A hotspot temperature rise infrared positioning method and system for a magnetizing transformer

Abstract

The present application relates to the technical field of image reconstruction, and particularly relates to a hotspot temperature rise infrared positioning method for a magnetizing transformer, comprising the following steps: constructing a three-dimensional reference surface model based on a plurality of reference image sets of the magnetizing transformer; constructing a three-dimensional reference surface temperature model and a three-dimensional real-time surface temperature model based on a plurality of reference temperature images and a plurality of real-time temperature images of the magnetizing transformer after starting; calculating a temperature difference between the three-dimensional real-time surface temperature model and the three-dimensional reference surface temperature model to obtain a three-dimensional surface temperature rise model; determining a surface abnormal temperature rise area based on the three-dimensional surface temperature rise model and a sampling interval; and generating a hotspot temperature rise infrared positioning result based on the surface abnormal temperature rise area. The present application can correct infrared temperature measurement deviation caused by observation angle and surface emissivity, construct a three-dimensional surface temperature rise distribution, and improve the accuracy of magnetizing transformer temperature rise hotspot positioning.

Description

Technical Field

[0001] This invention relates to the field of image reconstruction technology, and in particular to a method and system for infrared positioning of hot spot temperature rise in magnetized transformers. Background Technology

[0002] Permanent magnet materials and magnetic components are widely used in new energy vehicles, industrial motors, power generation equipment, and electronic products. Magnetizing transformers, as key equipment in the magnetization production of magnetic components, are essential for ensuring the manufacturing of related products. If localized temperature rise hotspots formed during their operation are not detected in time, they may lead to insulation aging, component damage, or equipment downtime. Therefore, it is necessary to detect and locate these temperature rise hotspots.

[0003] Current technology typically uses infrared thermal imagers to acquire temperature images of magnetized transformers and determines the locations of higher temperatures based on the temperature distribution in the images. However, different surfaces of a magnetized transformer face different orientations, requiring different observation angles for the infrared thermal imager to measure the temperature at each surface location. Furthermore, the surface materials of windings, core, terminals, and insulation materials have different emissivities. These differences in observation angle and surface emissivity affect the temperature measured by the infrared thermal imager, causing a deviation between the measured temperature in the temperature image and the actual temperature at the corresponding surface location, which may consequently affect the identification of hot spots.

[0004] Furthermore, infrared temperature images represent temperature distribution using two-dimensional pixel positions, making it difficult to directly characterize the three-dimensional spatial coordinates of temperature rise locations on the surface of a magnetizing transformer. Temperature images obtained from different acquisition perspectives also correspond to different image coordinate systems. Directly determining hotspots based on these images makes it difficult to unify temperature data from different acquisition perspectives onto the same three-dimensional surface and accurately determine the spatial location and coverage area of ​​the hotspots on the magnetizing transformer surface. This could lead to deviations in hotspot location determination, affecting subsequent fault location confirmation. Therefore, a method is needed to improve the accuracy of hotspot location in magnetizing transformers by correcting for infrared temperature measurement deviations caused by observation angles and surface emissivity, and by constructing a three-dimensional surface temperature rise distribution. Summary of the Invention

[0005] This invention provides an infrared method for locating hotspots in magnetized transformers and a computer-readable storage medium. Its main purpose is to improve the accuracy of hotspot location in magnetized transformers by correcting infrared temperature measurement deviations caused by observation angle and surface emissivity and constructing a three-dimensional surface temperature rise distribution.

[0006] To achieve the above objectives, the present invention provides an infrared positioning method for hot spot temperature rise in a magnetizing transformer, comprising:

[0007] Receive the hot spot temperature rise infrared positioning command, and confirm the magnetizing transformer based on the hot spot temperature rise infrared positioning command;

[0008] Based on multiple pre-constructed reference image acquisition perspectives, reference multi-view image acquisition is performed on the magnetizing transformer to obtain multiple reference image sets, which include: reference visible light images and reference depth images;

[0009] A three-dimensional reference surface model of the magnetized transformer was constructed based on multiple reference image sets;

[0010] Start the magnetizing transformer and perform multi-view reference temperature image acquisition on the magnetizing transformer after startup to obtain multiple reference temperature images;

[0011] A three-dimensional reference surface temperature model of the magnetized transformer was constructed based on multiple reference temperature images and a three-dimensional reference surface model.

[0012] Multi-view real-time temperature image acquisition was performed on the magnetizing transformer after startup to obtain multiple real-time temperature images. Based on the multiple real-time temperature images and the three-dimensional reference surface model, a three-dimensional real-time surface temperature model of the magnetizing transformer was constructed.

[0013] The temperature difference between the three-dimensional real-time surface temperature model and the three-dimensional reference surface temperature model is calculated to obtain the three-dimensional surface temperature rise model.

[0014] The sampling times of the multiple three-dimensional reference images and the sampling times of the multiple real-time image sets are obtained to obtain the reference sampling time and the real-time sampling time. The sampling interval is obtained by subtracting the reference sampling time from the real-time sampling time.

[0015] The abnormal surface temperature rise region is determined based on the three-dimensional surface temperature rise model and sampling interval, and the hot spot temperature rise infrared positioning result is generated based on the abnormal surface temperature rise region.

[0016] Optionally, the construction of a three-dimensional reference surface model of the magnetized transformer based on multiple reference image sets includes:

[0017] A target reference image set is obtained by sequentially extracting one reference image set from multiple reference image sets. The target reference image set includes a target reference visible light image and a target reference depth image.

[0018] Perform the following operations on the target reference image set:

[0019] Perform effective depth pixel filtering on the target reference depth image to obtain multiple effective depth pixels of the target;

[0020] For each of the multiple target effective depth pixels, perform the following operation:

[0021] Obtain the intrinsic parameters of the depth camera, the depth pixel coordinates and depth values ​​of the target's effective depth pixels;

[0022] Based on the intrinsic parameters of the depth camera, the coordinates of the depth pixels, and the depth value, spatial projection is performed on the effective depth pixels of the target to obtain the three-dimensional pixels of the target.

[0023] Pixel alignment is performed on the target reference depth image and the target reference visible light image to obtain the target aligned visible light image;

[0024] Extract visible light pixels with the same depth pixel coordinates from the target-aligned visible light image, and obtain the grayscale value of the visible light pixels to obtain the visible light grayscale value;

[0025] The target 3D pixel is assigned a gray value by using the visible light gray value as the gray value of the target 3D pixel, and the target reference point is obtained.

[0026] By summing up the target benchmark points, multiple target benchmark points are obtained;

[0027] An initial 3D point cloud model is constructed based on multiple target reference points;

[0028] By summing up the initial 3D point cloud models, multiple initial 3D point cloud models are obtained;

[0029] Point cloud fusion is performed on multiple initial 3D point cloud models to obtain a 3D reference surface model.

[0030] Optionally, the step of performing point cloud fusion on multiple initial 3D point cloud models to obtain a 3D reference surface model includes:

[0031] A sliding window extraction operation is performed on multiple initial 3D point cloud models to obtain a combination of multiple 3D point cloud models, wherein the combination of 3D point cloud models includes two initial 3D point cloud models.

[0032] For each of the multiple 3D point cloud model combinations, adjacent viewpoint pose calculations are performed to obtain multiple adjacent viewpoint transformation parameter sets, which include rotation matrices and translation vectors:

[0033] Extract one reference image acquisition view from multiple reference image acquisition viewpoints to obtain the target reference image acquisition viewpoint;

[0034] Perform a retention operation based on the elimination of target reference image acquisition viewpoints on multiple reference image acquisition viewpoints to obtain multiple retained image acquisition viewpoints;

[0035] Based on multiple adjacent viewpoint transformation parameters, the target viewpoint transformation parameters between each of the multiple preserved image acquisition viewpoints and the target reference image acquisition viewpoint are calculated respectively, thus obtaining multiple target viewpoint transformation parameters;

[0036] Based on multiple target viewpoint transformation parameters, point cloud merging is performed on multiple initial 3D point cloud models to obtain a 3D reference surface model.

[0037] Optionally, the adjacent viewpoint pose calculation is performed on each of the multiple 3D point cloud model combinations to obtain multiple adjacent viewpoint transformation parameter sets, including:

[0038] Extract one 3D point cloud model combination from multiple 3D point cloud model combinations sequentially to obtain the target 3D point cloud model combination, and perform the following operations on the target 3D point cloud model combination:

[0039] Extract the front-end 3D point cloud model and the back-end 3D point cloud model from the target 3D point cloud model combination;

[0040] Obtain the corresponding front reference visible light image and rear reference visible light image of the front 3D point cloud model and the rear 3D point cloud model;

[0041] Feature point extraction is performed on the front reference visible light image and the rear reference visible light image respectively to obtain the front feature point set and the rear feature point set;

[0042] Feature point matching is performed on the pre-set and post-set feature points to obtain multiple initial matching feature point pairs;

[0043] Multiple initial matching feature point pairs are filtered for matching errors to obtain multiple target matching point pairs;

[0044] Multiple preceding matching points and multiple following matching points are extracted from multiple target matching point pairs. Multiple preceding spatial coordinates of multiple preceding matching points and multiple following spatial coordinates of multiple following matching points are extracted from the preceding 3D point cloud model and the following 3D point cloud model, respectively.

[0045] Based on multiple front spatial coordinates and multiple back spatial coordinates, the rotation matrix and translation vector between the front 3D point cloud model and the back 3D point cloud model are calculated to obtain the adjacent viewpoint transformation parameter set.

[0046] By summing up the adjacent viewpoint transformation parameter sets, multiple adjacent viewpoint transformation parameter sets are obtained.

[0047] Optionally, based on multiple adjacent viewpoint transformation parameters, the target viewpoint transformation parameters between each of the multiple preserved image acquisition viewpoints and the target reference image acquisition viewpoint are calculated respectively, resulting in multiple target viewpoint transformation parameters, including:

[0048] Extract one preserved image acquisition viewpoint from multiple preserved image acquisition viewpoints sequentially to obtain the target preserved image acquisition viewpoint, and perform the following operations on the target preserved image acquisition viewpoint:

[0049] Determine the viewpoint sequence between the target preserved image acquisition viewpoint and the target reference image acquisition viewpoint;

[0050] Based on the viewpoint sequence, the adjacent viewpoint transformation parameter set located between the target reference image acquisition viewpoint and the target preserved image acquisition viewpoint is extracted from multiple adjacent viewpoint transformation parameter sets to obtain multiple target adjacent viewpoint transformation parameter sets. Among them, the target adjacent viewpoint transformation parameter set includes the target adjacent rotation matrix and the target adjacent translation vector.

[0051] Extract multiple target adjacent rotation matrices from a set of transformation parameters for multiple target adjacent viewpoints;

[0052] For each of the multiple target adjacent rotation matrices, perform the following operation:

[0053] Determine the target's adjacent translation vectors corresponding to the target's adjacent rotation matrices to obtain the corresponding adjacent translation vectors;

[0054] Construct a homogeneous transformation matrix based on the target's adjacent rotation matrices and corresponding adjacent translation vectors;

[0055] By summing the homogeneous transformation matrices, we obtain multiple homogeneous transformation matrices;

[0056] Based on the view sequence, matrix multiplication is performed on multiple homogeneous transformation matrices to obtain a multiplied transformation matrix. The target view rotation matrix and target view translation vector are extracted from the multiplied transformation matrix and denoted as the target view transformation parameters.

