A method and apparatus for locating temperature gradient anomalies in power equipment.

Abstract

This invention discloses a method and apparatus for locating temperature gradient anomalies in power equipment, comprising: acquiring visible light and infrared images of the power equipment; fusing the visible light and infrared images to obtain a fused image; dividing the fused image into pixel regions of the target equipment and pixel regions of non-target equipment, and obtaining a temperature matrix corresponding to the pixel regions of the target equipment based on the mapping relationship between pixels and temperature values; analyzing the temperature matrix using spatial temperature gradients to obtain a spatial temperature gradient distribution corresponding to the temperature matrix, and obtaining pixels with abnormal temperature gradients in the target region based on the spatial temperature gradient distribution; and spatially locating the location of the temperature anomaly in the power equipment based on the pixels with abnormal temperature gradients in the target region. This addresses the problem of low target recognition capability in existing infrared temperature measurement technologies.

Description

Technical Field

[0001] This invention relates to the field of temperature anomaly diagnosis and location in power equipment, and specifically to a location method and apparatus for diagnosing temperature gradient anomalies in power equipment. Background Technology

[0002] With continuous economic development, the demand for electricity from all sectors is constantly increasing, making electricity one of the fastest-growing sectors in the energy industry. Compared to 2011, the number of substations / converter stations operated by the State Grid's 27 provincial companies more than doubled in 2021, reaching over 40,000. As the core link in the safe and stable operation of the power grid, the stability of main power grid equipment is becoming increasingly prominent with the integration of more and more new energy sources, requiring timely monitoring of the status of more and more power equipment. Routine inspections are a crucial link in ensuring the safe and stable operation of main power grid equipment, but the number of power grid maintenance personnel has not increased in recent years. This combination of increasing power equipment and a shortage of maintenance personnel presents both significant challenges and opportunities for transformation in the maintenance of main power grid equipment. Artificial intelligence technology and new sensing technologies can improve the intelligence level of power grid equipment inspection.

[0003] Currently, infrared thermal imaging technology is used for power equipment inspection. Infrared thermal imaging technology uses photoelectric technology to detect the infrared signal of thermal radiation from an object in a specific band, converting this signal into salient images and graphics that can be distinguished by human vision. Infrared thermal imaging technology is a primary daily inspection method for detecting overheating defects in equipment, but it has the following problems:

[0004] 1) Infrared images have difficulty distinguishing between the subject and the background.

[0005] Infrared images reflect the temperature of the object being photographed, but lack information about its edges and textures. When the background of the object is complex, it is difficult to distinguish the subject from the background using infrared images. Therefore, a high degree of precision is required when shooting from the right angle. However, for densely populated substations where robots are used for automated inspections, it is difficult to achieve a clean background simply by changing the angle.

[0006] 2) The standard shape of infrared spectra makes it difficult to obtain the temperature field of the target body.

[0007] Using fixed-point and preset-position temperature measurement methods such as point temperature measurement, line temperature measurement, rectangular frame temperature measurement, and polygonal frame temperature measurement is difficult to effectively obtain the temperature field of the target surface. Point and line temperature measurement cannot reflect the complete temperature field of the target surface; due to the irregular shape of the target, rectangular frame temperature measurement may result in the rectangle being too large or too small, failing to reflect the complete temperature field of the target surface or reflecting the background temperature of the target, leading to missed and false detections in temperature analysis; using polygonal frame temperature measurement, since the polygonal frame is marked on the lens imaging area, there are still significant mechanical positional errors due to the movement of the camera gimbal or robot. Therefore, after long-term operation, polygonal frame temperature measurement also suffers from polygonal frame drift, leading to false detections in temperature analysis.

[0008] 3) Lack of intelligent analysis methods for temperature field differences

[0009] Currently, the basis for determining the surface temperature of equipment is mainly the "DL / T 664 Infrared Diagnostic Application Specification for Live Equipment". However, this specification lacks detailed consideration of regional differences, equipment differences, and operating condition differences. The relatively fixed judgment standard is difficult to adapt to the needs of differentiated operation and maintenance of new power system equipment. There is an urgent need for more intelligent and proactive differentiated analysis methods for equipment surface temperature field analysis. Summary of the Invention

[0010] During the operation of substations / converter stations, maintenance personnel need to conduct regular inspections of the equipment within the station to ensure that the facilities are operating normally. This invention proposes a method and apparatus for diagnosing temperature gradient anomalies in power equipment, aiming to solve the following problems: 1) the difficulty in distinguishing the target subject from the background in infrared images; 2) the difficulty in obtaining the temperature field of the target subject due to the standard shape of infrared spectra; and 3) the lack of intelligent analysis methods for temperature field differentiation.

