dual-light images to improve image quality and provide clear and stable input data for subsequent analysis. Feature extraction extracts key local and regional features from the images, including texture, shape, and edge information. These features help accurately identify various components of the device and potential problem areas. Spatial registration aligns the infrared thermal imaging image and the visible light image, ensuring precise matching of their temperature and structural information to avoid image discrepancies. Information fusion combines the registered images, integrating the temperature data from the infrared image with the structural information from the visible light image to generate a final fused image, providing a precise data foundation for subsequent temperature analysis and hotspot detection.

[0038] For example, the terminal first preprocesses the acquired dual-light images, using Gaussian filtering or median filtering to remove noise and histogram equalization to enhance image contrast. Next, image processing algorithms, such as Scale-Invariant Feature Transform (SIFT) or Speeded-Up Robust Features (SURF), are used to extract important feature points and region descriptors from the preprocessed images. These feature points and regions will serve as the basis for subsequent registration and matching. In the spatial registration stage, the terminal uses feature point matching algorithms, such as Random Sample Consensus (RANSAC), to accurately align the infrared thermal imaging image and the visible light image, ensuring accurate spatial correspondence. Finally, the terminal merges the registered images using a weighted fusion technique, generating a fused image that utilizes temperature information from the infrared image and structural information from the visible light image. This fused image has higher temperature detection accuracy and structural detail.

[0039] Operation S103 performs temperature field reconstruction and hotspot detection processing on the fused image to obtain temperature field distribution information and hotspot region information.

[0040] Temperature field reconstruction refers to reconstructing the temperature field of a device surface based on temperature data from a fused image, using interpolation algorithms and heat conduction models. The goal of temperature field reconstruction is to expand discrete temperature data points into a complete temperature distribution map, providing information on temperature change trends in different areas of the device surface. Hotspot detection identifies areas with abnormal temperatures—those exceeding a preset safe temperature range—by setting temperature thresholds or employing regional statistical analysis. Hotspot area information includes the spatial location, area, and temperature value of the hotspot; this information is crucial for equipment maintenance and fault early warning.

[0041] For example, the terminal smooths the temperature data points in the fused image by interpolating the temperature information (such as bilinear interpolation or kriging interpolation) to generate a continuous temperature field distribution map. Furthermore, the terminal utilizes temperature gradient analysis and local statistical analysis to identify areas of excessive temperature within the device, i.e., hotspot areas, by setting temperature thresholds or using standard deviation-based anomaly detection methods (such as the Z-score method). Through these algorithms, the terminal can accurately identify and mark potential fault areas within the device, providing crucial data for subsequent early warning systems.

[0042] Operate S104 to perform temperature analysis on the power station cabinet equipment based on temperature field distribution information and hot spot area information, and obtain temperature warning information for the power station cabinet equipment.

[0043] Temperature analysis refers to assessing the overall temperature status of power plant cabinet equipment based on temperature field distribution and hotspot area information. By analyzing the temperature values ​​of various areas on the equipment surface and the expansion trend of hotspots, it is possible to accurately determine whether there are potential temperature anomalies or malfunctions in the equipment. Temperature early warning information is generated based on the analysis results, including the severity of the current temperature anomaly, the location of the abnormal area, the temperature value, and the predicted temperature change trend, providing maintenance personnel with timely and accurate early warning data.

[0044] For example, the terminal first performs global and local analysis of the temperature field distribution information to assess the overall temperature status of the equipment and determine whether overheating exists on the equipment surface. Combining hotspot area information, the terminal identifies localized overheated areas and calculates whether their temperatures exceed preset safety thresholds (such as the equipment's maximum safe temperature). Furthermore, based on the equipment temperature change trend (by comparing currently collected temperature data with historical data) and the expansion of hotspot areas, the terminal predicts whether the equipment temperature may continue to rise or whether a malfunction may occur, and generates a temperature warning message. This warning message includes the temperature anomaly level (e.g., mild, moderate, severe anomaly), the spatial location and temperature data of the anomaly area, and provides operational suggestions (e.g., whether shutdown, cooling, or maintenance is required).

[0045] In the aforementioned temperature early warning method for power plant cabinet equipment, firstly, according to the acquisition parameters corresponding to this acquisition, dual-light images of the power plant cabinet equipment to be analyzed are acquired. These dual-light images include infrared thermal imaging images and visible light images. By combining the information from these two images, a comprehensive perception of the equipment status is achieved. The infrared image provides the temperature distribution on the equipment surface, helping to detect temperature anomalies, while the visible light image provides information on the equipment's appearance, structure, and details, aiding in fault location. This overcomes the limited monitoring range of a single sensor, ensuring full coverage monitoring of the equipment status. Furthermore, dynamically adjusting the acquisition parameters optimizes resource utilization, ensuring efficient acquisition in critical areas. Next, the dual-light images undergo preprocessing, feature extraction, spatial registration, and information fusion to generate a fused image. Preprocessing the acquired dual-light images improves image quality, reduces noise, and enhances contrast. Feature extraction obtains multi-scale local and regional features, ensuring the complete preservation of key information in the image. Spatial registration technology is used to further refine the infrared thermal imaging. Precise alignment of external and visible light images ensures data consistency. Information fusion is then used to synthesize the two images, generating a fused image that enhances the reliability and accuracy of the image data, laying a solid foundation for subsequent temperature field reconstruction and hotspot detection. Next, temperature field reconstruction and hotspot detection processing are performed on the fused image to obtain temperature field distribution and hotspot area information, improving the system's ability to detect temperature anomalies, especially the ability to identify potential risks such as localized overheating, providing crucial support for temperature early warning. Finally, based on the temperature field distribution and hotspot area information, temperature analysis is performed on the power plant cabinet equipment to obtain temperature early warning information. By analyzing the overall temperature field distribution and the characteristics of hotspot areas (such as temperature value, location, and size), the system accurately assesses the temperature anomaly level of the equipment and generates temperature early warning information. This warning information includes potential abnormal areas and corresponding temperature states, helping maintenance personnel to take timely measures and effectively reduce the risk of equipment failure. In the above method, spatial registration technology between infrared thermal imaging images and visible light images is used to achieve precise alignment of temperature data and equipment structural information; high-quality fused images are generated through image preprocessing and information fusion, providing a reliable foundation for temperature field reconstruction; hotspot detection accurately locates high-temperature areas and quickly identifies potential anomalies through threshold judgment and region segmentation; finally, based on temperature distribution and hotspot information, the temperature anomaly level is accurately assessed and early warning information is generated, realizing comprehensive monitoring and accurate early warning of equipment temperature status.

[0046] In an exemplary embodiment, before collecting the dual - light images of the power station panel equipment to be analyzed according to the acquisition parameters corresponding to the current acquisition in the above - mentioned operation S101, it further includes: in the case where the current acquisition is the first acquisition, taking the acquisition parameters corresponding to the lowest - level temperature anomaly level as the acquisition parameters for the current acquisition; in the case where the current acquisition is not the first acquisition, taking the acquisition parameters corresponding to the temperature anomaly level determined in the previous acquisition as the acquisition parameters for the current acquisition.

[0047] Among them, the temperature anomaly level is an index classified according to the severity of the equipment temperature state. The acquisition parameters include the setting parameters of the infrared camera and the visible - light camera, such as the temperature range, emissivity, exposure time, white balance, etc. of the infrared image. The setting of these parameters directly affects the quality of the collected images and the accuracy of the temperature data.