[0057] By summarizing the target view transformation parameters, multiple target view transformation parameters are obtained.

[0058] Optionally, the construction of the three-dimensional reference surface temperature model of the magnetizing transformer based on multiple reference temperature images and a three-dimensional reference surface model includes:

[0059] Extract a target reference temperature image from multiple reference temperature images sequentially, and then perform the following operations on the target reference temperature image:

[0060] Determine the reference depth image corresponding to the target reference temperature image to obtain the target reference depth image;

[0061] Pixel alignment is performed on the target reference depth image and the target reference temperature image to obtain the aligned reference temperature image;

[0062] By summing the alignment reference temperature images, multiple alignment reference temperature images are obtained;

[0063] Based on multiple aligned reference temperature images, temperature mapping is performed on a three-dimensional reference surface model to obtain a three-dimensional temperature model.

[0064] Temperature compensation correction is performed on the three-dimensional temperature model to obtain a three-dimensional reference surface temperature model.

[0065] Optionally, performing temperature compensation correction on the three-dimensional temperature model to obtain a three-dimensional reference surface temperature model includes:

[0066] Multiple surface temperature points are extracted from the 3D temperature model. Then, one surface temperature point is extracted from each of these multiple surface temperature points to obtain the target surface temperature point. The following operations are then performed on the target surface temperature point:

[0067] Obtain the target surface normal vector and target surface spatial coordinates of the target surface temperature point, and obtain the optical center spatial coordinates;

[0068] Based on the spatial coordinates of the optical center and the spatial coordinates of the target surface, determine the target observation direction vector corresponding to the temperature point on the target surface;

[0069] Calculate the target observation angle corresponding to the target surface temperature point based on the target surface normal vector and the target observation direction vector;

[0070] Based on the target observation angle and preset surface emissivity parameters, determine the target temperature compensation value for the target surface temperature point;

[0071] The surface correction temperature value is obtained by calculating the sum of the target surface temperature value and the target temperature compensation value.

[0072] The surface correction temperature values ​​are summarized to obtain multiple surface correction temperature values, and a three-dimensional reference surface temperature model is constructed based on these multiple surface correction temperature values.

[0073] Optionally, the formula for calculating the target observation angle is as follows:

[0074] ;

[0075] in, Indicates the target observation angle. Represents the inverse cosine function. Represents the target surface normal vector. Represents the target observation direction vector. Represents the magnitude of a vector.

[0076] Optionally, determining the abnormal surface temperature rise region based on the three-dimensional surface temperature rise model and sampling interval includes:

[0077] Multiple 3D temperature rise points are extracted from the 3D surface temperature rise model. Then, one 3D temperature rise point is extracted from each of these multiple points to obtain the target 3D temperature rise point. The following operations are then performed on the target 3D temperature rise point:

[0078] Obtain the target temperature rise value at the target three-dimensional temperature rise point;

[0079] The target temperature rise rate is obtained by dividing the target temperature rise value by the sampling interval.

[0080] If the target temperature rise rate is greater than or equal to the preset temperature rise rate threshold, the target three-dimensional temperature rise point is recorded as an abnormal temperature rise point.

[0081] By summarizing the abnormal temperature rise points, multiple abnormal temperature rise points are obtained;

[0082] Perform connected component analysis on multiple abnormal temperature rise points to obtain multiple abnormal temperature rise regions, and record these multiple abnormal temperature rise regions as surface abnormal temperature rise regions.

[0083] To achieve the above objectives, the present invention also provides an infrared positioning system for hot spot temperature rise in a magnetizing transformer, comprising:

[0084] The benchmark model construction module is used to receive the hot spot temperature rise infrared positioning command, confirm the magnetizing transformer based on the hot spot temperature rise infrared positioning command, and perform benchmark multi-view image acquisition on the magnetizing transformer based on the pre-constructed multiple benchmark image acquisition perspectives to obtain multiple benchmark image sets. The benchmark image sets include: benchmark visible light images and benchmark depth images. A three-dimensional benchmark surface model of the magnetizing transformer is constructed based on the multiple benchmark image sets.

[0085] The temperature model construction module is used to start the magnetizing transformer, perform multi-view reference temperature image acquisition on the magnetizing transformer after startup, obtain multiple reference temperature images, construct a three-dimensional reference surface temperature model of the magnetizing transformer based on the multiple reference temperature images and the three-dimensional reference surface model, perform multi-view real-time temperature image acquisition on the magnetizing transformer after startup, obtain multiple real-time temperature images, and construct a three-dimensional real-time surface temperature model of the magnetizing transformer based on the multiple real-time temperature images and the three-dimensional reference surface model.

[0086] The temperature rise model generation module is used to calculate the temperature difference between the three-dimensional real-time surface temperature model and the three-dimensional reference surface temperature model to obtain the three-dimensional surface temperature rise model.

[0087] The hotspot localization module is used to acquire the sampling times of the multiple three-dimensional reference images and the sampling times of the multiple real-time image sets, obtain the reference sampling time and the real-time sampling time, subtract the reference sampling time from the real-time sampling time to obtain the sampling interval, determine the abnormal surface temperature rise area based on the three-dimensional surface temperature rise model and the sampling interval, and generate hotspot temperature rise infrared localization results based on the abnormal surface temperature rise area.

[0088] To address the above problems, the present invention also provides an electronic device, the electronic device comprising:

[0089] Memory, storing at least one instruction;

[0090] The processor executes the instructions stored in the memory to implement the above-described infrared positioning method for hot spot temperature rise of a magnetizing transformer.

[0091] To address the aforementioned problems, the present invention also provides a computer-readable storage medium storing at least one instruction, which is executed by a processor in an electronic device to implement the aforementioned infrared positioning method for hot spot temperature rise of a magnetizing transformer.

[0092] To address the problems described in the background art, this invention receives a hotspot temperature rise infrared positioning command and confirms the magnetizing transformer to be detected. After the magnetizing transformer is started, multiple reference temperature images, multiple reference visible light images, and multiple reference depth images are acquired based on multiple reference image acquisition perspectives. Initial three-dimensional point cloud models under different acquisition perspectives are constructed based on the reference depth images and reference visible light images, and the temperature values ​​in the reference temperature images are mapped to the corresponding initial three-dimensional point cloud models. This invention calculates the target observation angle based on the surface normal vector of each surface temperature point and the optical center spatial coordinates of the infrared image acquisition device, and compensates and corrects the surface temperature value based on the target observation angle and surface emissivity parameters. Then, the three-dimensional temperature correction models under multiple acquisition perspectives are converted to the same spatial coordinate system and point cloud fusion is performed to obtain a three-dimensional reference surface temperature model. A three-dimensional real-time surface temperature model is constructed, and the temperature difference between the corresponding surface locations in the three-dimensional real-time surface temperature model and the three-dimensional reference surface temperature model is calculated to obtain a three-dimensional surface temperature rise model. Multiple three-dimensional temperature rise points are extracted from the three-dimensional surface temperature rise model. Three-dimensional temperature rise points with temperature rise values ​​greater than or equal to a preset temperature rise threshold are identified as abnormal temperature rise points. Based on the spatial relationship between multiple abnormal temperature rise points, the abnormal temperature rise region on the surface is determined, and then the infrared positioning result of hotspot temperature rise is generated based on the abnormal temperature rise region. Therefore, this invention improves the accuracy of surface temperature rise calculation and hotspot positioning of the magnetizing transformer by correcting the infrared temperature measurement deviation caused by the observation angle and surface emissivity, and by unifying surface temperature data from different acquisition angles to the same spatial coordinate system. Thus, this invention improves the accuracy of surface temperature data of the magnetizing transformer by compensating for and correcting the infrared temperature measurement deviation caused by the observation angle and surface emissivity. Unifying surface structure data and temperature data from different reference image acquisition angles to the same spatial coordinate system allows for the determination of the three-dimensional spatial location and coverage area of ​​the abnormal temperature rise region on the surface of the magnetizing transformer. By calculating the temperature rise rate at each three-dimensional temperature rise point using the sampling interval, the impact of the different sampling intervals between the reference temperature image and the real-time temperature image on the judgment of abnormal temperature rise can be reduced. Therefore, this invention can improve the accuracy of locating temperature rise hotspots in magnetized transformers by correcting infrared thermometry deviations caused by observation angle and surface emissivity and constructing a three-dimensional surface temperature rise distribution. Attached Figure Description

[0093] Figure 1 This is a flowchart illustrating a hotspot temperature rise infrared positioning method for a magnetizing transformer according to an embodiment of the present invention.

[0094] Figure 2 This is a functional block diagram of a hotspot temperature rise infrared positioning system for a magnetizing transformer provided in an embodiment of the present invention;

[0095] Figure 3 This is a schematic diagram of the structure of an electronic device for implementing the infrared positioning method for hot spot temperature rise of a magnetizing transformer, according to an embodiment of the present invention.

[0096] Explanation of reference numerals in the attached figures:

[0097] 10. Electronic device; 11. Processor; 12. Memory; 13. Bus.

[0098] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0099] It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0100] This application provides a method for infrared location of hotspot temperature rise in magnetizing transformers. The executing entity of this method includes, but is not limited to, at least one of the following electronic devices that can be configured to execute the method provided in this application: a server, a terminal, etc. In other words, the method for infrared location of hotspot temperature rise in magnetizing transformers can be executed by software or hardware installed on a terminal device or a server device, and the software can be a blockchain platform. The server includes, but is not limited to, a single server, a server cluster, a cloud server, or a cloud server cluster.

[0101] Reference Figure 1 The diagram shown is a flowchart illustrating a hotspot temperature rise infrared location method for a magnetizing transformer according to an embodiment of the present invention. In this embodiment, the hotspot temperature rise infrared location method for a magnetizing transformer includes:

[0102] S1. Receive the hot spot temperature rise infrared positioning command and confirm the magnetizing transformer based on the hot spot temperature rise infrared positioning command.

[0103] It should be noted that the hotspot temperature rise infrared positioning command is a command used to initiate hotspot temperature rise infrared positioning. Hotspot temperature rise infrared positioning refers to using an infrared thermal imager to identify hotspots in the magnetizing transformer where the temperature is abnormally rising. The magnetizing transformer refers to a transformer used to magnetize the equipment.

[0104] Understandably, during the magnetization process, the electronic components of the magnetizing transformer, such as the windings, core, and terminals, may heat up due to resistance losses, hysteresis losses, eddy current losses, or increased contact resistance. Excessive temperature rise in the magnetizing transformer will accelerate the aging and damage of electronic components; therefore, it is necessary to locate the hotspots of temperature rise to determine the possible location of a fault.

[0105] S2. Based on multiple pre-constructed reference image acquisition perspectives, perform reference multi-view image acquisition on the magnetizing transformer to obtain multiple reference image sets. The reference image sets include: reference visible light images and reference depth images. Construct a three-dimensional reference surface model of the magnetizing transformer based on the multiple reference image sets.

[0106] It should be understood that the reference image acquisition perspective refers to the perspective used when performing reference multi-view image acquisition on the magnetizing transformer; different reference image sets correspond to different perspectives. The reference visible light image refers to an image of a portion of the surface of the magnetizing transformer acquired by a visible light image acquisition device (such as an industrial camera) from a reference image acquisition perspective before the magnetizing transformer is started. The reference depth image refers to a depth image of a portion of the surface of the magnetizing transformer acquired by a depth image acquisition device (such as a depth camera) from a reference image acquisition perspective before the magnetizing transformer is started.

