Power equipment temperature visual positioning method and device and storage medium

Abstract

The application relates to the technical field of power equipment detection, and discloses a power equipment temperature visual positioning method, a device and a storage medium, which comprise the following steps: extracting a temperature visual matrix from collected insulator information data; respectively performing zero padding processing on the upper and lower short sides of the temperature visual matrix to obtain a filled temperature visual matrix; taking the filled temperature visual matrix as input, training a basic model of a preset directional box detection model, inputting insulator temperature value matrix data of a target equipment to be detected into the power equipment directional box detection model, and obtaining a power equipment directional box of the target equipment; and based on relative position information in the power equipment directional box of the target equipment, determining a component box of the target equipment according to a maximum connected domain extracted from the power equipment directional box of the target equipment. The application can solve the problems in the prior art, such as the inability to accurately distinguish power equipment and background and the difficulty in separating densely distributed power equipment targets.

Description

Technical Field

[0001] This invention relates to the field of power equipment testing technology, and in particular to a method, apparatus and computer-readable storage medium for visually locating the temperature of power equipment. Background Technology

[0002] Current work on power equipment inspection primarily focuses on infrared images to study abnormal heating issues in power equipment. However, infrared images suffer from problems such as the presence of multiple false colors and severe interference from the photographed equipment due to complex shooting environments.

[0003] Existing methods suffer from drawbacks such as low detection accuracy and poor model generalization ability. Currently, most deep learning-based power equipment detection models mark the power equipment area with axis-aligned rectangles, which fails to represent the true shape of the power equipment, cannot accurately distinguish between power equipment and the background, and is difficult to separate densely distributed power equipment targets, thus making it difficult to further improve the localization accuracy. Summary of the Invention

[0004] This invention provides a method, apparatus, and computer-readable storage medium for visual positioning of power equipment temperature. Its main purpose is to solve the problems in the prior art, where the power equipment detection model based on deep learning marks the power equipment area with axis-aligned rectangles, which cannot express the true shape of the power equipment, cannot accurately distinguish between the power equipment and the background, and is difficult to separate densely distributed power equipment targets, thus resulting in low positioning accuracy.

[0005] In a first aspect, to achieve the above objective, the present invention provides a visual positioning method for the temperature of power equipment, the method comprising:

[0006] A temperature visual matrix is ​​extracted from the collected insulator information data; wherein, in the temperature visual matrix, the value of each element corresponds to the actual temperature value of a region in the environment.

[0007] The short side of the temperature visual matrix is ​​zero-filled both above and below, so that the length of the short side is equal to that of the long side, thus obtaining the filled temperature visual matrix.

[0008] The filled temperature visual matrix is ​​used as input to train the base model of the preset bounding box detection model to obtain the power equipment bounding box detection model; wherein, in the base model, the angle of the bounding box is classified by rotation target detection technology;

[0009] The insulator temperature value matrix data of the target device to be detected is input into the power equipment orientation frame detection model to obtain the power equipment orientation frame of the target device;

[0010] Based on the relative position information in the power equipment orientation frame of the target device, and according to the maximum connected component extracted from the power equipment orientation frame of the target device, the component frame of the target device is determined to complete the temperature visual positioning of the target device.

[0011] Furthermore, in a preferred embodiment, the extraction of the temperature visual matrix from the collected insulator information data includes:

[0012] The information data of the insulator is collected by an infrared thermal imager, and an image in the infrared universal data file storage format is obtained.

[0013] Extract a temperature visual matrix with a size of 640*480*1 from the image in the infrared universal data file storage format.

[0014] Furthermore, a preferred embodiment is that the short side of the temperature visual matrix is ​​zero-padded both above and below, so that the length of the short side is equal to the length of the long side, resulting in a filled temperature visual matrix.

[0015] The short side of the temperature visual matrix is ​​zero-filled at the top and bottom, so that the size of the temperature visual matrix is ​​filled from 640*480*1 to 640*640*1, thus obtaining the primary filled temperature visual matrix.

[0016] The range of numerical values ​​in the initial filled temperature visual matrix is ​​narrowed to 0~255 to obtain the filled temperature visual matrix.

[0017] Furthermore, a preferred approach is that the base model uses the RetinaNet network as the framework and ResNet50 and FPN as the backbone feature extraction networks.