[0011] This invention provides a method for locating temperature gradient anomalies in power equipment, comprising:

[0012] Acquire visible light and infrared images of power equipment;

[0013] The visible light band image and the infrared band image are fused to obtain a fused image;

[0014] The fused image is divided into pixel regions of the target device and pixel regions of non-target device. Based on the mapping relationship between pixels and temperature values, the temperature matrix corresponding to the pixel regions of the target device is obtained.

[0015] Spatial temperature gradient analysis is used to analyze the temperature matrix to obtain the spatial temperature gradient distribution corresponding to the temperature matrix. Based on the spatial temperature gradient distribution, pixels with abnormal temperature gradients in the target area are obtained.

[0016] Based on the pixels showing abnormal temperature gradients in the target area, the location of the temperature anomaly in the power equipment is spatially located.

[0017] Furthermore, visible light and infrared images of the power equipment are acquired, including:

[0018] Visible light band image (V) and infrared band image (I) of power equipment are acquired using image acquisition equipment.

[0019] Furthermore, after the steps of acquiring visible light and infrared images of the power equipment, the process also includes:

[0020] The visible light band image V of the power equipment is preprocessed to remove the influence of different light intensities on the visible light band imaging of the power equipment, and the preprocessed visible light band image V is obtained.

[0021] Furthermore, it also includes:

[0022] The infrared band image I and the visible light band image V are subjected to high-frequency filtering, grayscale conversion, and edge extraction.

[0023] Feature points of the infrared band image I and the visible light band image V are extracted using a fully affine invariant image feature matching algorithm.

[0024] The feature points of the infrared band image I and the visible light band image V are matched using a quick nearest search method.

[0025] The registration of the outer band image I and the visible light band image V is achieved by removing mismatched points through a random sampling consistency algorithm.

[0026] Furthermore, the visible light band image and the infrared band image are fused to obtain a fused image, including:

[0027] The low-frequency and high-frequency subband coefficients of the outer band image I and the visible light band image V are obtained by non-downsampling contour wave transform.

[0028] The low-frequency and high-frequency subband coefficients are subjected to guided filtering to obtain the fused low-frequency subband coefficients Fl and FH;

[0029] Based on the low-frequency subband coefficients Fl and FH, a preliminary fused image F1 is obtained through inverse transformation;

[0030] The preliminary fused image F1 and the infrared band image V are fed into a pre-constructed convolutional neural network, which outputs the final fused image F.

[0031] Furthermore, the fused image is divided into pixel regions of the target device and pixel regions of non-target device, including:

[0032] The fused image F is segmented into instances, and according to the semantics of the pixels in the fused image, the fused image is divided into the pixel region Zt of the target device and the pixel region Zu of the non-target device.

[0033] This invention also provides a location device for diagnosing temperature gradient anomalies in power equipment, comprising:

[0034] The image acquisition unit is used to acquire visible light and infrared images of power equipment.

[0035] The fusion unit is used to fuse the visible light band image and the infrared band image to obtain a fused image;

[0036] The temperature matrix acquisition unit is used to divide the fused image into pixel regions of the target device and pixel regions of non-target device, and obtain the temperature matrix corresponding to the pixel regions of the target device according to the mapping relationship between pixels and temperature values.

[0037] The gradient distribution acquisition unit is used to analyze the temperature matrix using spatial temperature gradient to obtain the spatial temperature gradient distribution corresponding to the temperature matrix, and to acquire the pixels with abnormal temperature gradients in the target area based on the spatial temperature gradient distribution.

[0038] The spatial positioning unit is used to spatially locate the location of the temperature anomaly in the power equipment based on the pixels with abnormal temperature gradients in the target area.

[0039] Furthermore, it also includes:

[0040] The preprocessing unit is used to preprocess the visible light band image V of the power equipment, remove the influence of different light intensities on the visible light band imaging of the power equipment, and obtain the preprocessed visible light band image V.