[0048] Exemplarily, the terminal first determines whether the current acquisition is the first acquisition. If it is the first acquisition, the temperature anomaly level is set to the lowest level (i.e., the "normal" or slightly abnormal level), and the corresponding acquisition parameters are set according to this level, such as a lower temperature range of the infrared thermal imaging camera and standard exposure settings, to ensure that the acquisition parameters are not overly sensitive. For subsequent acquisitions, the terminal will adjust the parameters of the current acquisition according to the temperature anomaly level determined in the previous acquisition. If the previous acquisition result shows that the temperature anomaly is moderate or severe, the terminal will adjust the acquisition parameters, such as increasing the temperature range, increasing the exposure time, etc., in order to capture more detailed temperature information, so as to provide higher - precision data support for subsequent temperature analysis and anomaly detection.

[0049] In one embodiment, the temperature anomaly level can be simply judged by the highest temperature, that is, if the highest temperature reaches the temperature range of a certain temperature anomaly level, it is that temperature anomaly level. When the temperature T < T1, within the normal temperature range, the reference frequency f0 is adopted; when T1 ≤ T < T2, slightly abnormal, 2 times the reference frequency is adopted; when T2 ≤ T < T3, moderately abnormal, 4 times the reference frequency is adopted; when T ≥ T3, severely abnormal, 8 times the reference frequency is adopted; where T1, T2, and T3 are preset temperature thresholds; for example, setting T1 = 45°C, T2 = 65°C, T3 = 85°C, and f0 = 5 min / time, the acquisition frequency is adjusted as follows: in the normal state, f = f0 = 5 minutes / time; slightly abnormal, f = 2f0 = 2.5 minutes / time; moderately abnormal, f = 4f0 = 1.25 minutes / time; severely abnormal, f = 8f0 = 0.625 minutes / time.

[0050] The parameter adjustment of the dual - light camera also determines the specific parameter values according to the real - time feedback. Among them, the adjustable parameters of the infrared camera include the temperature measurement range and emissivity; the adjustable parameters of the visible - light camera include the exposure time and ISO sensitivity.

[0051] Furthermore, the images can be 3 to 5 sets of dual-light images, each set of data including an infrared thermal image, a visible light image, and an acquisition timestamp.

[0052] In this embodiment, by dynamically adjusting the acquisition parameters, the acquisition process can be intelligently optimized according to the real-time temperature status and anomaly level of the device, ensuring the accuracy and reliability of temperature data, avoiding excessive data acquisition when the device temperature is normal, reducing unnecessary resource waste, and improving data sensitivity and detection accuracy when the device temperature is abnormal.

[0053] In an exemplary embodiment, as shown in FIG2, the above operation S102 performs preprocessing, feature extraction, spatial registration, and information fusion processing on the two-light image to generate a fused image. This can also be achieved through the following operations:

[0054] Operation S201 preprocesses the dual-light image to obtain the preprocessed dual-light image.

[0055] Operation S202 extracts multi-scale local features and regional features from the preprocessed dual-light image to construct feature description information.

[0056] Operation S203 involves registering the preprocessed dual-light image based on the feature description information to obtain the affine transformation matrix of the infrared thermal imaging image relative to the visible light image.

[0057] Operation S204 performs a fusion process on the preprocessed dual-light image based on the affine transformation matrix to generate a fused image.

[0058] The preprocessing includes at least quality assessment, image enhancement, and region segmentation.

[0059] Dual-light image preprocessing refers to performing necessary image processing on the acquired infrared thermal imaging image and visible light image to improve image quality and remove potential noise. Quality assessment is the process of evaluating image quality to ensure that the temperature information in the image is clear and reliable, often using metrics such as signal-to-noise ratio and contrast ratio. Image enhancement includes contrast enhancement, brightness adjustment, and edge sharpening, aiming to highlight important information in the image and improve feature extraction. Region segmentation methods (such as superpixel segmentation and edge-based segmentation algorithms) divide the image into multiple regions, facilitating independent analysis of the temperature characteristics of different regions during subsequent processing.

[0060] For example, the terminal performs quality assessment on the acquired dual-light images, using algorithms such as peak signal-to-noise ratio and structural similarity to evaluate image quality, and removes noise from the images through Gaussian filtering or median filtering. Next, histogram equalization is used to enhance image contrast, making temperature information more prominent, and superpixel segmentation is used to divide the image into regions, thereby extracting device regions with different functions.

[0061] During the feature extraction stage, the terminal extracts multi-scale local and regional features from the preprocessed image. Commonly used feature extraction methods include SIFT and SURF, as well as regional features such as local binary patterns. This feature description information is used for subsequent image registration and fusion processing.

[0062] In the spatial registration process, the terminal uses feature point matching algorithms (such as SIFT or SURF) to match the infrared thermal imaging image and the visible light image, calculating the affine transformation matrix between them. The matching results are then optimized by applying the RANSAC algorithm, resulting in an accurate registration.

[0063] Finally, the terminal performs information fusion processing on the infrared thermal imaging image and the visible light image based on the calculated affine transformation matrix. Information fusion methods include weighted averaging and pixel-level fusion, ensuring that the final fused image can simultaneously retain temperature data (infrared image) and structural data (visible light image), providing high-quality data support for subsequent temperature field reconstruction and hotspot detection.

[0064] In one embodiment, the following operation can be performed.

[0065] ① Calculate the quality score of the two-light image: Q(I)=w1·SNR(I)+w2·E(I)+w3·C(I).

[0066] Where SNR(I) is the signal-to-noise ratio, E(I) is the edge sharpness, C(I) is the contrast ratio, and w1, w2, and w3 are weighting coefficients, with w1 + w2 + w3 = 1. For example, w1 = 0.4, w2 = 0.3, and w3 = 0.3. The SNR is calculated using a signal-to-noise separation method based on Gaussian filtering. Edge sharpness is based on image gradient magnitude statistics. Contrast ratio is calculated using the Michelson contrast ratio of local regions.

[0067] Determine whether the quality score Q(I) of the two-light image is less than the threshold Q. thre If yes, then image enhancement processing is performed followed by region segmentation; otherwise, region segmentation is performed directly. For example, Q thre It can be set to 0.6.

[0068] ② Perform region segmentation on the dual-light image, extract the outline of the cabinet equipment based on the level set method; use a multi-scale superpixel segmentation algorithm to divide the image into several sub-regions; combine prior knowledge to identify key functional regions, and construct a region importance weight map W(x,y)=γ1·W temp (x,y)+γ2·W str (x,y), where W temp W represents the importance weight of temperature. str Let γ1 and γ2 be the structural importance weights, and γ1 and γ2 be the weight coefficients, with a sum of 1. Based on the structural design and operating principles of the power plant equipment, determine which areas are crucial to the reliability and stability of the equipment, such as power modules, control panels, and heat dissipation areas. These areas are critical to the normal operation of the equipment and therefore have higher weights. The energy function design for the level set method is as follows: Among them, E reg For the regularization term, E area For the region term, E edge For marginal terms, μ, ν, and λ are weighting coefficients.

[0069] A physical constraint model was established based on the structural characteristics of the power plant cabinet equipment, including the relative positional relationships of equipment components and the spatial distribution patterns of key functional areas. The positional relationships of various components in the image provided spatial constraints for temperature field reconstruction. By modeling the positions of these components, the temperature distribution can be optimized and adjusted according to the known equipment structure. For example, the power module is located at the bottom of the equipment or near the heat dissipation area, while the control board and sensors are located at the top or middle of the equipment.