[0107] Furthermore, the construction of the three-dimensional reference surface model of the magnetized transformer based on multiple reference image sets includes:

[0108] A target reference image set is obtained by sequentially extracting one reference image set from multiple reference image sets. The target reference image set includes a target reference visible light image and a target reference depth image.

[0109] Perform the following operations on the target reference image set:

[0110] Perform effective depth pixel filtering on the target reference depth image to obtain multiple effective depth pixels of the target;

[0111] For each of the multiple target effective depth pixels, perform the following operation:

[0112] Obtain the intrinsic parameters of the depth camera, the depth pixel coordinates and depth values ​​of the target's effective depth pixels;

[0113] Based on the intrinsic parameters of the depth camera, the coordinates of the depth pixels, and the depth value, spatial projection is performed on the effective depth pixels of the target to obtain the three-dimensional pixels of the target.

[0114] Pixel alignment is performed on the target reference depth image and the target reference visible light image to obtain the target aligned visible light image;

[0115] Extract visible light pixels with the same depth pixel coordinates from the target-aligned visible light image, and obtain the grayscale value of the visible light pixels to obtain the visible light grayscale value;

[0116] The target 3D pixel is assigned a gray value by using the visible light gray value as the gray value of the target 3D pixel, and the target reference point is obtained.

[0117] By summing up the target benchmark points, multiple target benchmark points are obtained;

[0118] An initial 3D point cloud model is constructed based on multiple target reference points;

[0119] By summing up the initial 3D point cloud models, multiple initial 3D point cloud models are obtained;

[0120] Point cloud fusion is performed on multiple initial 3D point cloud models to obtain a 3D reference surface model.

[0121] It should be noted that the effective depth pixel filtering of the target reference depth image refers to extracting multiple target depth pixels from the target reference depth image and obtaining multiple depth values ​​corresponding to these target depth pixels. If a depth value is not zero or is null, the target depth pixel corresponding to that depth value is recorded as a target effective depth pixel. These target effective depth pixels are then aggregated to obtain multiple target effective depth pixels. It is understood that the target depth pixel refers to a pixel in the target reference depth image.

[0122] It should be understood that when a depth camera acquires depth images of the magnetized transformer surface, factors such as surface reflection and edge occlusion may prevent the acquisition of depth values ​​usable for back-projection calculations at some sampling points. This could result in null or zero-valued target depth pixels in the target reference depth image. Therefore, it is necessary to perform effective depth pixel filtering on the target reference depth image.

[0123] Specifically, the depth camera intrinsic parameters refer to the camera intrinsic parameters of the depth image acquisition device. The process of performing spatial projection on the target effective depth pixels based on the depth camera intrinsic parameters, depth pixel coordinates, and depth values ​​involves extracting the lateral focal length, longitudinal focal length, principal point x-coordinate, and principal point y-coordinate from the depth camera intrinsic parameters; extracting the depth pixel x-coordinate and depth pixel y-coordinate from the depth pixel coordinates; and calculating the target 3D pixel x-coordinate and y-coordinate using the formulas for calculating the target 3D pixel x-coordinate and y-coordinate. The depth value is then used as the vertical coordinate to obtain the target 3D pixel. Therefore, the target 3D pixel is used to characterize a discrete point of a sampling point on the surface of the magnetizing transformer in a pre-constructed spatial coordinate system. The formula for calculating the 3D x-coordinate is shown below:

[0124] ;

[0125] The formula for calculating the three-dimensional ordinate is as follows:

[0126] ;

[0127] in, Represents the three-dimensional x-axis. Represents the three-dimensional ordinate. Indicates the x-coordinate of the principal point. Represents the ordinate of the principal point. Represents the x-coordinate of the depth pixels. Represents the depth pixel ordinate. Indicates the depth value. Indicates the lateral focal length. Indicates the longitudinal focal length.

[0128] It should be explained that the lateral focal length refers to the focal length of the depth image acquisition device along the horizontal axis. The longitudinal focal length refers to the focal length of the depth image acquisition device along the vertical axis. The principal point x-coordinate refers to the x-coordinate of the principal point in the image coordinate system of the target reference depth image. The principal point y-coordinate refers to the y-coordinate of the principal point in the image coordinate system of the target reference depth image. The principal point refers to the intersection of the optical axis of the depth image acquisition device and the plane represented by the photosensitive element of the depth image acquisition device. It is understood that the lateral focal length, longitudinal focal length, principal point x-coordinate, and principal point y-coordinate can all be obtained from the technical parameters in the instruction manual of the depth image acquisition device.

[0129] It is understood that the x-coordinate of the depth pixel refers to the horizontal coordinate of the depth pixel coordinates, and the y-coordinate of the depth pixel refers to the vertical coordinate of the depth pixel coordinates. The depth pixel coordinates refer to the pixel coordinates of the effective depth pixels of the target in the image coordinate system of the depth image.

[0130] Importantly, the spatial coordinate system is constructed as follows: The origin is the optical center of the depth camera, the optical axis of the depth camera is the vertical axis, the direction from the depth camera to the magnetizing transformer is the positive direction of the vertical axis, and the plane perpendicular to the optical axis and passing through the optical center is the XOY plane. Two perpendicular lines are drawn on the XOY plane, passing through the optical center. One of these lines is used as the horizontal axis, and the other as the vertical axis, thus obtaining the spatial coordinate system. It should be noted that this spatial coordinate system corresponds to an initial 3D point cloud model constructed from the target reference depth image and the target reference visible light image from a reference image acquisition perspective.

[0131] It should be understood that the target effective depth pixels represent pixels in the target reference depth image, which cannot correspond to the position on the three-dimensional transformer surface. Therefore, it is necessary to convert the two-dimensional target effective depth pixels into three-dimensional target three-dimensional pixels.

[0132] It is clear that performing pixel alignment on the target reference depth image and the target reference visible light image refers to converting the pixel coordinates of pixels in the target reference visible light image into pixel coordinates in the image coordinate system of the target reference depth image. This ensures that pixels with the same pixel coordinates in the target reference visible light image and the target reference depth image correspond to the same target 3D pixel in their spatial coordinate systems. This allows for the extraction of visible light pixels (pixels in the target aligned visible light image) with the same pixel coordinates as the depth pixels, and the acquisition of visible light grayscale values. Pixel alignment is existing technology and will not be elaborated upon here.

[0133] It should be noted that the depth value only represents depth and does not display color. Therefore, it is not possible to intuitively see which actual location on the surface of the magnetizing transformer corresponds to the target 3D pixel. Thus, the visible light grayscale value needs to be used as the grayscale value of the target 3D pixel to assign grayscale values. This allows the initial 3D point cloud model to intuitively show the location of each target reference point on the surface of the magnetizing transformer, thereby determining the location of hotspots. The grayscale value assignment refers to the process of assigning visible light grayscale values ​​as the grayscale values ​​of the target 3D pixels.

[0134] Specifically, the target reference point refers to a target three-dimensional pixel containing grayscale values ​​in a spatial coordinate system. Constructing an initial three-dimensional point cloud model based on multiple target reference points refers to using the target reference points as discrete points in the point cloud to construct the point cloud, and using this point cloud as the initial three-dimensional point cloud model.

[0135] Furthermore, the step of performing point cloud fusion on multiple initial 3D point cloud models to obtain a 3D reference surface model includes:

[0136] A sliding window extraction operation is performed on multiple initial 3D point cloud models to obtain a combination of multiple 3D point cloud models, wherein the combination of 3D point cloud models includes two initial 3D point cloud models.

[0137] For each of the multiple 3D point cloud model combinations, adjacent viewpoint pose calculations are performed to obtain multiple adjacent viewpoint transformation parameter sets, which include rotation matrices and translation vectors:

[0138] Extract one reference image acquisition view from multiple reference image acquisition viewpoints to obtain the target reference image acquisition viewpoint;

[0139] Perform a retention operation based on the elimination of target reference image acquisition viewpoints on multiple reference image acquisition viewpoints to obtain multiple retained image acquisition viewpoints;

[0140] Based on multiple adjacent viewpoint transformation parameters, the target viewpoint transformation parameters between each of the multiple preserved image acquisition viewpoints and the target reference image acquisition viewpoint are calculated respectively, thus obtaining multiple target viewpoint transformation parameters;

[0141] Based on multiple target viewpoint transformation parameters, point cloud merging is performed on multiple initial 3D point cloud models to obtain a 3D reference surface model.

[0142] It should be noted that the sliding window extraction operation on multiple initial 3D point cloud models refers to arranging multiple initial 3D point cloud models according to the acquisition order of multiple reference depth images from first to last, obtaining an initial 3D point cloud model sequence, using two initial 3D point cloud models as the sliding window size and one initial 3D point cloud model as the step size, and extracting multiple 3D point cloud model combinations from the initial 3D point cloud model sequence.

[0143] For example, there exist a first initial 3D point cloud model, a second initial 3D point cloud model, and a third initial 3D point cloud model, corresponding to a first reference depth image, a second reference depth image, and a third reference depth image, respectively. The acquisition order of the first reference depth image, the second reference depth image, and the third reference depth image is: first acquire the first reference depth image, then acquire the second reference depth image, and finally acquire the third reference depth image. Then, the initial 3D point cloud model sequence is {first initial 3D point cloud model, second initial 3D point cloud model, third initial 3D point cloud model}. Multiple 3D point cloud model combinations extracted from the initial 3D point cloud model sequence are: a first 3D point cloud model combination {first initial 3D point cloud model, second initial 3D point cloud model} and a second 3D point cloud model combination {second initial 3D point cloud model, third initial 3D point cloud model}.

[0144] Furthermore, for each of the multiple 3D point cloud model combinations, adjacent viewpoint pose calculations are performed to obtain multiple adjacent viewpoint transformation parameter sets, including:

[0145] Extract one 3D point cloud model combination from multiple 3D point cloud model combinations sequentially to obtain the target 3D point cloud model combination, and perform the following operations on the target 3D point cloud model combination:

[0146] Extract the front-end 3D point cloud model and the back-end 3D point cloud model from the target 3D point cloud model combination;

[0147] Obtain the corresponding front reference visible light image and rear reference visible light image of the front 3D point cloud model and the rear 3D point cloud model;

[0148] Feature point extraction is performed on the front reference visible light image and the rear reference visible light image respectively to obtain the front feature point set and the rear feature point set;

[0149] Feature point matching is performed on the pre-set and post-set feature points to obtain multiple initial matching feature point pairs;

[0150] Multiple initial matching feature point pairs are filtered for matching errors to obtain multiple target matching point pairs;

[0151] Multiple preceding matching points and multiple following matching points are extracted from multiple target matching point pairs. Multiple preceding spatial coordinates of multiple preceding matching points and multiple following spatial coordinates of multiple following matching points are extracted from the preceding 3D point cloud model and the following 3D point cloud model, respectively.

[0152] Based on multiple front spatial coordinates and multiple back spatial coordinates, the rotation matrix and translation vector between the front 3D point cloud model and the back 3D point cloud model are calculated to obtain the adjacent viewpoint transformation parameter set.

[0153] By summing up the adjacent viewpoint transformation parameter sets, multiple adjacent viewpoint transformation parameter sets are obtained.

[0154] It should be understood that the "preceding 3D point cloud model" refers to the first initial 3D point cloud model extracted during the sliding window extraction process in the target 3D point cloud model combination. The second extracted initial 3D point cloud model is then used as the subsequent 3D point cloud model.