[0041] Furthermore, it also includes:

[0042] The image processing unit is used to perform high-frequency filtering, grayscale conversion, and edge extraction on the infrared band image I and the visible light band image V;

[0043] The feature point extraction unit is used to extract feature points of the infrared band image I and the visible light band image V using a fully affine invariant image feature matching algorithm.

[0044] The feature point matching unit matches the feature points of the infrared band image I and the visible light band image V using a quick nearest search method.

[0045] The registration unit is used to remove mismatched points through a random sampling consensus algorithm to achieve the registration of the outer band image I and the visible light band image V.

[0046] Furthermore, the fusion unit includes:

[0047] The coefficient acquisition subunit is used to obtain the low-frequency and high-frequency sub-band coefficients of the outer band image I and the visible light band image V through non-downsampled contour wave transform.

[0048] The filtering subunit is used to perform guided filtering on the low-frequency and high-frequency subband coefficients to obtain the fused low-frequency subband coefficients Fl and FH.

[0049] The preliminary fused image acquisition subunit is used to obtain the preliminary fused image F1 by inverse transformation based on the low-frequency subband coefficients Fl and FH;

[0050] The fused image output subunit is used to feed the preliminary fused image F1 and the infrared band image V into a pre-constructed convolutional neural network, and output the final fused image F through the convolutional neural network.

[0051] The present invention provides a method and apparatus for diagnosing temperature gradient anomalies in power equipment, which solves the problem of low target recognition capability in existing infrared thermometry technologies. Attached Figure Description

[0052] Figure 1 This is a flowchart illustrating a method for locating abnormal temperature gradients in power equipment, provided in an embodiment of the present invention.

[0053] Figure 2 This is a schematic diagram of the composition of the positioning system for diagnosing temperature gradient anomalies in power equipment according to an embodiment of the present invention;

[0054] Figure 3 This is a schematic diagram of the process for removing the effects of illumination from visible light band images of power equipment according to an embodiment of the present invention;

[0055] Figure 4 This is a schematic diagram of the process for registering infrared and visible light band images of power equipment according to an embodiment of the present invention;

[0056] Figure 5 This is a schematic diagram of the process of infrared and visible light band image fusion for power equipment according to an embodiment of the present invention;

[0057] Figure 6 This is a schematic diagram of the process for segmenting the target region of an infrared and visible light band image of power equipment according to an embodiment of the present invention;

[0058] Figure 7 This is a schematic diagram of the process for spatially locating abnormal temperature gradients in power equipment according to an embodiment of the present invention;

[0059] Figure 8 This is a schematic diagram of a positioning device for diagnosing temperature gradient anomalies in power equipment, provided in an embodiment of the present invention. Detailed Implementation

[0060] Numerous specific details are set forth in the following description to provide a full understanding of the invention. However, the invention can be practiced in many other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0061] Figure 1 This is a flowchart illustrating a method for locating abnormal temperature gradients in power equipment, provided by an embodiment of the present invention. The method provided in the first embodiment of the present invention will be described in detail.

[0062] Step S101: Acquire visible light and infrared images of the power equipment.

[0063] Visible light band image (V) and infrared band image (I) of power equipment are acquired using image acquisition equipment.

[0064] The visible light band image V of the power equipment is preprocessed to remove the influence of different light intensities on the visible light band imaging of the power equipment, and the preprocessed visible light band image V is obtained.

[0065] The infrared band image I and the visible light band image V are subjected to high-frequency filtering, grayscale conversion, and edge extraction; feature points of the infrared band image I and the visible light band image V are extracted using a fully affine invariant image feature matching algorithm; the feature points of the infrared band image I and the visible light band image V are matched using a quick nearest search method; and mismatched points are removed using a random sampling consensus algorithm to achieve registration of the infrared band image I and the visible light band image V.

[0066] Step S102: The visible light band image and the infrared band image are fused to obtain a fused image.

[0067] The low-frequency and high-frequency subband coefficients of the outer band image I and the visible light band image V are obtained by non-downsampling contour wave transform; the low-frequency and high-frequency subband coefficients are subjected to guided filtering to obtain the fused low-frequency subband coefficients Fl and FH; based on the low-frequency subband coefficients Fl and FH, a preliminary fused image F1 is obtained by inverse transform; the preliminary fused image F1 and the infrared band image V are fed into a pre-constructed convolutional neural network, and the final fused image F is output through the convolutional neural network.