[0070] ③ The enhancement process for visible light images includes histogram equalization and contrast enhancement; the enhancement process for infrared thermal imaging images includes temperature normalization and noise reduction.

[0071] Histogram equalization includes: setting a contrast limit threshold α = 3.0; histograms exceeding the threshold are cropped and redistributed; the image is divided into 8×8 sub-blocks, and histogram equalization is performed on each sub-block; bilinear interpolation is used to merge the boundaries between blocks. Contrast enhancement includes: employing piecewise linear transformation, grayscale mapping: g(x) = α·f(x) + β, where α is the contrast gain, ranging from [1.2 to 1.8], and β is the brightness adjustment parameter, ranging from [-30, 30], which is adaptively adjusted according to the image statistical characteristics. Temperature normalization is: In(x, y) = (I(x, y) - T min ) / (T max -T min ), T min T represents the minimum temperature value in the current image. max This represents the maximum temperature value in the current image. A median filter is used for noise reduction.

[0072] ④ Extract multi-scale local features from dual-light images, including: extracting SIFT feature points and their 128-dimensional descriptors from visible light images; extracting SURF feature points and their 64-dimensional descriptors from infrared images; and extracting Harris corner features from both dual-light images.

[0073] Extracting regional features from dual-light images includes: extracting shape descriptors based on the HOG operator; extracting temperature distribution pattern features from infrared images, including temperature gradient direction histograms and hotspot distribution features; extracting texture features based on the LBP operator; for example, the hotspot distribution features are segmented using a threshold and connected component marking is performed, and a spatial distribution description is formed through centroid calculation, including a hotspot density map, hotspot distance statistics, and shape features.

[0074] Constructing a context-aware feature descriptor D: D = [D local D context D prior ].

[0075] Among them, D local For local feature description, SIFT / SURF descriptors and Harris corner features are fused; D context This provides neighborhood structure information, including HOG shape descriptors, temperature distribution pattern features, and LBP texture features, used to describe the structural characteristics of the region surrounding the feature point; D prior These are constraints based on prior knowledge, including regional importance weights and physical constraint information.

[0076] Feature descriptor D achieves an organic combination of multi-scale and multi-modal features by fusing local feature descriptions (SIFT / SURF and Harris features), contextual structure information (HOG shape descriptor, temperature distribution features, and LBP texture features), and constraints based on prior knowledge (regional importance weights and physical constraint information). This not only ensures the accuracy of feature matching but also makes full use of regional structure information and domain expertise, thereby improving the accuracy and reliability of image registration and anomaly detection for power plant cabinet equipment.

[0077] In this embodiment, by preprocessing, feature extraction, spatial registration, and information fusion of the dual-light images, image data optimization and precise alignment can be achieved, eliminating errors and noise that may occur during image acquisition. By effectively fusing temperature data from infrared images with structural information from visible light images, the resulting fused image provides accurate and comprehensive data support for subsequent temperature field reconstruction and hotspot detection, significantly improving the precision and accuracy of temperature monitoring. Ultimately, this enhances the real-time temperature monitoring capability of power plant cabinet equipment, especially demonstrating high reliability in identifying localized temperature anomalies and potential fault areas.

[0078] In an exemplary embodiment, the above operation S203, which performs registration processing on the preprocessed dual-light image based on feature description information to obtain the affine transformation matrix of the infrared thermal imaging image relative to the visible light image, further includes: selecting initial registration points from feature points corresponding to the feature description information based on the region segmentation result of the dual-light image; obtaining the region segmentation result through region segmentation; selecting candidate registration points from the initial registration points based on the feature description information; using a random sampling consistency model to eliminate erroneous registration points among the candidate registration points to obtain the target registration point; and calculating and determining the affine transformation matrix based on the target registration point.

[0079] For example, the terminal first segments the infrared thermal imaging image and the visible light image into different regions by region segmentation of the dual-light image. Within these regions, key feature points (such as corner points and edge points) are extracted, and initial registration points are selected based on the feature descriptors of each region. These initial registration points serve as the starting point for registration, and a feature matching algorithm further filters out candidate registration points with high matching degrees. Then, the terminal uses the RANSAC algorithm for optimization, eliminating erroneous matching points. RANSAC calculates a preliminary affine transformation matrix by randomly selecting a subset from the candidate points, and filters the optimal registration point by verifying the applicability of this transformation matrix to other points. Finally, the target registration point is selected, and the accurate affine transformation matrix is ​​calculated. The affine transformation matrix contains transformation parameters such as translation, rotation, and scaling, enabling accurate alignment of the infrared thermal imaging image onto the visible light image. Through this matrix, the terminal can accurately align the infrared thermal imaging image and the visible light image, ensuring that subsequent image analysis is based on consistent spatial coordinates, thus providing a reliable data foundation for temperature field reconstruction and hotspot detection.

[0080] In one embodiment, a visible light image can provide clear edge and structural information, so it is selected as the reference image for registration, and an infrared thermal image is selected as the image to be registered.

[0081] ① Coarse matching based on region segmentation results: The matching region is divided using the region feature map G(V, E), where the vertex set V represents the segmented region and the edge set E represents the spatial adjacency relationship between regions. For example, in region connectivity analysis, the common edge determination is used, i.e., whether two regions share boundary pixels, and an adjacency matrix is ​​constructed based on the common edge determination; the priority search region is determined according to the region importance weight W(x, y); within the priority search region, the nearest neighbor search of the feature descriptor D is performed using a kd-tree to obtain the initial matching pair set M1; during the nearest neighbor search, k can take the value 2, i.e., take 2 nearest neighbors, the feature dimension of the kd-tree is the dimension of D, and it is split according to the dimension with the largest variance, the leaf node capacity is a maximum of 10 feature points, and during the search, the distance ratio threshold is set to 0.8, and the search radius is limited to 100 pixels.

[0082] ② Perform fine-grained matching based on spatial constraints: calculate the Euclidean distance of feature descriptors as a similarity measure; construct the spatial constraint SC: SC(p i ,p j )=a1d(p i ,p j )+a2·|θ(p i ,p j )|.

[0083] Where, p i ,p j Let d(p) represent two feature point matching pairs to be evaluated. Each matching pair contains a feature point in the reference image and a feature point in the image to be registered. i ,p j θ(p) represents the distance between pairs of feature points. i ,p j ) represents the angle between the line connecting the feature point pairs and the horizontal direction, where a1 and a2 are weighting coefficients; d(p i ,p j ) is the distance normalization, i.e., d(p) i ,p j )=d(p i ,p j ) / d max ;|θ(p i ,p j ) is angle normalization, i.e., |θ(p) i ,p j )|=θ(p i ,p j ) / θ max d max θ max These are the maximum distance threshold and the maximum angle threshold, respectively.

[0084] The initial set of matching pairs M1 is selected based on spatial constraints SC, and the optimized set of matching pairs M2 is obtained; for example, the SC threshold is set to 0.8, that is, matching pairs with SC less than 0.8 are retained.

[0085] ③ Estimate the affine transformation matrix H using the RANSAC algorithm, and define the confidence evaluation function as follows:

[0086] Score(p i ,p j )=ω1·S f (p i ,p j )+ω2S g (p i ,p j )+ω3·S p (pi ,p j )

[0087] Among them, S f S is the cosine similarity of the feature descriptor D; g For point pairs (p i ,p j The degree of agreement of S with respect to the affine transformation matrix H; p The prior constraints are based on the weights of regional importance; ω1, ω2, and ω3 are weight coefficients and ω1+ω2+ω3=1.