[0155] It is understood that the preceding reference visible light image refers to the reference visible light image used to construct the preceding 3D point cloud model. The following reference visible light image refers to the reference visible light image used to construct the following 3D point cloud model. The preceding feature point set refers to the set of multiple feature points extracted after feature point extraction is performed on the preceding reference visible light image, where each feature point is a preceding feature point. The following feature point set refers to the set of multiple feature points extracted after feature point extraction is performed on the following reference visible light image, where each feature point is a following feature point. The feature point extraction can be implemented using existing feature point extraction algorithms such as the SIFT feature point extraction algorithm, which will not be elaborated upon here.

[0156] It is clear that performing feature point matching on the preceding and following feature point sets refers to obtaining the preceding feature description vector corresponding to each preceding feature point in the preceding feature point set and the following feature description vector corresponding to each following feature point in the following feature point set. For each preceding and following feature description vector, perform pairwise combinations to obtain multiple feature description vector combinations. Calculate the Euclidean distance between the preceding and following feature description vectors in each feature description vector combination to obtain multiple Euclidean distances. For each preceding feature point, extract the following feature point with the smallest Euclidean distance from the following feature point set based on the multiple Euclidean distances to obtain candidate following feature points. Finally, form an initial matching feature point pair between the preceding feature point and the corresponding candidate following feature point.

[0157] It should be noted that the preceding feature description vector refers to the feature description vector of the preceding feature points, and the following feature description vector refers to the feature description vector of the following feature points. The feature description vector is a vector constructed from the local image features of feature points extracted from a visible light image using existing feature point extraction algorithms such as SIFT. Each element in the vector represents the gray-level gradient of the feature point in one direction. For example, the gray-level gradients in eight directions centered on the feature point: 0 degrees, 45 degrees, 90 degrees, 135 degrees, 180 degrees, 225 degrees, 270 degrees, and 315 degrees.

[0158] It should be understood that the initial matching feature point pair is used to characterize two feature points in the preceding and following reference visible light images that may correspond to the same location on the surface of the magnetized transformer. The smaller the Euclidean distance, the closer the texture, grayscale, and structure around the two feature points are, and the higher the probability that the two feature points correspond to the same location.

[0159] Importantly, since the preceding and following reference visible light images are matched for feature points based on the Euclidean distance between their feature description vectors, when the magnetizing transformer surface has repetitive heat dissipation structures, similar edges, areas of the same color, or local occlusion, feature points corresponding to different surface locations may have similar feature description vectors, thus forming initial matched feature point pairs. Furthermore, changes in the acquisition viewpoint, illumination, and image noise can also cause differences in the feature description vectors corresponding to the same surface location. If all initial matched feature point pairs are directly used to extract 3D coordinates, it may establish a correspondence between preceding 3D points (target reference points in the preceding 3D point cloud model) that are not at the same surface location and following 3D points (target reference points in the following 3D point cloud model), leading to errors in the calculated rotation matrix and translation vector. This results in multiple target reference points representing the same surface location in multiple 3D point cloud models corresponding to different positions on the actual magnetizing transformer surface, thus affecting the determination of subsequent hotspot locations. Therefore, it is necessary to perform matching error filtering on multiple initial matched feature point pairs, deleting initial matched feature point pairs that are incorrectly matched due to errors. The matching error screening can be implemented using existing technologies such as random sampling consensus algorithms, which will not be elaborated upon here.

[0160] Specifically, the preceding matching point refers to the preceding feature point extracted from a target matching point pair, and the following matching point refers to the following feature point extracted from a target matching point pair. The preceding spatial coordinates refer to the three-dimensional coordinates of the preceding matching point extracted from the preceding 3D point cloud model. The following spatial coordinates refer to the three-dimensional coordinates of the following matching point extracted from the following 3D point cloud model.

[0161] In detail, the calculation of the rotation matrix and translation vector between the front 3D point cloud model and the back 3D point cloud model refers to solving the rotation matrix and translation vector based on multiple front spatial coordinates and multiple back spatial coordinates, using existing technologies such as the Kabsch algorithm. This invention will not elaborate further here.

[0162] It should be explained that each 3D point cloud model corresponds to a spatial coordinate system, while the multiple initial 3D point cloud models correspond to multiple spatial coordinate systems. Constructing a 3D reference surface temperature model requires converting multiple initial 3D point cloud models in different spatial coordinate systems into multiple initial 3D point cloud models in the same spatial coordinate system. Specifically, the rotation matrix and translation vector are used to transform the subsequent 3D point cloud model into a 3D point cloud model in the spatial coordinate system corresponding to the preceding 3D point cloud model.

[0163] It is understood that the target reference image acquisition viewpoint refers to a reference image acquisition viewpoint randomly extracted from multiple reference image acquisition viewpoints. The spatial coordinate system corresponding to this viewpoint is used as the unique spatial coordinate system, and the initial 3D point cloud models corresponding to the multiple retained image acquisition viewpoints are transformed into the initial 3D point cloud models under this unique spatial coordinate system, thereby constructing a 3D reference surface model. The multiple retained image acquisition viewpoints refer to multiple reference image acquisition viewpoints after removing the target reference image acquisition viewpoint.

[0164] Furthermore, based on multiple adjacent viewpoint transformation parameters, the target viewpoint transformation parameters between each of the multiple preserved image acquisition viewpoints and the target reference image acquisition viewpoint are calculated respectively, resulting in multiple target viewpoint transformation parameters, including:

[0165] Extract one preserved image acquisition viewpoint from multiple preserved image acquisition viewpoints sequentially to obtain the target preserved image acquisition viewpoint, and perform the following operations on the target preserved image acquisition viewpoint:

[0166] Determine the viewpoint sequence between the target preserved image acquisition viewpoint and the target reference image acquisition viewpoint;

[0167] Based on the viewpoint sequence, the adjacent viewpoint transformation parameter set located between the target reference image acquisition viewpoint and the target preserved image acquisition viewpoint is extracted from multiple adjacent viewpoint transformation parameter sets to obtain multiple target adjacent viewpoint transformation parameter sets. Among them, the target adjacent viewpoint transformation parameter set includes the target adjacent rotation matrix and the target adjacent translation vector.

[0168] Extract multiple target adjacent rotation matrices from a set of transformation parameters for multiple target adjacent viewpoints;

[0169] For each of the multiple target adjacent rotation matrices, perform the following operation:

[0170] Determine the target's adjacent translation vectors corresponding to the target's adjacent rotation matrices to obtain the corresponding adjacent translation vectors;

[0171] Construct a homogeneous transformation matrix based on the target's adjacent rotation matrices and corresponding adjacent translation vectors;

[0172] By summing the homogeneous transformation matrices, we obtain multiple homogeneous transformation matrices;

[0173] Based on the view sequence, matrix multiplication is performed on multiple homogeneous transformation matrices to obtain a multiplied transformation matrix. The target view rotation matrix and target view translation vector are extracted from the multiplied transformation matrix and denoted as the target view transformation parameters.

[0174] By summarizing the target view transformation parameters, multiple target view transformation parameters are obtained.

[0175] It should be noted that the viewpoint sequence refers to the image acquisition views traversed from the target reference image acquisition viewpoint to the target retained image acquisition viewpoint. For example, following the order in which multiple reference image sets are obtained during the multi-view image acquisition, there are sequentially target reference image acquisition viewpoint, first retained image acquisition viewpoint, second retained image acquisition viewpoint, and third retained image acquisition viewpoint. If the extracted target retained image acquisition viewpoint is the second retained image acquisition viewpoint, then the viewpoint sequence is {target reference image acquisition viewpoint, first retained image acquisition viewpoint, second retained image acquisition viewpoint}. If the extracted target retained image acquisition viewpoint is the third retained image acquisition viewpoint, then the viewpoint sequence is {target reference image acquisition viewpoint, first retained image acquisition viewpoint, second retained image acquisition viewpoint, third retained image acquisition viewpoint}.

[0176] It is understood that the target adjacent viewpoint transformation parameter set refers to an adjacent viewpoint transformation parameter set between the target reference image acquisition viewpoint and the target retained image acquisition viewpoint. For example, following the above example, if the viewpoint sequence is {target reference image acquisition viewpoint, first retained image acquisition viewpoint, second retained image acquisition viewpoint}, then the target reference image acquisition viewpoint corresponds to the front 3D point cloud model in a 3D point cloud model combination, and the first retained image acquisition viewpoint corresponds to the rear 3D point cloud model in a 3D point cloud model combination. Therefore, the target reference image acquisition viewpoint and the first retained image acquisition viewpoint correspond to an adjacent viewpoint transformation parameter set. Similarly, the first retained image acquisition viewpoint and the second retained image acquisition viewpoint correspond to the front 3D point cloud model and the rear 3D point cloud model in another 3D point cloud model combination, respectively. Therefore, the first retained image acquisition viewpoint and the second retained image acquisition viewpoint also correspond to an adjacent viewpoint transformation parameter set. The target adjacent rotation matrix and the target adjacent translation vector refer to the rotation matrix and translation vector in the target adjacent viewpoint transformation parameter set, respectively.

[0177] It is clear that the corresponding adjacent translation vector refers to the target adjacent translation vector in the target adjacent viewpoint transformation parameter set corresponding to the target adjacent rotation matrix. The homogeneous transformation matrix is ​​shown below:

[0178] ;

[0179] in, Represents the target adjacent rotation matrix. This represents the corresponding adjacent translation vector.

[0180] It should be understood that the aforementioned matrix multiplication based on the viewpoint sequence refers to performing matrix multiplication on multiple homogeneous transformation matrices according to the arrangement order of the viewpoints acquired from multiple reference images in the viewpoint sequence, to obtain a multiplied transformation matrix. Matrix multiplication is a prior art technique, and will not be elaborated upon here.

[0181] For example, there exists a first homogeneous transformation matrix: The second homogeneous transformation matrix: The third homogeneous transformation matrix: Wherein, the first homogeneous transformation matrix corresponds to the target reference image acquisition viewpoint and the first preserved image acquisition viewpoint, the second homogeneous transformation matrix corresponds to the first preserved image acquisition viewpoint and the second preserved image acquisition viewpoint, and the third homogeneous transformation matrix corresponds to the second preserved image acquisition viewpoint and the third preserved image acquisition viewpoint. Where the viewpoint sequence is {target reference image acquisition viewpoint, first preserved image acquisition viewpoint, second preserved image acquisition viewpoint, third preserved image acquisition viewpoint}, then performing matrix multiplication on multiple homogeneous transformation matrices is to calculate... The multiplication transformation matrix is ​​obtained as follows: ,in, This represents the target view rotation matrix. This represents the target viewpoint translation vector. Wherein, Let represent the target adjacent rotation matrix in the first homogeneous transformation matrix. This represents the corresponding adjacent translation vector in the first homogeneous transformation matrix. This represents the target adjacent rotation matrix in the second homogeneous transformation matrix. This represents the corresponding adjacent translation vector in the second homogeneous transformation matrix. This represents the target adjacent rotation matrix in the third homogeneous transformation matrix. This represents the corresponding adjacent translation vector in the third homogeneous transformation matrix. Let the first homogeneous transformation matrix be represented. Denotes the second homogeneous transformation matrix. Let represent the third homogeneous transformation matrix. This represents the multiplication transformation matrix.