[0068] Step S103: Divide the fused image into pixel regions of the target device and pixel regions of non-target device, and obtain the temperature matrix corresponding to the pixel regions of the target device according to the mapping relationship between pixels and temperature values.

[0069] The fused image F is segmented into pixel regions Zt (target device) and Zu (non-target device) according to the semantics of the pixels. The temperature matrix T corresponding to the pixel region of the target device is obtained based on the mapping relationship between pixels and temperature values.

[0070] Step S104: Use spatial temperature gradient analysis to analyze the temperature matrix to obtain the spatial temperature gradient distribution corresponding to the temperature matrix, and obtain the pixels with abnormal temperature gradients in the target area based on the spatial temperature gradient distribution.

[0071] For the temperature matrix T of the target area, spatial temperature gradient analysis is used to obtain the spatial temperature gradient distribution Ds. Based on the spatial temperature gradient distribution, the pixels with abnormal temperature gradients in the target area are obtained.

[0072] Step S105: Based on the pixels showing abnormal temperature gradients in the target area, spatially locate the location of the temperature anomaly in the power equipment.

[0073] Spatial location and proactive warning are performed based on pixels with abnormal temperature gradients in the target area.

[0074] Based on the same inventive concept, this invention also provides a location system for diagnosing temperature gradient anomalies in power equipment, such as... Figure 2 As shown, it includes:

[0075] Visible light image acquisition module, infrared image acquisition module, binocular camera integrated housing, gimbal, edge computing module, airborne communication module, station communication module, and station analysis and computing module;

[0076] The visible light image acquisition module and the infrared image acquisition module are encapsulated in an integrated housing of a binocular camera, and are used to simultaneously acquire visible spectral images and infrared spectral images of substations / converter stations;

[0077] The base of the gimbal can be fixed to the mobile base of the patrol robot or to the top or vertical surface of a wall or column.

[0078] The edge computing module is used for real-time pre-analysis of visible and infrared spectral images. It performs fast edge calculations to address issues such as the influence of light, defocusing, and missing targets during the acquisition process of the visible and infrared image acquisition modules, and issues automatic adjustment commands to the visible and infrared image acquisition modules and the gimbal.

[0079] The airborne communication module is used to transmit visible spectrum image and infrared spectrum image data to the station communication module.

[0080] The station-end communication module is used to receive visible spectral image data and infrared spectral image data.

[0081] The analysis and calculation module is used for image registration, image fusion, target region extraction, and temperature field analysis of visible and infrared spectral images, providing the spatial location of temperature field anomalies and proactive early warning.

[0082] Exemplary embodiments of the invention will now be described with reference to the accompanying drawings. However, the invention may be embodied in many different forms and is not limited to the embodiments described herein. These embodiments are provided to fully and completely disclose the invention and to fully convey its scope to those skilled in the art. The terminology used in the exemplary embodiments illustrated in the drawings is not intended to limit the invention. In the drawings, the same units / elements are referred to by the same reference numerals.

[0083] The visible light image V and infrared image I of the power equipment are acquired simultaneously through the visible light and infrared image acquisition modules; specifically, such as Figure 2 As shown, the edge computing module of the system sends synchronous acquisition commands to the visible light image acquisition module and the infrared image acquisition module. The visible light image acquisition module acquires electromagnetic waves with wavelengths of 380nm-700nm to reflect the shape outline, appearance color and surface texture of the power equipment; the infrared image acquisition module acquires electromagnetic waves with wavelengths of 7μm-14μm to reflect the appearance temperature of the power equipment. The acquired visible light band images and infrared band images are transmitted to the edge computing module.

[0084] Preprocessing is performed on the visible light image I of the power equipment to remove the influence of different light intensities on the visible light imaging of the power equipment, resulting in a preprocessed visible light image V; the specific process is as follows. Figure 3 As shown, the process by which a visible light image acquisition module captures information is determined by two factors: the reflective properties of the object itself and the intensity of light around the object. Light intensity determines the dynamic range of all pixels in the original image, while the inherent properties of the original image are determined by the object's reflectance coefficient. By removing the influence of light, the inherent properties of the object are preserved.