[0088] For example, the affine transformation matrix estimation process is as follows: (1) Random sampling: select 3 pairs of matching points each time; (2) Calculate the affine matrix: 6 degrees of freedom; (3) Inner point statistics: calculate the projection error.

[0089] The final matching pair set M3 is obtained by filtering based on the Score threshold. The affine transformation matrix is ​​then calculated using the final matching pair set M3 to complete image registration. For example, the Score threshold can be set to 0.7, meaning that matching pairs with a Score > 0.7 are retained as the final matching pair set M3.

[0090] ④ Then, the parameters of the affine transformation matrix can be solved using the least squares method. After that, the pixels of the infrared thermal imaging image are transformed to the pixel coordinate system of the visible light image according to the affine transformation matrix, thus completing the image registration.

[0091] For example, there are n pairs of matching points (x i y i ) and (x i ′,y i (Each pair of matching points contains the coordinates of the visible light image and the coordinates of the infrared thermal image). For each pair of matching points, the following two equations can be obtained:

[0092] ax i +by i +c=x′ i dx i +ey i +f=y′ i

[0093] Where a, b, c, d, e, and f are the affine transformation parameters to be determined.

[0094] Rearrange these equations into matrix form: A·θ=B.

[0095] Since the system of equations may be overdetermined (i.e., 2N > 6), the least squares method is needed to solve for the parameter vector θ. The formula for the least squares solution is: θ = (A T A) -1 AT B. T is the matrix transpose.

[0096] Alternatively, it can be solved using methods such as QR decomposition (orthogonal triangular decomposition) and singular value decomposition (SVD).

[0097] By employing a priority search mechanism guided by the regional feature map G(V,E) and a fine matching process based on spatial constraints SC, combined with regional importance weights W(x,y) and equipment physical characteristic constraints, multiple constraints and optimizations are achieved in the registration process. In conjunction with the screening mechanism of the confidence evaluation function Score, both registration accuracy and computational complexity are guaranteed. This approach is particularly suitable for scenarios with specific structural characteristics, such as power plant cabinet equipment, and exhibits strong robustness and practicality.

[0098] In this embodiment, high-precision image registration is achieved by combining region segmentation, feature description information filtering, and the RANSAC optimization algorithm. Through the selection of initial registration points, candidate registration points, and RANSAC algorithm optimization, accurate alignment of the infrared thermal imaging image and the visible light image is ensured, eliminating registration errors caused by different image shooting angles or other factors. The resulting affine transformation matrix provides a precise spatial coordinate basis for subsequent temperature field reconstruction and hotspot detection, thereby significantly improving the accuracy and reliability of temperature monitoring.

[0099] In an exemplary embodiment, the above operation S204, which fuses the preprocessed dual-light image according to the affine transformation matrix to generate a fused image, further includes: obtaining a registered infrared thermal imaging image based on the affine transformation matrix and the infrared thermal imaging image; obtaining an edge map of the visible light image and a temperature gradient map of the registered infrared thermal imaging image; determining the fusion weight of the dual-light image based on the edge map and the temperature gradient map; and fusing the registered infrared thermal imaging image and the visible light image according to the fusion weight to obtain the fused image.

[0100] For example, the terminal first aligns the infrared thermal imaging image using an affine transformation matrix, precisely registering it to the spatial coordinate system of the visible light image. Then, edge detection is applied to the visible light image to extract edge information. The edge map, generated from edge information extracted from the visible light image, highlights obvious structural boundaries and contours on the device surface. Edge information plays a crucial role in the fusion process, determining which regions require high weight. Temperature gradients are then calculated on the infrared image to generate a temperature gradient map. Next, the terminal calculates the weight of each image region using a weighted fusion algorithm. Hotspot and edge regions are assigned higher weights, indicating higher structural importance for edge regions and greater temperature gradients indicating temperature anomalies, ensuring that key information from these regions is fully represented in the final fused image. Finally, based on these fusion weights, the terminal fuses the infrared thermal imaging image with the visible light image to generate a high-quality fused image. Common fusion methods include weighted averaging, which adjusts the contribution ratio of the infrared and visible light images according to the importance (weight) of the regions, ensuring that the final fused image simultaneously retains structural information (from the visible light image) and temperature information (from the infrared image). For example, temperature information in hotspot areas may rely more on infrared images, while edge areas may rely more on visible light images. This image combines temperature and structural information, providing a precise data foundation for subsequent temperature field reconstruction and hotspot detection.

[0101] In one embodiment, the registered infrared thermal imaging image IR and visible light image VIS are normalized to unify the image value range to the [0, 1] interval, thereby unifying the numerical range of the two images and eliminating dimensional differences.

[0102] Calculate the Sobel edge map Ed(x, y) of a visible light image. Sobel operator calculation:

[0103] Horizontal gradient:

[0104] Vertical gradient:

[0105] Edge strength: Edge maps reflect structural information and contour features in visible light images.

[0106] Extract the temperature gradient map Tg(x, y) from the infrared thermal imaging image, including the horizontal temperature gradient: Vertical temperature gradient: Temperature gradient magnitude: Temperature gradient maps reflect the degree of temperature change in infrared images.

[0107] Generate adaptive weights Z(x,y) = μ1·Ed(x,y) + μ2·Tg(x,y), where μ1 and μ2 are adjustment coefficients and μ1 + μ2 = 1; when setting, μ1 < μ2 can be set so that the high edge response region tends to retain visible light information and the high temperature gradient region tends to retain infrared information.

[0108] Image fusion generation: F(x,y)=Z(x,y)·IR(x,y)+(1-Z(x,y))·VIS(x,y).

[0109] Where F(x, y) is the fused image, IR(x, y) is the normalized infrared thermal image, and VIS(x, y) is the normalized visible light image. Using this fusion formula, more visible light details are preserved in areas with significant edges, and more infrared information is preserved in areas with large temperature gradients, achieving adaptive fusion of structural and temperature information.

[0110] The dual-light image fusion algorithm introduces adaptive weights Z(x,y) to organically combine the Sobel edge map Ed(x,y) and the temperature gradient map Tg(x,y), achieving complementary advantages between the structural details of visible light images and the temperature information of infrared images. Through adjustable weight coefficients μ1 and μ2, the system can flexibly adjust the fusion ratio of structural and temperature information according to actual application requirements, ensuring both the visual quality of the fused image and the accurate transmission of temperature information.

[0111] In this embodiment, through precise registration of the affine transformation matrix and weighted fusion based on the edge map and temperature gradient map, a precise fused image can be generated. This image not only retains the temperature distribution information of the power plant cabinet equipment but also accurately presents the structural details of the equipment surface. The weighted fusion method ensures that information from hotspot areas, edge areas, and other critical areas is preferentially preserved, improving the overall information content and accuracy of the image. This provides a reliable foundation for subsequent temperature field reconstruction, hotspot detection, and temperature early warning, significantly improving the accuracy of equipment monitoring and fault prediction.

[0112] In an exemplary embodiment, the above operation S103 performs temperature field reconstruction processing and hotspot detection processing on the fused image to obtain temperature field distribution information and hotspot region information. It further includes: obtaining initial temperature field distribution information based on the registered infrared thermal imaging image; obtaining temperature field distribution information based on the initial temperature field distribution information, temperature gradient map, fused image, preset reference temperature, and preset power station cabinet equipment region importance weights; determining hotspot regions based on the temperature field distribution information and the region segmentation results of the dual-light image using a threshold judgment method to obtain hotspot region information; and obtaining the region segmentation results through region segmentation processing.