[0182] It is known that the target view rotation matrix and the target view translation vector are used to transform the initial three-dimensional point cloud model corresponding to the target retained image acquisition view to the initial three-dimensional point cloud model in the spatial coordinate system corresponding to the target reference image acquisition view.

[0183] Specifically, based on multiple target viewpoint transformation parameters, a point cloud merging pointer is executed on multiple initial 3D point cloud models. For each target viewpoint transformation parameter, the initial 3D point cloud model corresponding to the target viewpoint transformation parameter is determined. For each target reference point in the initial 3D point cloud model, the 3D coordinates of each target reference point are obtained, and a column vector of three rows and one column is formed in the order of 3D horizontal coordinate, 3D vertical coordinate, and 3D vertical coordinate. The target viewpoint rotation matrix in the target viewpoint transformation parameters is multiplied by the column vector to obtain the coordinate vector. The target viewpoint translation vector is added to the coordinate vector to obtain the transformed 3D coordinates in the spatial coordinate system corresponding to the target reference image acquisition viewpoint.

[0184] S3. Start the magnetizing transformer and perform multi-view reference temperature image acquisition on the magnetizing transformer after startup to obtain multiple reference temperature images. Based on the multiple reference temperature images and the three-dimensional reference surface model, construct a three-dimensional reference surface temperature model of the magnetizing transformer.

[0185] It should be understood that "starting the magnetizing transformer" refers to energizing the magnetizing transformer, causing it to begin operation. "Performing multi-view reference temperature image acquisition on the started magnetizing transformer" refers to using an infrared image acquisition device to acquire reference temperature images of the started magnetizing transformer from multiple reference image acquisition perspectives, thereby obtaining multiple reference temperature images.

[0186] It is understood that the reference temperature image refers to the temperature image of a portion of the surface of the magnetizing transformer, acquired by an infrared image acquisition device (such as an infrared thermal imager) from a reference image acquisition perspective at the reference sampling time after the magnetizing transformer is started. This image is used to provide the initial temperature for the point cloud in the three-dimensional reference surface temperature model. It should be noted that the infrared image acquisition device, the visible light image acquisition device, and the depth image acquisition device can be integrated into a single image acquisition device (such as an RGBDT camera).

[0187] Furthermore, the construction of the three-dimensional reference surface temperature model of the magnetizing transformer based on multiple reference temperature images and a three-dimensional reference surface model includes:

[0188] Extract a target reference temperature image from multiple reference temperature images sequentially, and then perform the following operations on the target reference temperature image:

[0189] Determine the reference depth image corresponding to the target reference temperature image to obtain the target reference depth image;

[0190] Pixel alignment is performed on the target reference depth image and the target reference temperature image to obtain the aligned reference temperature image;

[0191] By summing the alignment reference temperature images, multiple alignment reference temperature images are obtained;

[0192] Based on multiple aligned reference temperature images, temperature mapping is performed on a three-dimensional reference surface model to obtain a three-dimensional temperature model.

[0193] Temperature compensation correction is performed on the three-dimensional temperature model to obtain a three-dimensional reference surface temperature model.

[0194] It is understood that the reference depth image, reference temperature image, and reference visible light image correspond one-to-one with the reference image acquisition viewpoint. Therefore, a reference depth image with the same reference image acquisition viewpoint as the target reference temperature image can be extracted as the target reference depth image.

[0195] It should be noted that the pixel alignment of the target reference depth image and the target reference temperature image involves converting the pixel coordinates of pixels in the target reference temperature image into pixel coordinates in the image coordinate system of the target reference depth image. This ensures that pixels with the same pixel coordinates in the target reference temperature image and the target reference depth image correspond to the same target reference point, thereby facilitating temperature mapping of the three-dimensional reference surface model to obtain a three-dimensional temperature model.

[0196] Understandably, performing temperature mapping on the three-dimensional reference surface model refers to replicating a three-dimensional reference surface model to obtain a three-dimensional replicated surface model. For each target reference point in the three-dimensional replicated surface model, the pixel coordinates of the target effective depth pixel corresponding to the target reference point are obtained, and the temperature of the pixel corresponding to the pixel coordinates with the same depth pixel coordinates is extracted from multiple aligned reference temperature images. The corresponding temperature is then marked in the three-dimensional reference surface model to obtain a three-dimensional temperature model. Therefore, this invention can determine whether the corresponding target reference point is a hotspot and the location of the hotspot on the surface of the magnetizing transformer by measuring temperature.

[0197] Furthermore, the step of performing temperature compensation correction on the three-dimensional temperature model to obtain a three-dimensional reference surface temperature model includes:

[0198] Multiple surface temperature points are extracted from the 3D temperature model. Then, one surface temperature point is extracted from each of these multiple surface temperature points to obtain the target surface temperature point. The following operations are then performed on the target surface temperature point:

[0199] Obtain the target surface normal vector and target surface spatial coordinates of the target surface temperature point, and obtain the optical center spatial coordinates;

[0200] Based on the spatial coordinates of the optical center and the spatial coordinates of the target surface, determine the target observation direction vector corresponding to the temperature point on the target surface;

[0201] Calculate the target observation angle corresponding to the target surface temperature point based on the target surface normal vector and the target observation direction vector;

[0202] Based on the target observation angle and preset surface emissivity parameters, determine the target temperature compensation value for the target surface temperature point;

[0203] The surface correction temperature value is obtained by calculating the sum of the target surface temperature value and the target temperature compensation value.

[0204] The surface correction temperature values ​​are summarized to obtain multiple surface correction temperature values, and a three-dimensional reference surface temperature model is constructed based on these multiple surface correction temperature values.

[0205] It should be noted that the surface temperature point refers to the target reference point in the three-dimensional temperature model. The target surface spatial coordinates refer to the three-dimensional coordinates of the target surface temperature point in the spatial coordinate system. The target surface temperature value refers to the temperature of the target surface temperature point. The target surface normal vector refers to the normal vector of the target surface temperature point on the surface of the magnetizing transformer in the three-dimensional temperature model. The normal vector is prior art, and will not be elaborated upon here.

[0206] Specifically, the optical center spatial coordinates refer to the three-dimensional coordinates of the optical center of the infrared image acquisition device in the spatial coordinate system. The specific process of determining the target observation direction vector corresponding to the target surface temperature point is as follows: subtract the target surface spatial coordinates from the optical center spatial coordinates, thereby using the vector pointing from the target surface temperature point to the optical center of the infrared image acquisition device as the target observation direction vector.

[0207] Furthermore, the formula for calculating the target observation angle is as follows:

[0208] ;

[0209] in, Indicates the target observation angle. Represents the inverse cosine function. Represents the target surface normal vector. Represents the target observation direction vector. Represents the magnitude of a vector.

[0210] It should be noted that when an infrared image acquisition device measures the temperature of a target surface, the measured temperature is affected by the target observation angle of the infrared image acquisition device and the emissivity of the target surface temperature point. Therefore, the target observation angle and the emissivity (surface emissivity parameter) of the target surface temperature point are used as independent variables to input into the pre-constructed temperature compensation regression equation to calculate the correction value of the temperature that needs to be increased or decreased at that temperature point. This correction value is the target temperature compensation value.

[0211] It should be understood that the target observation angle refers to the angle at which the infrared image acquisition device acquires an infrared image of the target surface temperature point on the surface of the magnetizing transformer from a reference image acquisition perspective; it is also the angle in the spatial coordinate system corresponding to that reference image acquisition perspective. The emissivity of the target surface temperature point refers to the emissivity of the material corresponding to that target surface temperature point, which can be obtained by consulting the emissivity of that material recorded in books, papers, etc.

[0212] It's important to understand that infrared image acquisition devices determine temperature based on the energy radiated from the surface of a magnetized transformer. However, the energy of the received infrared radiation is affected by the surface emissivity parameters of the target surface temperature point and the target observation angle. The surface of a magnetized transformer may include different materials such as the core, windings, terminals, and insulation, each with varying surface emissivity. Furthermore, the acquired temperature differs at different target observation angles. To eliminate temperature errors caused by the target observation angle and to acquire the true temperature of the target surface temperature point, avoiding the influence of temperature errors on the temperature rise value and preventing misjudgments or omissions of hot spots, the target surface temperature value needs to be corrected based on the surface emissivity parameters of the target surface temperature point and the target observation angle. The target observation angle is calculated based on the target observation direction vector and the target surface normal vector.

[0213] It should be noted that the temperature compensation regression equation refers to the regression equation constructed by taking the target observation angle and surface emissivity parameter as independent variables and the target temperature compensation value as the dependent variable, and constructing a regression equation between the target observation angle, surface emissivity parameter, and target temperature compensation value. This can be obtained experimentally. For example, a magnetized transformer is heated to multiple set temperatures, and the actual temperature of the magnetized transformer is measured using a contact temperature sensor (such as a thermocouple thermometer). Multiple sampling points are set, and for each sampling point, the surface emissivity parameter of that sampling point is obtained. Multiple target observation angles are set, and at each target observation angle, an infrared image acquisition device is used to acquire the temperature image of the magnetized transformer. For each sampling point, at each target observation angle, the actual temperature of the sampling point is subtracted from the image temperature (the temperature of the sampling point in the temperature image) to obtain the target temperature compensation value of that sampling point at one target observation angle. The target temperature compensation values ​​are then summed to obtain the target temperature compensation value set for that sampling point at multiple target observation angles. Finally, the target temperature compensation value set for multiple sampling points is obtained. It can be seen that one sampling point corresponds to one surface emissivity parameter. Using multiple target observation angles and multiple surface emissivity parameters as independent variables, and multiple target temperature compensation value sets as dependent variables, the regression equation between the target observation angles and surface emissivity parameters and the target temperature compensation values ​​is fitted by the response surface methodology. The resulting regression equation is the temperature compensation regression equation.

[0214] Understandably, constructing a three-dimensional reference surface temperature model based on multiple surface correction temperature values ​​refers to annotating the corrected surface correction temperature values ​​onto the three-dimensional temperature model and deleting the originally annotated target surface temperature values ​​from the three-dimensional temperature model to obtain the three-dimensional reference surface temperature model. The three-dimensional reference surface temperature model is used to calculate the three-dimensional surface temperature rise model.

[0215] S4. Perform multi-view real-time temperature image acquisition on the magnetizing transformer after startup to obtain multiple real-time temperature images. Based on the multiple real-time temperature images and the three-dimensional reference surface model, construct a three-dimensional real-time surface temperature model of the magnetizing transformer.

[0216] It is understood that the multi-view real-time temperature image acquisition of the magnetizing transformer after startup refers to using an infrared image acquisition device to acquire real-time temperature images of the magnetizing transformer after startup from multiple reference image acquisition perspectives, thereby obtaining multiple real-time temperature images.

[0217] It should be understood that the real-time temperature image refers to the temperature image obtained by performing multi-view real-time temperature image acquisition on the magnetizing transformer after startup. The process of constructing the three-dimensional real-time surface temperature model of the magnetizing transformer based on multiple real-time temperature images and a three-dimensional reference surface model is the same as the process of constructing the three-dimensional reference surface temperature model of the magnetizing transformer based on multiple reference temperature images and a three-dimensional reference surface model, and will not be described again here.

[0218] It is clear that the real-time temperature image refers to the temperature image of the magnetizing transformer during operation. The three-dimensional real-time surface temperature model refers to a model of the surface of the magnetizing transformer during operation, with the temperature labeled at each point on the model.