[0085] The image expression for the visible light band is:

[0086] I(x,y)=L(x,y)×V(x,y) (1)

[0087] Where I(x,y) represents the image signal received by the human eye or camera; L(x,y) represents the ambient light intensity information component, i.e. the incident light component; V(x,y) represents the inherent property information component of the object itself; and (x,y) represents the coordinates of any pixel in the image.

[0088] Taking the logarithm of both sides of equation (1) yields:

[0089] log[I(x,y)]=log[L(x,y)×V(x,y)]=log[L(x,y)]+log[V(x,y)] (2)

[0090] Let i(x,y)=log[I(x,y)], v(x,y)=log[V(x,y)], l(x,y)=log[L(x,y)], then:

[0091] i(x,y)=l(x,y)+v(x,y) (3)

[0092] Where i(x,y) represents the logarithmic component of the image information captured by the human eye or camera; l(x,y) represents the logarithmic component of the ambient light intensity information; and v(x,y) represents the logarithmic component of the inherent properties of the object itself. The human eye's perception of brightness is not linear, but closer to a logarithmic curve, and complex multiplication and division are transformed into simple addition and subtraction in the logarithmic domain. Therefore, converting to logarithms can significantly reduce the complexity of the algorithm.

[0093] Gaussian filtering, grayscale conversion, and edge extraction are performed on the infrared image I and the processed visible light image V, respectively. Feature points are extracted from both the processed infrared and visible light images. Feature point mapping between the infrared and visible light images is achieved through fast feature point matching. Mismatched points are removed using a random sampling consensus algorithm, thus achieving registration between the infrared image I and the visible light image V. Specifically, as follows... Figure 4 As shown, filtering, grayscale conversion, and edge extraction are all standard image operations. The filtering method used is Gaussian filtering, a linear smoothing filter suitable for eliminating noise in images.

[0094] Specifically, the grayscale conversion mentioned above uses mean grayscale conversion, which converts the mean of the RGB three-channel information into single-channel grayscale information, thereby speeding up information processing.

[0095] Specifically, the edge extraction method used is Canny edge extraction, which is a first-order differential operator detection algorithm. Canny edge extraction uses non-maximum suppression to effectively suppress multi-response edges, and uses dual thresholds to effectively reduce the false negative rate of edges.

[0096] Specifically, the feature point extraction method employed is a fully affine-invariant image feature matching feature point extraction method. Since the optical axis directions of the infrared acquisition module and the visible light acquisition module are inconsistent, the fully affine-invariant image feature matching feature points achieve complete affine invariance by simulating longitude and latitude.

[0097] Specifically, the feature point matching employs a fast nearest neighbor search method. To exclude key points with no matching relationship due to image occlusion and background clutter, the matching method compares nearest neighbor distance and second nearest neighbor distance: A key point in the infrared image is selected, and its two closest Euclidean distances in the visible light image are found. If the ratio (ratio) of the nearest distance divided by the second nearest distance is less than a threshold, the pair of matching points is accepted. Matching numerous images with arbitrary scale, rotation, and brightness variations shows that a ratio value between 0.4 and 0.6 is optimal. Values ​​less than 0.4 rarely match, while values ​​greater than 0.6 contain a large number of incorrect matches. The recommended principle for determining the ratio is as follows:

[0098] ratio = 0.4: For matching with high accuracy requirements;

[0099] ratio = 0.6: For matching where a relatively large number of matching points are required;

[0100] ratio = 0.5: General case.

[0101] Specifically, the mismatch elimination method employs random sampling consensus, which effectively fits the fitting function under noisy models, dividing points into "inside points" (correctly matched points) and "outside points" (mismatched points). In a dataset containing "outside points," the optimal parameter model is iteratively searched, and points that do not conform to the optimal model are defined as "outside points."

[0102] The registered visible light image V and infrared band image I are transformed using non-downsampled contour wave transformation to obtain the low-frequency and high-frequency subband coefficients of the visible light band image and infrared band image, respectively. Guided filtering is then applied to the low-frequency and high-frequency subband coefficients of the visible light band image and infrared band image to obtain the fused low-frequency subband coefficients Fl and FH. The preliminary fused image F1 is obtained through inverse transformation. The final fused image F is obtained by feeding the preliminary fused image and the infrared band image into a convolutional neural network.