[0113] For example, the terminal first obtains preliminary temperature field distribution information from the registered infrared thermal imaging image, which shows the temperature changes in various areas of the device surface. The infrared thermal imaging image provides temperature information for each area of ​​the device surface. After preprocessing and registration, a preliminary temperature field distribution map is obtained, presenting a rough distribution of the device surface temperature. Then, combining the temperature gradient map and the fused image, an optimized temperature field distribution map is obtained through weighted calculation based on a preset reference temperature and regional importance weights. Considering the different functional areas of the device and the importance of each area, key areas (such as power supplies and control boards) are given higher weights to ensure they receive more attention in the temperature field analysis. The preset reference temperature provides a standard reference temperature value for the device, used to compare with the actual temperature of the device surface and identify areas with abnormal temperatures. Based on the temperature field distribution and the segmentation of device areas, the terminal uses a threshold judgment method to identify areas with excessively high temperatures (i.e., hot spots) and outputs hot spot information including the hot spot location, temperature value, and size.

[0114] In one embodiment, ① temperature field reconstruction is performed based on the fused image F(x,y), and the reconstruction function is: T(x,y) = T base +K(x,y)·F(x,y)·W(x,y)+div(T g (x,y)).

[0115] Where T(x, y) is the reconstructed temperature field distribution function, representing the actual temperature value at point (x, y) in the plane coordinate system of the cabinet equipment; T base The baseline temperature value can be set to the ambient temperature; F(x, y) represents the fused image; W(x, y) represents the region importance weight; T g (x, y) represents the temperature gradient plot; div represents the divergence operator; K(x, y) is the correction coefficient, K(x, y) = K0· ( 1+β1·G(x,y))·(1+β2·L(x,y)), where K0 is the baseline correction coefficient obtained from camera calibration, G(x,y) is the normalized temperature gradient magnitude, L(x,y) is the local temperature uniformity, and β1 and β2 are weighting coefficients.

[0116] Furthermore, the divergence operator acts on the temperature gradient map T g (x, y), the calculation formula is: Where T g (x) and T g (y) represents the temperature gradient components in the x and y directions, respectively. This operator is used to describe the spatial diffusion characteristics of the temperature field and reflects the degree of heat accumulation or dispersion.

[0117] Calculation of the normalized temperature gradient magnitude G(x, y): First, calculate the partial derivatives of the temperature in the x and y directions: Calculate the gradient magnitude: Normalization: G(x,y)=G'(x,y) / max(G'(x,y)), where G(x,y) represents the degree of temperature change, and the larger the value, the faster the temperature changes.

[0118] Calculation of local temperature uniformity L(x, y): Calculate the local standard deviation within the local neighborhood N(x, y) of point (x, y): Where μ is the local mean; local uniformity: L(x,y)=1-σ(x,y) / max(σ); L(x,y) describes the smoothness of the temperature distribution, and the larger the value, the more uniform the temperature distribution.

[0119] The reconstruction function achieves the mapping transformation from image grayscale values ​​to actual temperature values ​​by combining reference temperature, fused image information, regional weights, and temperature gradient characteristics, while also taking into account the spatial distribution characteristics of the temperature field.

[0120] ② Based on T(x, y), calculate local statistical features, use the threshold method to determine hotspot areas, extract the feature parameters of hotspot areas, including the highest temperature, area and shape features, establish a hotspot tracking table, and record the location and temperature changes.

[0121] The criteria for determining hotspot regions are: H(x,y)={(x,y)|T(x,y)>min(μ+κσ(N(x,y)),T thre )}.

[0122] Where H(x, y) represents the hotspot region; N(x, y) represents the local neighborhood; μ is the local mean; σ is the local standard deviation; κ is the detection coefficient; T thre This is the preset temperature threshold.

[0123] Furthermore, when determining the local neighborhood, with point (x, y) as the center, combined with the above-mentioned region segmentation results and considering the physical structural characteristics of the device, N(x, y) = {(i, j) | (i, j) belong to the same superpixel region and |ix|≤r and |jy|≤r}, where r is the search radius, for example, a value of 5-7 pixels. The same superpixel region refers to the result of the above multi-scale superpixel segmentation, requiring (i, j) and (x, y) to be in the same functional region. During calculation, the superpixel region where point (x, y) is located is determined, and a window of (2r+1)×(2r+1) is taken within this region. The mean temperature μ and standard deviation σ within this window are calculated, and it is determined whether T(x, y) exceeds the threshold.

[0124] In this embodiment, by combining temperature field reconstruction, temperature gradient mapping, region segmentation, and hotspot detection methods, accurate monitoring of the temperature status of power plant cabinet equipment can be achieved. The temperature field distribution information has been optimized through comprehensive analysis of multiple factors, particularly focusing on critical and high-temperature areas, significantly improving the ability to identify temperature anomalies. The hotspot detection method, based on threshold judgment, can quickly locate high-temperature areas within the equipment and provide timely early warning information, effectively preventing equipment failures or catastrophic events.

[0125] In an exemplary embodiment, the above-described operation S104, which analyzes the temperature of the power station cabinet equipment based on temperature field distribution information and hotspot area information to obtain temperature warning information for the power station cabinet equipment, further includes: acquiring historical temperature field distribution information and historical hotspot area information corresponding to historical data collection; determining temperature magnitude change trend prediction information and hotspot area expansion trend information based on historical temperature field distribution information, historical hotspot area information, temperature field distribution information, and hotspot area information; determining the temperature anomaly level of the current data collection based on the temperature magnitude change trend prediction information, hotspot area expansion trend information, and temperature field distribution information; and generating temperature warning information for the power station cabinet equipment based on the temperature anomaly level of the current data collection, the temperature magnitude change trend prediction information, the hotspot area expansion trend information, and the temperature field distribution information.

[0126] For example, the terminal acquires historical data on historical temperature field distribution and historical hotspot areas, and combines this with current temperature field distribution information to analyze temperature change trends and hotspot expansion trends. Temperature magnitude change trend prediction information, based on a comparison of historical temperature data and the current temperature field distribution, predicts whether the temperature will rise or fall in the future. Hotspot expansion trend information, by comparing the current temperature of hotspot areas with the historical trends of hotspot areas, predicts whether the hotspot areas will further expand. Then, based on this data, the terminal calculates the temperature anomaly level and generates corresponding warning information.

[0127] For example, if the current temperature trend shows an upward movement and the hotspot area continues to expand, the terminal may issue a "severe anomaly" warning and recommend further cooling or equipment inspection measures. The temperature anomaly level is comprehensively assessed based on the currently collected temperature field distribution and historical temperature field distribution data, combined with predicted temperature change trends and hotspot area expansion trends. By analyzing the rate of temperature change and the expansion of hotspot areas, the current temperature anomaly level can be accurately determined, such as: normal, mild anomaly, moderate anomaly, severe anomaly, etc. Finally, based on the determined temperature anomaly level, temperature change trend, hotspot area expansion trend, and temperature field distribution information, a temperature warning message is generated. The temperature warning message will clearly indicate the degree of anomaly in the current equipment temperature and provide targeted warning response suggestions, such as whether shutdown, cooling, or other maintenance measures are required.