[0219] S5. Calculate the temperature difference between the three-dimensional real-time surface temperature model and the three-dimensional reference surface temperature model to obtain the three-dimensional surface temperature rise model.

[0220] It should be understood that calculating the temperature difference between the 3D real-time surface temperature model and the 3D reference surface temperature model refers to obtaining the reference temperature and real-time temperature of each target reference point in the 3D reference surface temperature model, subtracting the reference temperature from the real-time temperature to obtain the temperature rise value. This temperature rise value represents the numerical increase in temperature at the target reference point. Multiple temperature rise values ​​are then summed, and these multiple temperature rise values ​​are used to replace multiple real-time temperatures in the 3D real-time surface temperature model to obtain the 3D surface temperature rise model. The reference temperature refers to the temperature of the target reference point extracted from the 3D reference surface temperature model. The real-time temperature refers to the temperature of the sampling point with the same 3D coordinates as the target reference point extracted from the 3D real-time surface temperature model. The 3D surface temperature rise model represents the temperature rise value of each sampling point on the surface of the magnetized transformer.

[0221] S6. Obtain the sampling time of the multiple three-dimensional reference images and the sampling time of the multiple real-time image sets to obtain the reference sampling time and the real-time sampling time. Subtract the reference sampling time from the real-time sampling time to obtain the sampling interval. Determine the abnormal surface temperature rise area based on the three-dimensional surface temperature rise model and the sampling interval. Generate hot spot temperature rise infrared positioning results based on the abnormal surface temperature rise area.

[0222] Furthermore, the determination of abnormal surface temperature rise regions based on the three-dimensional surface temperature rise model and sampling interval includes:

[0223] Multiple 3D temperature rise points are extracted from the 3D surface temperature rise model. Then, one 3D temperature rise point is extracted from each of these multiple points to obtain the target 3D temperature rise point. The following operations are then performed on the target 3D temperature rise point:

[0224] Obtain the target temperature rise value at the target three-dimensional temperature rise point;

[0225] The target temperature rise rate is obtained by dividing the target temperature rise value by the sampling interval.

[0226] If the target temperature rise rate is greater than or equal to the preset temperature rise rate threshold, the target three-dimensional temperature rise point is recorded as an abnormal temperature rise point.

[0227] By summarizing the abnormal temperature rise points, multiple abnormal temperature rise points are obtained;

[0228] Perform connected component analysis on multiple abnormal temperature rise points to obtain multiple abnormal temperature rise regions, and record these multiple abnormal temperature rise regions as surface abnormal temperature rise regions.

[0229] It is understood that the three-dimensional temperature rise point refers to a sampling point in the three-dimensional surface temperature rise model, and the target temperature rise value refers to the temperature rise value of the target three-dimensional temperature rise point. The reference sampling time refers to the moment when multi-view reference temperature image acquisition is performed on the magnetizing transformer after startup. The real-time sampling time refers to the moment when multi-view real-time temperature image acquisition is performed on the magnetizing transformer after startup.

[0230] It should be understood that the reference sampling time and the real-time sampling time are two adjacent sampling times, with the real-time sampling time lagging behind the reference sampling time. For example, if the preset sampling frequency is 30 seconds / time, that is, multi-view temperature image acquisition (multi-view reference temperature image acquisition or multi-view real-time temperature image acquisition) is performed once every 30 seconds, then after the first multi-view image acquisition is performed at 10:01, the second multi-view image acquisition is performed at 10:01:30, and the third multi-view image acquisition is performed at 10:02. If the first multi-view image acquisition performed at 10:01 is confirmed as the multi-view reference temperature image acquisition, and 10:01 is confirmed as the reference sampling time, then the second multi-view image acquisition performed at 10:01:30 is confirmed as the multi-view real-time temperature image acquisition, and 10:01:30 is confirmed as the real-time sampling time. If the second multi-view image acquisition performed at 10:01:30 is confirmed as the multi-view reference temperature image acquisition, and 10:01:30 is confirmed as the reference sampling time, then the third multi-view image acquisition performed at 10:02 is confirmed as the multi-view real-time temperature image acquisition, and 10:02 is confirmed as the real-time sampling time.

[0231] It is clear that the sampling interval refers to the difference between the real-time sampling time and the reference sampling time, used to characterize the time required for the target three-dimensional temperature rise point to rise from the reference temperature to the real-time temperature. The target temperature rise rate refers to the speed at which the temperature of the target three-dimensional temperature rise point rises, characterizing the magnitude of the temperature rise of the target three-dimensional temperature rise point per unit time. The unit time can be set to 1 minute.

[0232] It should be noted that the temperature rise rate threshold represents the upper limit of the temperature rise value of the sampling point on the surface of the magnetizing transformer corresponding to the target three-dimensional temperature rise point per unit time. If the target temperature rise rate is greater than or equal to the preset temperature rise rate threshold, it means that the temperature rise value of the target three-dimensional temperature rise point exceeds the upper limit of the temperature rise value per unit time. The generated heat cannot be released in time, causing heat to accumulate at the target three-dimensional temperature rise point. This may accelerate the aging of the insulation material of the magnetizing transformer windings, cause oxidation of the terminals or increase in contact resistance, and further increase the heat generation. This may lead to problems such as decreased insulation performance, winding short circuits, terminal burning, or shortened service life of the magnetizing transformer. Therefore, the target three-dimensional temperature rise point is identified as an abnormal temperature rise point to identify sampling points on the surface of the magnetizing transformer that may be abnormal temperature rise points, and to promptly locate abnormal temperature rise areas that may cause insulation aging or component damage.

[0233] It should be explained that the temperature rise threshold can be obtained through historical experience. For example, multiple normal three-dimensional surface temperature rise models are obtained from historical data. For a normal three-dimensional surface temperature rise model, the reference sampling time and real-time sampling time corresponding to the normal three-dimensional surface temperature rise model are obtained to obtain the historical reference sampling time and historical real-time sampling time. The historical sampling interval is obtained by subtracting the historical reference sampling time from the historical real-time sampling time. The maximum value of multiple temperature rise values ​​in the normal three-dimensional surface temperature rise model is extracted to obtain the maximum temperature rise value. The maximum temperature rise value is divided by the historical sampling interval to obtain the maximum temperature rise rate. The maximum temperature rise rates are summarized to obtain multiple maximum temperature rise rates corresponding to multiple normal three-dimensional surface temperature rise models. The average value and standard deviation of the multiple maximum temperature rise rates are calculated, and then the sum of the average value and three times the standard deviation is calculated to obtain the temperature rise rate threshold. It should be noted that the normal three-dimensional surface temperature rise model refers to a model in which multiple temperature rise values ​​will not cause the winding insulation material of the magnetizing transformer to age rapidly, the terminals to oxidize, or the contact resistance to increase.

[0234] It should be understood that performing connected component analysis on multiple abnormal temperature rise points refers to dividing interconnected abnormal temperature rise points into the same abnormal temperature rise region based on multiple three-dimensional coordinates of these points, using existing techniques such as the 8-connectivity algorithm, thus obtaining multiple abnormal temperature rise regions. It is clear that generating hotspot temperature rise infrared positioning results based on surface abnormal temperature rise regions means marking the surface abnormal temperature rise regions as hotspots in the three-dimensional surface temperature rise model, and outputting the three-dimensional surface temperature rise model with marked hotspots as the hotspot temperature rise infrared positioning result.

[0235] Importantly, the infrared location results of hotspot temperature rise determined by this invention are used to characterize areas on the surface of the magnetizing transformer that may have abnormal temperature rise points. This can serve as a basis for screening suspected hotspots during routine inspections of the magnetizing transformer, but it does not directly characterize hotspots in internal components such as windings and cores. To determine internal hotspots, it is necessary to use internal temperature detection techniques (such as inverting internal temperature by combining transformer surface temperature with internal structural parameters of the magnetizing transformer, or fiber optic temperature measurement) to identify internal abnormal temperature rise points.

[0236] To address the problems described in the background art, this invention receives a hotspot temperature rise infrared positioning command and confirms the magnetizing transformer to be detected. After the magnetizing transformer is started, multiple reference temperature images, multiple reference visible light images, and multiple reference depth images are acquired based on multiple reference image acquisition perspectives. Initial three-dimensional point cloud models under different acquisition perspectives are constructed based on the reference depth images and reference visible light images, and the temperature values ​​in the reference temperature images are mapped to the corresponding initial three-dimensional point cloud models. This invention calculates the target observation angle based on the surface normal vector of each surface temperature point and the optical center spatial coordinates of the infrared image acquisition device, and compensates and corrects the surface temperature value based on the target observation angle and surface emissivity parameters. Then, the three-dimensional temperature correction models under multiple acquisition perspectives are converted to the same spatial coordinate system and point cloud fusion is performed to obtain a three-dimensional reference surface temperature model. A three-dimensional real-time surface temperature model is constructed, and the temperature difference between the corresponding surface locations in the three-dimensional real-time surface temperature model and the three-dimensional reference surface temperature model is calculated to obtain a three-dimensional surface temperature rise model. Multiple three-dimensional temperature rise points are extracted from the three-dimensional surface temperature rise model. Three-dimensional temperature rise points with temperature rise values ​​greater than or equal to a preset temperature rise threshold are identified as abnormal temperature rise points. Based on the spatial relationship between multiple abnormal temperature rise points, the abnormal temperature rise region on the surface is determined, and then the infrared positioning result of hotspot temperature rise is generated based on the abnormal temperature rise region. Therefore, this invention improves the accuracy of surface temperature rise calculation and hotspot positioning of the magnetizing transformer by correcting the infrared temperature measurement deviation caused by the observation angle and surface emissivity, and by unifying surface temperature data from different acquisition angles to the same spatial coordinate system. Thus, this invention improves the accuracy of surface temperature data of the magnetizing transformer by compensating for and correcting the infrared temperature measurement deviation caused by the observation angle and surface emissivity. Unifying surface structure data and temperature data from different reference image acquisition angles to the same spatial coordinate system allows for the determination of the three-dimensional spatial location and coverage area of ​​the abnormal temperature rise region on the surface of the magnetizing transformer. By calculating the temperature rise rate at each three-dimensional temperature rise point using the sampling interval, the impact of the different sampling intervals between the reference temperature image and the real-time temperature image on the judgment of abnormal temperature rise can be reduced. Therefore, this invention can improve the accuracy of locating temperature rise hotspots in magnetized transformers by correcting infrared thermometry deviations caused by observation angle and surface emissivity and constructing a three-dimensional surface temperature rise distribution.

[0237] like Figure 2 The diagram shown is a functional block diagram of a hotspot temperature rise infrared positioning system for a magnetizing transformer provided in an embodiment of the present invention.

[0238] The hotspot temperature rise infrared positioning system 100 for magnetizing transformers described in this invention can be installed in electronic devices. Depending on the functions implemented, the hotspot temperature rise infrared positioning system 100 for magnetizing transformers may include a reference model construction module 101, a temperature model construction module 102, a temperature rise model generation module 103, and a hotspot positioning module 104. The module described in this invention can also be referred to as a unit, which refers to a series of computer program segments that can be executed by the processor of an electronic device and can perform a fixed function, and which are stored in the memory of the electronic device.