[0103] Specifically, such as Figure 5 As shown, the steps for fusing visible light images and infrared images are as follows:

[0104] (1) The infrared band image I and the visible light image V are decomposed by non-downsampling contour wave transform to obtain the low-frequency subband coefficients and high-frequency subband coefficients of the infrared band image and the low-frequency subband coefficients and high-frequency subband coefficients of the visible light image, respectively.

[0105] (2) The low-frequency subband coefficients of the infrared image and the low-frequency subband coefficients of the visible light image are fused by guided filtering to obtain a new low-frequency subband coefficient FL.

[0106] (3) The high-frequency subband coefficients of the infrared image and the high-frequency subband coefficients of the visible light image are fused by guided filtering to obtain a new high-frequency subband coefficient FH.

[0107] (4) The new low-frequency subband coefficients FL and high-frequency subband coefficients FH are transformed by non-downsampling contour wave inverse transformation to obtain the initial fused image F1.

[0108] (5) Feed the infrared image I and the initial fused image F1 into the convolutional neural network.

[0109] (6) The final fused image F is obtained by reconstructing the Laplace pyramid.

[0110] The fused image F is segmented into instances, and according to the semantics of the pixels in the fused image, it is divided into the pixel region Zt of the target device and the pixel region Zu of the non-target device. The temperature matrix T corresponding to the pixel region of the target device is obtained according to the mapping relationship between pixels and temperature values.

[0111] Specifically, such as Figure 6 As shown, the instance segmentation steps are as follows:

[0112] (1) Input the fused image and input it into a pre-trained neural network to obtain the corresponding feature map;

[0113] (2) Set a predetermined number of regions of interest for each point in this feature map to obtain multiple candidate regions of interest;

[0114] (3) These candidate regions of interest are fed into the region generation network for binary classification (foreground or background) and bounding box regression to filter out some candidate regions of interest.

[0115] (4) Perform region of interest alignment on these remaining regions of interest;

[0116] (5) Classify these regions of interest, perform border regression and instance segmentation to generate pixel regions.

[0117] For the temperature matrix T of the target area, spatial temperature gradient analysis is used to obtain the spatial temperature gradient distribution Ds, and spatial positioning and active early warning are performed based on the pixels with abnormal temperature gradients in the target area.

[0118] Specifically, such as Figure 7 As shown in (a), the temperature gradient of adjacent temperature matrices is calculated from the temperature matrix T of the target area, and the spatial temperature gradient distribution Ds is obtained. Based on the spatial location of the acquired images, combined with equipment component ledger information; as Figure 7 As shown in (b), spatial positioning is performed based on pixels with abnormal temperature gradient distribution. Compared with defect identification based solely on images, spatial defect analysis based on the spatial location of equipment components has higher accuracy and lower false detection rate. When the threshold specified in the operation and maintenance guidelines is exceeded, the defect level is determined and an active warning is given based on the spatial location of the collected images and the ledger information, which greatly reduces the workload of manual collection and analysis.

[0119] Based on the same inventive concept, the present invention also provides a positioning device 800 for diagnosing temperature gradient anomalies in power equipment, such as... Figure 8 As shown, it includes:

[0120] Image acquisition unit 810 is used to acquire visible light and infrared images of power equipment;

[0121] The fusion unit 820 is used to fuse the visible light band image and the infrared band image to obtain a fused image;

[0122] Temperature matrix acquisition unit 830 is used to divide the fused image into pixel regions of the target device and pixel regions of non-target device, and obtain the temperature matrix corresponding to the pixel regions of the target device according to the mapping relationship between pixels and temperature values.

[0123] The gradient distribution acquisition unit 840 is used to analyze the temperature matrix using spatial temperature gradient to obtain the spatial temperature gradient distribution corresponding to the temperature matrix, and to acquire the pixels with abnormal temperature gradients in the target area based on the spatial temperature gradient distribution.

[0124] The spatial positioning unit 85010 is used to spatially locate the location of the temperature anomaly of the power equipment based on the pixels of the abnormal temperature gradient in the target area.

[0125] Furthermore, it also includes:

[0126] The preprocessing unit is used to preprocess the visible light band image V of the power equipment, remove the influence of different light intensities on the visible light band imaging of the power equipment, and obtain the preprocessed visible light band image V.