[0128] In one embodiment, the temperature field T(x, y) output by the temperature monitoring module and the hotspot detection results are processed in a time series to establish a monitoring data sequence {T(x, y, t)}. The data is then spatially weighted using a regional importance weight W(x, y) to generate a regional temperature feature vector, including the highest temperature, average temperature, temperature standard deviation, and hotspot area. Then, a sliding time window method is used to calculate the temperature change rate and acceleration, and exponential smoothing is used to predict the temperature trend, establishing a trend evaluation index: R(t) = ξ1·(ΔT / Δt) + ξ2·(Δt / Δt) 2 T / Δt 2 )+ξ3·P(t).

[0129] Where ΔT / Δt represents the rate of change of temperature with time, Δ 2 T / Δt 2 Let P(t) represent the rate of change of temperature, P(t) be the temperature prediction error, and ξ1, ξ2, and ξ3 be weighting coefficients, with ξ1 + ξ2 + ξ3 = 1.

[0130] Furthermore, ΔT / Δt=(T(t)-T(t-1)) / Δt, Δ 2 T / Δt 2 = ( ΔT / Δt(t) - ΔT / Δt(t-1)) / Δt, the temperature prediction deviation P(t) is the temperature trend predicted using exponential smoothing, P(t) = |T pre (t)-T act (t)|,T pre (t) is the exponentially smoothed predicted value based on historical data, T act (t) represents the actual measured value.

[0131] Next, based on multi-level temperature thresholds and trend evaluation metrics, an anomaly discrimination model is constructed. The discrimination criteria include the degree of temperature overrun, the duration of overrun, the change trend, and the spatial distribution characteristics, and the temperature anomaly level is output. The degree of temperature overrun is the comparison between the actual temperature and the preset thresholds T1, T2, and T3. The duration of overrun is the duration of the abnormal state. The change trend is the numerical value and change direction of the trend evaluation metric R(t). The spatial distribution characteristics include the shape characteristics and expansion trend of the hot spot area. The anomaly recognition unit not only considers the static temperature threshold judgment but also incorporates the dynamic trend analysis results. Through the comprehensive evaluation of these characteristics, the accurate recognition and level classification of the abnormal state of the equipment are achieved, providing a reliable basis for early warning decision-making.

[0132] Generate early warning information based on the anomaly recognition results, including the temperature anomaly level, the location of the abnormal area, the highest temperature value, and the temperature change trend, and push the early warning information to the system management module.

[0133] For example, the temperature anomaly level is the same as the level corresponding to the acquisition control unit in the image acquisition module, and the specific settings are: normal: T < T1; mild anomaly: T1 ≤ T < T2; moderate anomaly: T2 ≤ T < T3; severe anomaly: T ≥ T3.

[0134] Generate an early warning report based on the anomaly recognition results, including key information such as the temperature anomaly level (normal, mild anomaly, moderate anomaly, severe anomaly), the precise position coordinates of the abnormal area, the current highest temperature value, and the temperature change trend based on the trend evaluation metric R(t).

[0135] In this embodiment, by combining historical data and real-time temperature data, the temperature change trend of the power station cabinet equipment and the expansion of the hot spot area can be accurately predicted, significantly improving the accuracy of temperature anomaly detection and early warning. Generate the temperature anomaly level based on the temperature field distribution information, historical data, and trend prediction, and generate the temperature early warning information in a timely manner, helping the operation and maintenance personnel quickly identify the potential failure risks of the equipment and take corresponding measures. It not only improves the early warning ability of equipment failures but also effectively avoids equipment damage and accidents caused by temperature anomalies.

[0136] In an exemplary embodiment, as shown in Figure 3, the present application provides a temperature early warning system for power station cabinet equipment, which includes:

[0137] An image acquisition module 301 for real-time acquisition of the dual-light images of the cabinet equipment, that is, infrared thermal imaging images and visible light images;

[0138] An image processing module 302 for preprocessing, feature extraction, spatial registration, and information fusion of the acquired dual-light images to generate a fused image;

[0139] Temperature monitoring module 303 is used for temperature field reconstruction and hotspot detection of the fused image;

[0140] The intelligent analysis module 304 is used to perform trend analysis on the output data of the temperature monitoring module, perform anomaly identification, and generate early warning information.

[0141] The system management module 305 is used to realize monitoring data storage, abnormal event recording, early warning information push and remote access control, and provides a human-machine interaction interface.

[0142] The image acquisition module 301 includes: an infrared image acquisition unit for controlling infrared camera parameters and acquiring infrared thermal imaging images; a visible light image acquisition unit for controlling visible light camera parameters and acquiring visible light images; and an acquisition control unit for receiving real-time feedback from the intelligent analysis module 304 through the system management module 305, thereby enabling the acquisition triggering and parameter adjustment of the dual-light camera.

[0143] The image processing module 302 includes: a preprocessing unit for quality assessment, image enhancement, and region segmentation of the acquired infrared thermal imaging image and visible light image; a feature extraction unit for extracting multi-scale local features and regional features from the preprocessed dual-light image and constructing feature descriptors; an image registration unit for registering the dual-light image using a hierarchical registration strategy; and an information fusion unit for fusing the registered dual-light image to generate a fused image.

[0144] The temperature monitoring module 303 includes: a temperature field reconstruction unit for reconstructing the temperature field based on the fused image; and a hotspot detection unit for calculating local statistical features based on the temperature field, determining hotspot regions using a threshold method, and extracting feature parameters of the hotspot regions.

[0145] The intelligent analysis module 304 includes: a data processing unit, used to process the temperature field and hotspot detection results output by the temperature monitoring module in a time series, establish a monitoring data sequence, spatially weight the data by combining regional importance weights, and generate a regional temperature feature vector, including the highest temperature, average temperature, temperature standard deviation, and hotspot area; a trend analysis unit, used to calculate the temperature change rate and acceleration using a sliding time window method, use exponential smoothing to predict the temperature trend, and establish a trend evaluation index; an anomaly identification unit, based on multi-level temperature thresholds and trend evaluation indexes, constructs an anomaly discrimination model, with discrimination criteria including the degree of temperature exceeding limits, duration of exceeding limits, change trend, and spatial distribution characteristics, and outputs the temperature anomaly level; and an early warning generation unit, which generates early warning information based on the anomaly identification results, including the temperature anomaly level, anomaly area location, highest temperature value, and temperature change trend, and pushes the early warning information to the system management module 305.

[0146] The system management module 305, as the control center of the entire early warning system, is responsible for implementing the storage management of monitoring data, recording and tracking abnormal events, and timely pushing early warning information. At the same time, it provides a human-computer interaction interface. This module receives the real-time feedback from the intelligent analysis module 304 and automatically adjusts the acquisition frequency of the acquisition control unit according to the temperature anomaly level (T<T1 is the normal temperature range, T1≤T<T2 is a mild anomaly, T2≤T<T3 is a moderate anomaly, T≥T3 is a severe anomaly), corresponding to the reference frequency f0, 2 times the reference frequency, 4 times the reference frequency, and 8 times the reference frequency respectively), and accordingly adjusts the parameter configuration of the dual-light camera. In addition, this module also provides a remote access control function, enabling operation and maintenance personnel to view the device status, adjust system parameters, and respond to abnormal early warnings at any time, realizing the intelligent management and control of the entire early warning system.