[0018] Furthermore, in a preferred embodiment, the classification of the orientation box angle using rotational target detection technology in the basic model includes:

[0019] The orientation frame information of the insulator is represented by the long side representation method, where c_x and c_y represent the center coordinates of the rotating rectangle; longside represents the longest side; shortside represents the short side; theta represents the angle between the longest side and the x-axis in a counterclockwise rotation. The counterclockwise angle is negative, and theta ranges from -180° to 0°.

[0020] Furthermore, in a preferred embodiment, the step of using the filled temperature visual matrix as input to train the base model of the preset bounding box detection model to obtain the power equipment bounding box detection model includes:

[0021] The dataset of power equipment temperature values ​​corresponding to the filled temperature visual matrix is ​​divided into a training set and a test set in an 8:2 ratio;

[0022] The basic model is trained using the training set to obtain a primary training model;

[0023] The primary training model is tested using the test set. When the loss curve of the primary training model meets the preset model training loss requirements, a power equipment bounding box detection model is obtained.

[0024] Furthermore, a preferred embodiment is that the evaluation method for the power equipment bounding box detection model includes: precision and recall; wherein,

[0025] The formula for calculating the accuracy rate is: P = TP / (TP + FP);

[0026] The recall rate is calculated using the formula: R = TP / (TP + FN); where,

[0027] TP represents the number of devices correctly identified as electrical equipment, FP represents the number of devices incorrectly identified as electrical equipment, and FN represents the number of devices incorrectly identified as non-electrical equipment.

[0028] Furthermore, a preferred embodiment is that, based on the relative position information in the power equipment orientation frame of the target device, and according to the maximum connected component extracted from the power equipment orientation frame of the target device, the component frame of the target device is determined to complete the temperature visual positioning of the target device, including:

[0029] Set the temperature value of the insulator outside the power equipment orientation frame of the target device to 0, and binarize the image corresponding to the insulator temperature value matrix of the target device to obtain a binarized image of the target device.

[0030] Obtain the largest connected component within the bounding box of the power equipment in the binarized image;

[0031] Based on the relative position information in the power equipment orientation frame of the target device, the component frame of the target device is determined according to the maximum connected component to complete the temperature visual positioning of the target device.

[0032] Secondly, in order to solve the above problems, the present invention also provides a visual positioning device for temperature of power equipment, the device comprising:

[0033] The matrix extraction module is used to extract a temperature visual matrix from the collected insulator information data; wherein, in the temperature visual matrix, the value of each element corresponds to the actual temperature value of a region in the environment.

[0034] The zero-fill module is used to perform zero-fill processing on the top and bottom of the short side of the temperature visual matrix, so that the length of the short side of the temperature visual matrix is ​​equal to that of the long side, thus obtaining the filled temperature visual matrix.

[0035] The model training module is used to train the base model of the preset bounding box detection model by taking the filled temperature visual matrix as input, so as to obtain the bounding box detection model of power equipment; wherein, in the base model, the angle of the bounding box is classified by rotating target detection technology;

[0036] The phase frame detection module is used to obtain the power equipment orientation frame of the target equipment by inputting the insulator temperature value matrix data of the target equipment to be detected into the power equipment orientation frame detection model;

[0037] The temperature visual positioning module is used to determine the component frame of the target device based on the relative position information in the power equipment orientation frame of the target device and according to the maximum connected component extracted from the power equipment orientation frame of the target device, so as to complete the temperature visual positioning of the target device.

[0038] Thirdly, in order to solve the above problems, the present invention also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the power equipment temperature visual positioning method as described above.

[0039] The present invention proposes a method, device, and storage medium for visually locating the temperature of power equipment. By incorporating angle classification during the training of the basic model, angle information is added to the axis-aligned directional boxes, thereby improving the accuracy of directional box detection for power equipment. Based on directional box detection, combined with image processing methods, the component boxes of the target equipment to be detected are determined, thereby accurately locating the positions of each component of the power equipment. This facilitates the accurate detection of abnormal heating and fault diagnosis of power equipment. Attached Figure Description

[0040] Figure 1 This is a flowchart illustrating a method for visually locating the temperature of power equipment according to an embodiment of the present invention.

[0041] Figure 2 This is a schematic diagram of a power equipment temperature visual positioning device according to an embodiment of the present invention;

[0042] Figure 3 This is a schematic diagram of the test results of a power equipment orientation frame detection model provided in an embodiment of the present invention;

[0043] Figure 4 This is a schematic diagram of the loss curve during the training process of a power equipment orientation frame detection model provided in an embodiment of the present invention.