[0127] Furthermore, it also includes:

[0128] The image processing unit is used to perform high-frequency filtering, grayscale conversion, and edge extraction on the infrared band image I and the visible light band image V;

[0129] The feature point extraction unit is used to extract feature points of the infrared band image I and the visible light band image V using a fully affine invariant image feature matching algorithm.

[0130] The feature point matching unit matches the feature points of the infrared band image I and the visible light band image V using a quick nearest search method.

[0131] The registration unit is used to remove mismatched points through a random sampling consensus algorithm to achieve the registration of the outer band image I and the visible light band image V.

[0132] Furthermore, the fusion unit includes:

[0133] The coefficient acquisition subunit is used to obtain the low-frequency and high-frequency sub-band coefficients of the outer band image I and the visible light band image V through non-downsampled contour wave transform.

[0134] The filtering subunit is used to perform guided filtering on the low-frequency and high-frequency subband coefficients to obtain the fused low-frequency subband coefficients Fl and FH.

[0135] The preliminary fused image acquisition subunit is used to obtain the preliminary fused image F1 by inverse transformation based on the low-frequency subband coefficients Fl and FH;

[0136] The fused image output subunit is used to feed the preliminary fused image F1 and the infrared band image V into a pre-constructed convolutional neural network, and output the final fused image F through the convolutional neural network.

[0137] This invention provides a method and apparatus for locating temperature gradient anomalies in power equipment. It involves registering and fusing infrared and visible light images acquired by a binocular pan-tilt unit, segmenting the target region using the fused image, and then performing temporal and spatial temperature gradient analysis on the segmented target region's temperature field using a temperature gradient algorithm. This enables spatial localization of abnormal temperature conditions in the target region. This invention addresses the problems of: 1) difficulty in distinguishing the target from the background in infrared images; 2) difficulty in obtaining the target's temperature field due to the standard shape of infrared spectra; and 3) lack of intelligent analysis methods for temperature field differentiation. Compared to traditional infrared thermometry methods, this method improves the ability to identify infrared targets, extract the temperature field of the target region, and intelligently identify anomalies. This method and system can be implemented with robots, effectively improving the efficiency and accuracy of infrared defect detection in power systems and playing a crucial role in ensuring the reliable operation of new power system equipment.

[0138] Meanwhile, this invention effectively utilizes the enhanced role of infrared thermometry in power equipment inspection. Firstly, infrared thermometry is primarily used for live-line testing of power equipment. Power equipment maintenance personnel use handheld portable infrared thermometers to inspect the external temperature of power equipment, detect overheating defects, and address them promptly. Improving the defect detection capability of infrared thermometry is a crucial means of ensuring the inherent safety of equipment. Secondly, maintenance personnel typically spend no less than two hours conducting live-line infrared testing of an entire substation. With more and more substations adopting unmanned operation and maintenance methods, the commuting time and costs for maintenance personnel between multiple stations are high. Statistics show that compared to over 20,000 operating substations / converter stations nationwide in 2012, this number has exceeded 40,000 in 2022, but the number of substation maintenance personnel has decreased rather than increased. Traditional manual live-line testing methods are no longer sufficient to meet the growing demand for live-line maintenance of substation equipment, necessitating the use of digital methods to improve the efficiency of infrared detection and analysis. This invention provides a temperature gradient anomaly diagnosis and location system for power equipment with enhanced infrared and visible light band image information. It achieves high-precision and high-efficiency identification, spatial location, and proactive early warning of temperature defects in substations / converter stations. By fusing visible and infrared spectral images, it enables temperature anomaly diagnosis based on the fused images, improving the efficiency of daily inspections by substation / converter station maintenance personnel, increasing the reliability of status awareness and defect early warning, reducing the workload of maintenance personnel, and enhancing their work safety.

[0139] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0140] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0141] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0142] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0143] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the scope of the claims of the present invention.