[0147] In this embodiment, (1) through adaptive acquisition control, multi-feature fusion registration, information fusion with regional importance weighting, and a multi-level early warning mechanism based on trend analysis, the accuracy of temperature monitoring of power station cabinet equipment is improved, timely early warning of temperature anomalies is achieved, and a strong guarantee is provided for the safe operation of the power system; (2) an adaptive acquisition frequency adjustment mechanism based on the temperature anomaly level is adopted to dynamically adjust the acquisition frequency according to the device temperature status, optimizing the utilization of system resources while ensuring the monitoring effect, improving the system operation efficiency, and enhancing the system's response speed to abnormal situations; (3) by constructing a feature descriptor that includes local features, context information, and prior constraints, combined with a hierarchical registration strategy, the accuracy and robustness of dual-light image registration are improved, effectively solving the problem of difficult image registration in complex environments; (4) an adaptive fusion method based on regional importance weights and temperature gradient information makes full use of the complementary characteristics of dual-light images, improving the quality of the fused image and providing a reliable data basis for subsequent temperature monitoring; (5) a temperature field reconstruction model considering local statistical features, combined with a hot spot detection method based on regional importance weights, improves the accuracy of temperature monitoring and can effectively identify and locate temperature anomaly regions; (6) a multi-level early warning mechanism based on time-series data analysis, through trend evaluation indicators and multi-dimensional discriminant models, realizes early warning of temperature anomalies, improves the accuracy and timeliness of system early warning, and provides an effective decision-making basis for equipment maintenance.

[0148] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0149] Based on the same inventive concept, this application also provides a power plant cabinet equipment temperature warning device for implementing the above-mentioned power plant cabinet equipment temperature warning method. The solution provided by this device is similar to the solution described in the above method. Therefore, the specific limitations of one or more power plant cabinet equipment temperature warning device embodiments provided below can be found in the limitations of the power plant cabinet equipment temperature warning method above, and will not be repeated here.

[0150] In an exemplary embodiment, as shown in FIG4, a temperature early warning device for power plant cabinet equipment is provided, including: an image acquisition module 401, an image fusion module 402, a temperature detection module 403, and a temperature analysis module 404, wherein:

[0151] The image acquisition module 401 is used to acquire dual-light images of the power station cabinet equipment to be analyzed according to the acquisition parameters corresponding to this acquisition; the dual-light images include infrared thermal imaging images and visible light images.

[0152] The image fusion module 402 is used to perform preprocessing, feature extraction, spatial registration and information fusion processing on the dual-light image to generate a fused image.

[0153] Temperature detection module 403 is used to perform temperature field reconstruction processing and hotspot detection processing on the fused image to obtain temperature field distribution information and hotspot region information;

[0154] The temperature analysis module 404 is used to perform temperature analysis on the power station cabinet equipment based on temperature field distribution information and hot spot area information, and obtain temperature warning information for the power station cabinet equipment.

[0155] In one embodiment, the above-mentioned power plant cabinet equipment temperature early warning device further includes a data acquisition parameter determination module, which is used to take the data acquisition parameter corresponding to the lowest temperature anomaly level as the data acquisition parameter for this data acquisition when this data acquisition is the first data acquisition; and to take the data acquisition parameter corresponding to the temperature anomaly level determined in the previous data acquisition as the data acquisition parameter for this data acquisition when this data acquisition is not the first data acquisition.

[0156] In one embodiment, the image fusion module 402 is further configured to preprocess the two-light image to obtain a preprocessed two-light image; the preprocessing includes at least quality assessment processing, image enhancement processing, and region segmentation processing; extract multi-scale local features and regional features from the preprocessed two-light image to construct feature description information; perform registration processing on the preprocessed two-light image according to the feature description information to obtain an affine transformation matrix of the infrared thermal imaging image relative to the visible light image; and perform fusion processing on the preprocessed two-light image according to the affine transformation matrix to generate a fused image.

[0157] In one embodiment, the image fusion module 402 is further configured to: select initial registration points from feature points corresponding to feature description information based on the region segmentation result of the two-light image; obtain the region segmentation result through region segmentation processing; select candidate registration points from the initial registration points based on feature description information; use a random sampling consistency model to remove erroneous registration points from the candidate registration points to obtain the target registration point; and calculate and determine the affine transformation matrix based on the target registration point.

[0158] In one embodiment, the image fusion module 402 is further configured to obtain a registered infrared thermal imaging image based on the affine transformation matrix and the infrared thermal imaging image; obtain an edge map of the visible light image and a temperature gradient map of the registered infrared thermal imaging image; determine the fusion weight of the two-light image based on the edge map and the temperature gradient map; and perform fusion processing on the registered infrared thermal imaging image and the visible light image based on the fusion weight to obtain a fused image.

[0159] In one embodiment, the temperature detection module 403 is further configured to obtain initial temperature field distribution information based on the registered infrared thermal imaging image; obtain temperature field distribution information based on the initial temperature field distribution information, temperature gradient map, fused image, preset reference temperature and preset power station cabinet equipment area importance weight; determine hot spot area information by using a threshold judgment method based on the temperature field distribution information and the area segmentation result of the dual-light image; and obtain the area segmentation result through area segmentation processing.

[0160] In one embodiment, the temperature analysis module 404 is further configured to acquire historical temperature field distribution information and historical hotspot area information corresponding to historical data collection; determine temperature magnitude change trend prediction information and hotspot area expansion trend information based on the historical temperature field distribution information, historical hotspot area information, temperature field distribution information, and hotspot area information; determine the temperature anomaly level of the current data collection based on the temperature magnitude change trend prediction information, hotspot area expansion trend information, and temperature field distribution information; and generate temperature warning information for the power station cabinet equipment based on the temperature anomaly level of the current data collection, the temperature magnitude change trend prediction information, the hotspot area expansion trend information, and the temperature field distribution information.

[0161] Each module in the aforementioned power plant cabinet temperature early warning device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the computer device's memory as software, so that the processor can call and execute the corresponding operations of each module.

[0162] In an exemplary embodiment, a computer device is provided, which may be a terminal, and its internal structure diagram is shown in Figure 5. The computer device includes a processor, memory, input / output interface, communication interface, display unit, and input device. The processor, memory, and input / output interface are connected via a system bus, and the communication interface, display unit, and input device are also connected to the system bus via the input / output interface. The processor of the computer device provides computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores an operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The input / output interface of the computer device is used for exchanging information between the processor and external devices. The communication interface of the computer device is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, mobile cellular networks, Near Field Communication (NFC), or other technologies. When the computer program is executed by the processor, it implements a method for temperature early warning of power plant cabinet equipment. The display unit of the computer device is used to form a visually visible image and may be a display screen, a projection device, or a virtual reality imaging device. The display screen can be an LCD screen or an e-ink screen. The input device of the computer device can be a touch layer covering the display screen, or buttons, trackballs, or touchpads set on the casing of the computer device, or external keyboards, touchpads, or mice, etc.

[0163] Those skilled in the art will understand that the structure shown in Figure 5 is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or may combine certain components, or may have different component arrangements.

[0164] In one embodiment, a computer device is also provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps in the above method embodiments.

[0165] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon that, when executed by a processor, implements the steps in the above method embodiments.

[0166] In one embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps in the above method embodiments.

[0167] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data must comply with relevant regulations.

[0168] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, artificial intelligence (AI) processors, etc., and are not limited to these.

[0169] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0170] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this patent application should be determined by the appended claims.