[0239] The reference model construction module 101 is used to receive the hot spot temperature rise infrared positioning command, confirm the magnetizing transformer based on the hot spot temperature rise infrared positioning command, and perform reference multi-view image acquisition on the magnetizing transformer based on the pre-constructed multiple reference image acquisition perspectives to obtain multiple reference image sets. The reference image sets include: reference visible light images and reference depth images. A three-dimensional reference surface model of the magnetizing transformer is constructed based on the multiple reference image sets.

[0240] The temperature model construction module 102 is used to start the magnetizing transformer, perform multi-view reference temperature image acquisition on the magnetizing transformer after startup to obtain multiple reference temperature images, construct a three-dimensional reference surface temperature model of the magnetizing transformer based on the multiple reference temperature images and the three-dimensional reference surface model, perform multi-view real-time temperature image acquisition on the magnetizing transformer after startup to obtain multiple real-time temperature images, and construct a three-dimensional real-time surface temperature model of the magnetizing transformer based on the multiple real-time temperature images and the three-dimensional reference surface model.

[0241] The temperature rise model generation module 103 is used to calculate the temperature difference between the three-dimensional real-time surface temperature model and the three-dimensional reference surface temperature model to obtain a three-dimensional surface temperature rise model.

[0242] The hotspot positioning module 104 is used to acquire the sampling time of the multiple three-dimensional reference images and the sampling time of the multiple real-time image sets, obtain the reference sampling time and the real-time sampling time, subtract the reference sampling time from the real-time sampling time to obtain the sampling interval, determine the abnormal surface temperature rise area based on the three-dimensional surface temperature rise model and the sampling interval, and generate hotspot temperature rise infrared positioning results based on the abnormal surface temperature rise area.

[0243] In detail, the modules in the infrared positioning system 100 for hot spot temperature rise of a magnetizing transformer described in this embodiment of the invention employ the same methods as described above during use. Figure 1 The method used is the same as the infrared positioning method for hot spot temperature rise of magnetized transformers described in the article, and can produce the same technical effect, so it will not be repeated here.

[0244] like Figure 3The diagram shown is a schematic representation of an electronic device for implementing an infrared positioning method for hot spot temperature rise in a magnetizing transformer, according to an embodiment of the present invention.

[0245] The electronic device 1 may include a processor 10, a memory 11 and a bus 12, and may also include a computer program stored in the memory 11 and executable on the processor 10, such as a hot spot temperature rise infrared positioning method program for magnetizing transformers.

[0246] The memory 11 includes at least one type of readable storage medium, such as flash memory, portable hard drive, multimedia card, card-type memory (e.g., SD or DX memory), magnetic memory, magnetic disk, optical disk, etc. In some embodiments, the memory 11 can be an internal storage unit of the electronic device 1, such as the portable hard drive of the electronic device 1. In other embodiments, the memory 11 can be an external storage device of the electronic device 1, such as a plug-in portable hard drive, smart media card (SMC), secure digital card (SD), flash card, etc., equipped on the electronic device 1. Furthermore, the memory 11 includes both internal storage units and external storage devices of the electronic device 1. The memory 11 can be used not only to store application software and various types of data installed on the electronic device 1, such as code for a hotspot temperature rise infrared positioning method program for a magnetizing transformer, but also to temporarily store data that has been output or will be output.

[0247] In some embodiments, the processor 10 may be composed of integrated circuits, such as a single packaged integrated circuit or multiple integrated circuits with the same or different functions, including combinations of one or more central processing units (CPUs), microprocessors, digital processing chips, graphics processors, and various control chips. The processor 10 is the control unit of the electronic device, connecting various components of the entire electronic device through various interfaces and lines. It executes programs or modules stored in the memory 11 (e.g., a hotspot temperature rise infrared positioning method program for a magnetizing transformer) and calls data stored in the memory 11 to perform various functions of the electronic device 1 and process data.

[0248] The bus 12 can be a peripheral component interconnect (PCI) bus or an extended industry standard architecture (EISA) bus, etc. The bus 12 can be divided into an address bus, a data bus, a control bus, etc. The bus 12 is configured to realize the connection and communication between the memory 11 and at least one processor 10, etc.

[0249] Figure 3 Only electronic devices with components are shown; it will be understood by those skilled in the art that... Figure 3 The structure shown does not constitute a limitation on the electronic device 1, and may include fewer or more components than shown, or combine certain components, or have different component arrangements.

[0250] For example, although not shown, the electronic device 1 may also include a power supply (such as a battery) to power the various components. Preferably, the power supply can be logically connected to the at least one processor 10 through a power management system, thereby enabling functions such as charging management, discharging management, and power consumption management through the power management system. The power supply may also include one or more DC or AC power supplies, recharging systems, power fault detection circuits, power converters or inverters, power status indicators, and other arbitrary components. The electronic device 1 may also include various sensors, Bluetooth modules, Wi-Fi modules, etc., which will not be described in detail here.

[0251] Furthermore, the electronic device 1 may also include a network interface. Optionally, the network interface may include a wired interface and / or a wireless interface (such as a Wi-Fi interface, a Bluetooth interface, etc.), which is typically used to establish communication connections between the electronic device 1 and other electronic devices.

[0252] Optionally, the electronic device 1 may further include a user interface, which may be a display, an input unit (such as a keyboard), and optionally, a standard wired interface or a wireless interface. Optionally, in some embodiments, the display may be an LED display, a liquid crystal display, a touch-sensitive liquid crystal display, or an OLED (Organic Light-Emitting Diode) touchscreen, etc. The display may also be appropriately referred to as a screen or display unit, used to display information processed in the electronic device 1 and to display a visual user interface.

[0253] The program for infrared location of hotspot temperature rise in magnetizing transformers, stored in the memory 11 of the electronic device 1, is a combination of multiple instructions. When run in the processor 10, it can achieve the following:

[0254] Receive the hot spot temperature rise infrared positioning command, and confirm the magnetizing transformer based on the hot spot temperature rise infrared positioning command;

[0255] Based on multiple pre-constructed reference image acquisition perspectives, reference multi-view image acquisition is performed on the magnetizing transformer to obtain multiple reference image sets, which include: reference visible light images and reference depth images;

[0256] A three-dimensional reference surface model of the magnetized transformer was constructed based on multiple reference image sets;

[0257] Start the magnetizing transformer and perform multi-view reference temperature image acquisition on the magnetizing transformer after startup to obtain multiple reference temperature images;

[0258] A three-dimensional reference surface temperature model of the magnetized transformer was constructed based on multiple reference temperature images and a three-dimensional reference surface model.

[0259] Multi-view real-time temperature image acquisition was performed on the magnetizing transformer after startup to obtain multiple real-time temperature images. Based on the multiple real-time temperature images and the three-dimensional reference surface model, a three-dimensional real-time surface temperature model of the magnetizing transformer was constructed.

[0260] The temperature difference between the three-dimensional real-time surface temperature model and the three-dimensional reference surface temperature model is calculated to obtain the three-dimensional surface temperature rise model.

[0261] The sampling times of the multiple three-dimensional reference images and the sampling times of the multiple real-time image sets are obtained to obtain the reference sampling time and the real-time sampling time. The sampling interval is obtained by subtracting the reference sampling time from the real-time sampling time.

[0262] The abnormal surface temperature rise region is determined based on the three-dimensional surface temperature rise model and sampling interval, and the hot spot temperature rise infrared positioning result is generated based on the abnormal surface temperature rise region.

[0263] Specifically, the processor 10's implementation method for the above instructions can be found in [reference needed]. Figures 1 to 3 The descriptions of the relevant steps in the corresponding embodiments are not repeated here.

[0264] Furthermore, if the modules / units integrated in the electronic device 1 are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. The computer-readable storage medium can be volatile or non-volatile. For example, the computer-readable medium may include: any entity or system capable of carrying the computer program code, a recording medium, a USB flash drive, a portable hard drive, a magnetic disk, an optical disk, a computer memory, or a read-only memory (ROM).

[0265] The present invention also provides a computer-readable storage medium storing a computer program, which, when executed by a processor of an electronic device, can perform the following:

[0266] Receive the hot spot temperature rise infrared positioning command, and confirm the magnetizing transformer based on the hot spot temperature rise infrared positioning command;

[0267] Based on multiple pre-constructed reference image acquisition perspectives, reference multi-view image acquisition is performed on the magnetizing transformer to obtain multiple reference image sets, which include: reference visible light images and reference depth images;

[0268] A three-dimensional reference surface model of the magnetized transformer was constructed based on multiple reference image sets;

[0269] Start the magnetizing transformer and perform multi-view reference temperature image acquisition on the magnetizing transformer after startup to obtain multiple reference temperature images;

[0270] A three-dimensional reference surface temperature model of the magnetized transformer was constructed based on multiple reference temperature images and a three-dimensional reference surface model.

[0271] Multi-view real-time temperature image acquisition was performed on the magnetizing transformer after startup to obtain multiple real-time temperature images. Based on the multiple real-time temperature images and the three-dimensional reference surface model, a three-dimensional real-time surface temperature model of the magnetizing transformer was constructed.

[0272] The temperature difference between the three-dimensional real-time surface temperature model and the three-dimensional reference surface temperature model is calculated to obtain the three-dimensional surface temperature rise model.

[0273] The sampling times of the multiple three-dimensional reference images and the sampling times of the multiple real-time image sets are obtained to obtain the reference sampling time and the real-time sampling time. The sampling interval is obtained by subtracting the reference sampling time from the real-time sampling time.

[0274] The abnormal surface temperature rise region is determined based on the three-dimensional surface temperature rise model and sampling interval, and the hot spot temperature rise infrared positioning result is generated based on the abnormal surface temperature rise region.

[0275] In the embodiments provided by this invention, it should be understood that the disclosed devices, systems, and methods can be implemented in other ways. For example, the system embodiments described above are merely illustrative, and actual implementations may have other classification methods.

[0276] The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0277] Furthermore, the functional modules in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or in the form of hardware plus software functional modules.

[0278] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0279] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A method for infrared location of hot spot temperature rise in a magnetizing transformer, characterized in that, The method includes: Receive the hot spot temperature rise infrared positioning command, and confirm the magnetizing transformer based on the hot spot temperature rise infrared positioning command; Based on multiple pre-constructed reference image acquisition perspectives, reference multi-view image acquisition is performed on the magnetizing transformer to obtain multiple reference image sets, which include: reference visible light images and reference depth images; A three-dimensional reference surface model of the magnetized transformer was constructed based on multiple reference image sets; Start the magnetizing transformer and perform multi-view reference temperature image acquisition on the magnetizing transformer after startup to obtain multiple reference temperature images; A three-dimensional reference surface temperature model of the magnetized transformer was constructed based on multiple reference temperature images and a three-dimensional reference surface model. Multi-view real-time temperature image acquisition was performed on the magnetizing transformer after startup to obtain multiple real-time temperature images. Based on the multiple real-time temperature images and the three-dimensional reference surface model, a three-dimensional real-time surface temperature model of the magnetizing transformer was constructed. The temperature difference between the three-dimensional real-time surface temperature model and the three-dimensional reference surface temperature model is calculated to obtain the three-dimensional surface temperature rise model. The sampling times of the multiple three-dimensional reference images and the sampling times of the multiple real-time image sets are obtained to obtain the reference sampling time and the real-time sampling time. The sampling interval is obtained by subtracting the reference sampling time from the real-time sampling time. The abnormal surface temperature rise region is determined based on the three-dimensional surface temperature rise model and sampling interval, and the hot spot temperature rise infrared positioning result is generated based on the abnormal surface temperature rise region.