Claims

1. A positioning method for temperature gradient anomaly diagnosis of a power device, characterized by, include: Acquire visible light and infrared images of power equipment; The visible light band image and the infrared band image are fused to obtain a fused image; The fused image is divided into pixel regions of the target device and pixel regions of non-target device. Based on the mapping relationship between pixels and temperature values, the temperature matrix corresponding to the pixel regions of the target device is obtained. The temperature matrix is ​​analyzed using spatial temperature gradient to obtain the spatial temperature gradient distribution corresponding to the temperature matrix. Based on the spatial temperature gradient distribution, the pixels with abnormal temperature gradients in the target area are obtained. Based on the pixels showing abnormal temperature gradients in the target area, the location of the temperature anomaly in the power equipment is spatially located. The visible light band image and the infrared band image are fused to obtain a fused image, including: The low-frequency and high-frequency subband coefficients of the infrared band image I and the visible light band image V are obtained by non-downsampling contour wave transform. Guided filtering is performed on the low-frequency and high-frequency subband coefficients to obtain the low-frequency subband coefficient Fl of the fused infrared band image and the low-frequency subband coefficient FH of the visible light band image; Based on the low-frequency subband coefficient Fl of the infrared band image and the low-frequency subband coefficient FH of the visible light band image, a preliminary fused image F1 is obtained through non-downsampling contour wave inverse transform; The preliminary fused image F1 and the infrared band image I are fed into a pre-constructed convolutional neural network, which outputs the final fused image F.

2. The method of claim 1, wherein, Acquire visible light and infrared images of power equipment, including: Visible light band image (V) and infrared band image (I) of power equipment are acquired using image acquisition equipment.

3. The method of claim 1, wherein, Following the steps of acquiring visible light and infrared images of power equipment, the process also includes: The visible light band image V of the power equipment is preprocessed to remove the influence of different light intensities on the visible light band imaging of the power equipment, and the preprocessed visible light band image V is obtained.

4. The method according to claim 1 or 3, characterized in that, Following the steps of acquiring visible light and infrared images of power equipment, the process also includes: Gaussian filtering, grayscale conversion, and edge extraction are performed on the infrared band image I and the visible light band image V; Feature points of the infrared band image I and the visible light band image V are extracted using a fully affine invariant image feature matching algorithm. The feature points of the infrared band image I and the visible light band image V are matched using a quick nearest search method; The infrared band image I and the visible light band image V are registered by removing mismatched points through a random sampling consistency algorithm.

5. The method according to claim 1, characterized in that, The fused image is divided into pixel regions of the target device and pixel regions of non-target device, including: The fused image F is segmented into instances, and according to the semantics of the pixels in the fused image, the fused image is divided into the pixel region Zt of the target device and the pixel region Zu of the non-target device.

6. A location device for diagnosing temperature gradient anomalies in power equipment, characterized in that, include: The image acquisition unit is used to acquire visible light and infrared images of power equipment. The fusion unit is used to fuse the visible light band image and the infrared band image to obtain a fused image; The temperature matrix acquisition unit is used to divide the fused image into pixel regions of the target device and pixel regions of non-target device, and obtain the temperature matrix corresponding to the pixel regions of the target device according to the mapping relationship between pixels and temperature values. The gradient distribution acquisition unit is used to analyze the temperature matrix using spatial temperature gradient to obtain the spatial temperature gradient distribution corresponding to the temperature matrix, and to acquire the pixels with abnormal temperature gradients in the target area based on the spatial temperature gradient distribution. A spatial positioning unit is used to spatially locate the location of the temperature anomaly in the power equipment based on the pixels with abnormal temperature gradients in the target area. The fusion unit includes: The coefficient acquisition subunit is used to obtain the low-frequency and high-frequency sub-band coefficients of the infrared band image I and the visible light band image V through non-downsampled contour wave transformation. The filtering subunit is used to perform guided filtering on the low-frequency and high-frequency subband coefficients to obtain the low-frequency subband coefficient Fl of the fused infrared band image and the low-frequency subband coefficient FH of the visible light band image. The preliminary fused image acquisition subunit is used to obtain the preliminary fused image F1 by non-downsampling contour wave inverse transform based on the low-frequency subband coefficient Fl of the infrared band image and the low-frequency subband coefficient FH of the visible light band image. The fused image output subunit is used to feed the preliminary fused image F1 and the infrared band image I into a pre-constructed convolutional neural network, and output the final fused image F through the convolutional neural network.

7. The apparatus according to claim 6, characterized in that, Also includes: The preprocessing unit is used to preprocess the visible light band image V of the power equipment, remove the influence of different light intensities on the visible light band imaging of the power equipment, and obtain the preprocessed visible light band image V.

Images