Claims

1. A method for early warning of temperature in power plant cabinet equipment, characterized in that, The method includes: If this is the first data collection, the data collection parameters corresponding to the lowest temperature anomaly level will be used as the data collection parameters for this data collection; if this is not the first data collection, the data collection parameters corresponding to the temperature anomaly level determined in the previous data collection will be used as the data collection parameters for this data collection. According to the acquisition parameters corresponding to this acquisition, dual-light images of the power station cabinet equipment to be analyzed are acquired; the dual-light images include infrared thermal imaging images and visible light images; The two-light image is preprocessed to obtain a preprocessed two-light image; the preprocessing includes at least quality assessment processing, image enhancement processing, and region segmentation processing; multi-scale local features and regional features are extracted from the preprocessed two-light image to construct feature description information; based on the region segmentation result of the two-light image, initial registration points are selected from the feature points corresponding to the feature description information; the region segmentation result is obtained through the region segmentation processing; based on the feature description information, candidate registration points are selected from the initial registration points; a random sampling consistency model is used to remove erroneous registration points from the candidate registration points to obtain the target registration point; based on the... The target registration points are defined, and the affine transformation matrix of the infrared thermal imaging image relative to the visible light image is calculated and determined. Based on the affine transformation matrix and the infrared thermal imaging image, a registered infrared thermal imaging image is obtained. Using edge information from the visible light image, an edge map of the visible light image is obtained, and a temperature gradient map of the registered infrared thermal imaging image is calculated. Based on the edge map and the temperature gradient map, the fusion weights of the two images are determined. Based on the fusion weights, the registered infrared thermal imaging image and the visible light image are fused to obtain a fused image. The fused image is subjected to temperature field reconstruction processing and hotspot detection processing to obtain temperature field distribution information and hotspot region information; the temperature field distribution information is determined based on the initial temperature field distribution information, the temperature gradient map of the registered infrared thermal imaging image, the fused image, the preset reference temperature, and the preset importance weight of the power station cabinet equipment area; the initial temperature field information is obtained based on the registered infrared thermal imaging image. Based on the temperature field distribution information and the hot spot area information, temperature analysis is performed on the power station cabinet equipment to obtain temperature early warning information for the power station cabinet equipment.

2. The method according to claim 1, characterized in that, The process of performing temperature field reconstruction and hotspot detection on the fused image to obtain temperature field distribution information and hotspot region information includes: Based on the registered infrared thermal imaging image, the initial temperature field distribution information is obtained; The temperature field distribution information is obtained based on the initial temperature field distribution information, the temperature gradient map, the fused image, the preset reference temperature, and the preset importance weight of the power station cabinet equipment area; Based on the temperature field distribution information and the region segmentation results of the dual-light image, hotspot region information is determined using a threshold judgment method; the region segmentation results are obtained through the region segmentation process.

3. The method according to claim 1, characterized in that, The step of performing temperature analysis on the power station cabinet equipment based on the temperature field distribution information and the hot spot area information to obtain temperature early warning information for the power station cabinet equipment includes: Obtain historical temperature field distribution information and historical hotspot area information corresponding to historical data collection; Based on the historical temperature field distribution information, the historical hotspot area information, the temperature field distribution information, and the hotspot area information, determine the temperature magnitude change trend prediction information and the hotspot area expansion trend information; Based on the predicted temperature magnitude change trend information, the hotspot area expansion trend information, and the temperature field distribution information, the temperature anomaly level of this data collection is determined; Based on the temperature anomaly level collected this time, the predicted temperature change trend information, the hot spot area expansion trend information, and the temperature field distribution information, a temperature warning information for the power station cabinet equipment is generated.

4. The method according to claim 1, characterized in that, The hotspot information includes the hotspot's spatial location, area, and temperature value.

5. The method according to claim 1, characterized in that, The temperature warning information includes the severity of the current temperature anomaly, the location of the abnormal area, the temperature value, and the predicted temperature change trend.

6. A temperature early warning device for power station cabinet equipment, characterized in that, The device includes: The data acquisition parameter determination module is used to determine the acquisition parameters of the lowest temperature anomaly level as the acquisition parameters of the current acquisition when the current acquisition is the first acquisition; and to determine the acquisition parameters of the temperature anomaly level determined in the previous acquisition as the acquisition parameters of the current acquisition when the current acquisition is not the first acquisition. The image acquisition module is used to acquire dual-light images of the power station cabinet equipment to be analyzed according to the acquisition parameters corresponding to this acquisition; the dual-light images include infrared thermal imaging images and visible light images. An image fusion module is used to preprocess the two-light image to obtain a preprocessed two-light image. The preprocessing includes at least quality assessment processing, image enhancement processing, and region segmentation processing. Multi-scale local features and regional features are extracted from the preprocessed two-light image to construct feature description information. Based on the region segmentation result of the two-light image, initial registration points are selected from the feature points corresponding to the feature description information. The region segmentation result is obtained through the region segmentation processing. Based on the feature description information, candidate registration points are selected from the initial registration points. A random sampling consistency model is used to eliminate erroneous registration points among the candidate registration points to obtain the target registration. Based on the target registration points, the affine transformation matrix of the infrared thermal imaging image relative to the visible light image is calculated and determined; based on the affine transformation matrix and the infrared thermal imaging image, a registered infrared thermal imaging image is obtained; using the edge information in the visible light image, an edge map of the visible light image is obtained, and a temperature gradient map of the registered infrared thermal imaging image is obtained by calculating the temperature gradient of the registered infrared thermal imaging image; based on the edge map and the temperature gradient map, the fusion weight of the two-light image is determined; based on the fusion weight, the registered infrared thermal imaging image and the visible light image are fused to obtain a fused image; The temperature detection module is used to perform temperature field reconstruction processing and hotspot detection processing on the fused image to obtain temperature field distribution information and hotspot region information. The temperature field distribution information is determined based on the initial temperature field distribution information, the temperature gradient map of the registered infrared thermal imaging image, the fused image, the preset reference temperature, and the preset importance weight of the power station cabinet equipment area. The initial temperature field distribution information is obtained based on the registered infrared thermal imaging image. The temperature analysis module is used to perform temperature analysis on the power station cabinet equipment based on the temperature field distribution information and the hot spot area information, and to obtain temperature warning information for the power station cabinet equipment.

7. The apparatus according to claim 6, characterized in that, The temperature detection module is further configured to obtain initial temperature field distribution information based on the registered infrared thermal imaging image; and to obtain the temperature field distribution information based on the initial temperature field distribution information, the temperature gradient map, the fused image, the preset reference temperature, and the preset importance weight of the power station cabinet equipment area. Based on the temperature field distribution information and the region segmentation results of the dual-light image, a threshold judgment method is used to determine the hot spot region information; The region segmentation result is obtained through the region segmentation process.

8. The apparatus according to claim 6, characterized in that, The temperature analysis module is further configured to acquire historical temperature field distribution information and historical hotspot area information corresponding to historical data collection; determine temperature magnitude change trend prediction information and hotspot area expansion trend information based on the historical temperature field distribution information, the historical hotspot area information, the temperature field distribution information, and the hotspot area information; determine the temperature anomaly level of the current data collection based on the temperature magnitude change trend prediction information, the hotspot area expansion trend information, and the temperature field distribution information; and generate temperature warning information for the power station cabinet equipment based on the temperature anomaly level of the current data collection, the temperature magnitude change trend prediction information, the hotspot area expansion trend information, and the temperature field distribution information.

9. The apparatus according to claim 6, characterized in that, The hotspot information includes the hotspot's spatial location, area, and temperature value.

10. The apparatus according to claim 6, characterized in that, The temperature warning information includes the severity of the current temperature anomaly, the location of the abnormal area, the temperature value, and the predicted temperature change trend.

11. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 5.

12. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 5.

13. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 5.

Images