2. The infrared positioning method for hot spot temperature rise in a magnetizing transformer as described in claim 1, characterized in that, The construction of the three-dimensional reference surface model of the magnetized transformer based on multiple reference image sets includes: A target reference image set is obtained by sequentially extracting one reference image set from multiple reference image sets. The target reference image set includes a target reference visible light image and a target reference depth image. Perform the following operations on the target reference image set: Perform effective depth pixel filtering on the target reference depth image to obtain multiple target effective depth pixels; For each of the multiple target effective depth pixels, perform the following operation: Obtain the intrinsic parameters of the depth camera, the depth pixel coordinates and depth values ​​of the target's effective depth pixels; Based on the intrinsic parameters of the depth camera, the coordinates of the depth pixels, and the depth value, spatial projection is performed on the effective depth pixels of the target to obtain the three-dimensional pixels of the target. Pixel alignment is performed on the target reference depth image and the target reference visible light image to obtain the target aligned visible light image; Extract visible light pixels with the same depth pixel coordinates from the target-aligned visible light image, and obtain the grayscale value of the visible light pixels to obtain the visible light grayscale value; The target 3D pixel is assigned a gray value by using the visible light gray value as the gray value of the target 3D pixel, and the target reference point is obtained. By summing up the target benchmark points, multiple target benchmark points are obtained; An initial 3D point cloud model is constructed based on multiple target reference points; By summing up the initial 3D point cloud models, multiple initial 3D point cloud models are obtained; Point cloud fusion is performed on multiple initial 3D point cloud models to obtain a 3D reference surface model.

3. The infrared positioning method for surface hotspot temperature rise of a magnetizing transformer as described in claim 2, characterized in that, The step of performing point cloud fusion on multiple initial 3D point cloud models to obtain a 3D reference surface model includes: A sliding window extraction operation is performed on multiple initial 3D point cloud models to obtain a combination of multiple 3D point cloud models, wherein the combination of 3D point cloud models includes two initial 3D point cloud models. For each of the multiple 3D point cloud model combinations, adjacent viewpoint pose calculations are performed to obtain multiple adjacent viewpoint transformation parameter sets, which include rotation matrices and translation vectors: Extract one reference image acquisition view from multiple reference image acquisition viewpoints to obtain the target reference image acquisition viewpoint; Perform a retention operation based on the elimination of target reference image acquisition viewpoints on multiple reference image acquisition viewpoints to obtain multiple retained image acquisition viewpoints; Based on multiple adjacent viewpoint transformation parameters, the target viewpoint transformation parameters between each of the multiple preserved image acquisition viewpoints and the target reference image acquisition viewpoint are calculated respectively, thus obtaining multiple target viewpoint transformation parameters; Based on multiple target viewpoint transformation parameters, point cloud merging is performed on multiple initial 3D point cloud models to obtain a 3D reference surface model.

4. The infrared positioning method for surface hot spot temperature rise of a magnetizing transformer as described in claim 3, characterized in that, The process involves performing adjacent viewpoint pose calculations on each of the multiple 3D point cloud model combinations to obtain multiple adjacent viewpoint transformation parameter sets, including: Extract one 3D point cloud model combination from multiple 3D point cloud model combinations sequentially to obtain the target 3D point cloud model combination, and perform the following operations on the target 3D point cloud model combination: Extract the front-end 3D point cloud model and the back-end 3D point cloud model from the target 3D point cloud model combination; Obtain the corresponding front reference visible light image and rear reference visible light image of the front 3D point cloud model and the rear 3D point cloud model; Feature point extraction is performed on the front reference visible light image and the rear reference visible light image respectively to obtain the front feature point set and the rear feature point set; Feature point matching is performed on the pre-set and post-set feature points to obtain multiple initial matching feature point pairs; Multiple initial matching feature point pairs are filtered for matching errors to obtain multiple target matching point pairs; Multiple preceding matching points and multiple following matching points are extracted from multiple target matching point pairs. Multiple preceding spatial coordinates of multiple preceding matching points and multiple following spatial coordinates of multiple following matching points are extracted from the preceding 3D point cloud model and the following 3D point cloud model, respectively. Based on multiple front spatial coordinates and multiple back spatial coordinates, the rotation matrix and translation vector between the front 3D point cloud model and the back 3D point cloud model are calculated to obtain the adjacent viewpoint transformation parameter set. By summing up the adjacent viewpoint transformation parameter sets, multiple adjacent viewpoint transformation parameter sets are obtained.

5. The infrared positioning method for surface hotspot temperature rise of a magnetizing transformer as described in claim 4, characterized in that, The method involves calculating target viewpoint transformation parameters between each of the multiple preserved image acquisition viewpoints and the target reference image acquisition viewpoint, based on multiple adjacent viewpoint transformation parameters, to obtain multiple target viewpoint transformation parameters, including: Extract one preserved image acquisition viewpoint from multiple preserved image acquisition viewpoints sequentially to obtain the target preserved image acquisition viewpoint, and perform the following operations on the target preserved image acquisition viewpoint: Determine the viewpoint sequence between the target preserved image acquisition viewpoint and the target reference image acquisition viewpoint; Based on the viewpoint sequence, the adjacent viewpoint transformation parameter set located between the target reference image acquisition viewpoint and the target preserved image acquisition viewpoint is extracted from multiple adjacent viewpoint transformation parameter sets to obtain multiple target adjacent viewpoint transformation parameter sets. Among them, the target adjacent viewpoint transformation parameter set includes the target adjacent rotation matrix and the target adjacent translation vector. Extract multiple target adjacent rotation matrices from a set of transformation parameters for multiple target adjacent viewpoints; For each of the multiple target adjacent rotation matrices, perform the following operation: Determine the target's adjacent translation vectors corresponding to the target's adjacent rotation matrices to obtain the corresponding adjacent translation vectors; Construct a homogeneous transformation matrix based on the target's adjacent rotation matrices and corresponding adjacent translation vectors; By summing the homogeneous transformation matrices, we obtain multiple homogeneous transformation matrices; Based on the view sequence, matrix multiplication is performed on multiple homogeneous transformation matrices to obtain a multiplied transformation matrix. The target view rotation matrix and target view translation vector are extracted from the multiplied transformation matrix and denoted as the target view transformation parameters. By summarizing the target view transformation parameters, multiple target view transformation parameters are obtained.

6. The infrared positioning method for hot spot temperature rise in a magnetizing transformer as described in claim 5, characterized in that, The construction of the three-dimensional reference surface temperature model of the magnetized transformer based on multiple reference temperature images and a three-dimensional reference surface model includes: Extract a target reference temperature image from multiple reference temperature images sequentially, and then perform the following operations on the target reference temperature image: Determine the reference depth image corresponding to the target reference temperature image to obtain the target reference depth image; Pixel alignment is performed on the target reference depth image and the target reference temperature image to obtain the aligned reference temperature image; By summing the alignment reference temperature images, multiple alignment reference temperature images are obtained; Based on multiple aligned reference temperature images, temperature mapping is performed on a three-dimensional reference surface model to obtain a three-dimensional temperature model. Temperature compensation correction is performed on the three-dimensional temperature model to obtain a three-dimensional reference surface temperature model.

7. The infrared positioning method for surface hotspot temperature rise of a magnetizing transformer as described in claim 6, characterized in that, The process of performing temperature compensation correction on the three-dimensional temperature model to obtain a three-dimensional reference surface temperature model includes: Multiple surface temperature points are extracted from the 3D temperature model. Then, one surface temperature point is extracted from each of these multiple surface temperature points to obtain the target surface temperature point. The following operations are then performed on the target surface temperature point: Obtain the target surface normal vector and target surface spatial coordinates of the target surface temperature point, and obtain the optical center spatial coordinates; Based on the spatial coordinates of the optical center and the spatial coordinates of the target surface, determine the target observation direction vector corresponding to the temperature point on the target surface; Calculate the target observation angle corresponding to the target surface temperature point based on the target surface normal vector and the target observation direction vector; Based on the target observation angle and preset surface emissivity parameters, determine the target temperature compensation value for the target surface temperature point; The surface correction temperature value is obtained by calculating the sum of the target surface temperature value and the target temperature compensation value. By summing up the surface correction temperature values, multiple surface correction temperature values ​​are obtained, and a three-dimensional reference surface temperature model is constructed based on these multiple surface correction temperature values.

8. The infrared positioning method for surface hot spot temperature rise of a magnetizing transformer as described in claim 7, characterized in that, The formula for calculating the target observation angle is as follows: ； in, Indicates the target observation angle. Represents the inverse cosine function. Represents the target surface normal vector. Represents the target observation direction vector. Represents the magnitude of a vector.

9. The infrared positioning method for surface hot spot temperature rise of a magnetizing transformer as described in claim 8, characterized in that, The determination of abnormal surface temperature rise regions based on a three-dimensional surface temperature rise model and sampling interval includes: Multiple 3D temperature rise points are extracted from the 3D surface temperature rise model. Then, one 3D temperature rise point is extracted from each of these multiple points to obtain the target 3D temperature rise point. The following operations are then performed on the target 3D temperature rise point: Obtain the target temperature rise value at the target three-dimensional temperature rise point; The target temperature rise rate is obtained by dividing the target temperature rise value by the sampling interval. If the target temperature rise rate is greater than or equal to the preset temperature rise rate threshold, the target three-dimensional temperature rise point is recorded as an abnormal temperature rise point. By summarizing the abnormal temperature rise points, multiple abnormal temperature rise points are obtained; Perform connected component analysis on multiple abnormal temperature rise points to obtain multiple abnormal temperature rise regions, and record these multiple abnormal temperature rise regions as surface abnormal temperature rise regions.

10. An infrared positioning system for surface hotspot temperature rise in a magnetizing transformer, characterized in that, The system includes: The benchmark model construction module is used to receive the hot spot temperature rise infrared positioning command, confirm the magnetizing transformer based on the hot spot temperature rise infrared positioning command, and perform benchmark multi-view image acquisition on the magnetizing transformer based on the pre-constructed multiple benchmark image acquisition perspectives to obtain multiple benchmark image sets. The benchmark image sets include: benchmark visible light images and benchmark depth images. A three-dimensional benchmark surface model of the magnetizing transformer is constructed based on the multiple benchmark image sets. The temperature model construction module is used to start the magnetizing transformer, perform multi-view reference temperature image acquisition on the magnetizing transformer after startup, obtain multiple reference temperature images, construct a three-dimensional reference surface temperature model of the magnetizing transformer based on the multiple reference temperature images and the three-dimensional reference surface model, perform multi-view real-time temperature image acquisition on the magnetizing transformer after startup, obtain multiple real-time temperature images, and construct a three-dimensional real-time surface temperature model of the magnetizing transformer based on the multiple real-time temperature images and the three-dimensional reference surface model. The temperature rise model generation module is used to calculate the temperature difference between the three-dimensional real-time surface temperature model and the three-dimensional reference surface temperature model to obtain the three-dimensional surface temperature rise model. The hotspot localization module is used to acquire the sampling times of the multiple three-dimensional reference images and the sampling times of the multiple real-time image sets, obtain the reference sampling time and the real-time sampling time, subtract the reference sampling time from the real-time sampling time to obtain the sampling interval, determine the abnormal surface temperature rise area based on the three-dimensional surface temperature rise model and the sampling interval, and generate hotspot temperature rise infrared localization results based on the abnormal surface temperature rise area.

Images