[0044] 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

[0045] It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0046] This invention provides a visual method for locating the temperature of power equipment. (Refer to...) Figure 1 The diagram shown is a flowchart illustrating a visual positioning method for the temperature of power equipment according to an embodiment of the present invention. This method can be executed by a device, which can be implemented in software and / or hardware.

[0047] In this embodiment, the visual positioning method for temperature of power equipment includes:

[0048] Step S110: Extract the temperature visual matrix from the collected insulator information data; wherein, in the temperature visual matrix, the value of each element corresponds to the actual temperature value of a region in the environment.

[0049] Specifically, infrared thermal imaging equipment is used to collect insulator information data. This data primarily includes the temperature value of each region within the insulator. To enrich the data, information such as the insulator's material and type can also be included, allowing the trained model to develop directional bounding box detection models for different insulator materials or types. A temperature visual matrix is ​​extracted from the collected insulator information data. In this matrix, each element corresponds to the actual temperature value of a region in the environment.

[0050] As an optional embodiment of the present invention, extracting the temperature visual matrix from the collected insulator information data includes:

[0051] The information data of the insulator is collected by an infrared thermal imager, and an image in the infrared universal data file storage format is obtained.

[0052] Extract a temperature visual matrix with a size of 640*480*1 from an image stored in the infrared universal data file format.

[0053] Specifically, the information data of the insulators collected by the infrared thermal imager is image data. Then, a temperature visual matrix with a size of 640*480*1 is extracted from the image data. In the temperature visual matrix, each element represents the actual temperature value of a corresponding area in the environment. The 640*480*1 matrix means that the longest side is 640 and the shortest side is 480.

[0054] Step S120: Zero-fill the top and bottom of the short side of the temperature visual matrix so that the length of the short side is equal to that of the long side, thus obtaining the filled temperature visual matrix.

[0055] Specifically, by filling the edges of the input matrix with zero values, the edges of the temperature vision matrix can be filtered to prevent the loss of image edge information during subsequent image adjustments.

[0056] As an optional embodiment of the present invention, the short side of the temperature visual matrix is ​​zero-filled both above and below, so that the length of the short side of the temperature visual matrix is ​​equal to that of the long side, resulting in a filled temperature visual matrix including:

[0057] The short side of the temperature visual matrix is ​​filled with zeros at the top and bottom, so that the size of the temperature visual matrix is ​​filled from 640*480*1 to 640*640*1, thus obtaining the initial filled temperature visual matrix.

[0058] The range of numerical values ​​in the initial filled temperature visual matrix is ​​narrowed to 0~255 to obtain the filled temperature visual matrix.

[0059] Step S130: Using the filled temperature visual matrix as input, train the basic model of the preset bounding box detection model to obtain the power equipment bounding box detection model; wherein, in the basic model, the rotation target detection technology is used to classify the angle of the bounding box.

[0060] Specifically, the base model of the preset bounding box detection model is obtained by pre-building the network structure of the model. In the base model, the angle of the bounding box is classified by using rotation target detection technology, thereby increasing the positional accuracy of the bounding box obtained when detecting power equipment. This helps to solve the problems in the existing technology of marking the power equipment area with axis-aligned rectangles, which cannot express the true shape of the power equipment, cannot accurately distinguish between the power equipment and the background, and is difficult to separate densely distributed power equipment targets.

[0061] As an optional embodiment of the present invention, the basic model uses the RetinaNet network as the framework and ResNet50 and FPN as the backbone feature extraction networks.

[0062] Specifically, the RetinaNet network structure in object detection algorithms uses ResNet as the backbone and FPN as the neck neural network, which provides good control over the number of training parameters. Overall, the RetinaNet network structure is simple and clear.

[0063] As an optional embodiment of the present invention, in the basic model, the classification of the angle of the bounding box using rotational target detection technology includes:

[0064] The orientation frame information of the insulator is represented by the long side representation method, where c_x and c_y represent the center coordinates of the rotating rectangle; longside represents the longest side; shortside represents the short side; theta represents the angle between the longest side and the x-axis in a counterclockwise rotation. The counterclockwise angle is negative, and theta ranges from -180° to 0°.

[0065] Specifically, during the basic model building process, Circular Smooth Label (CSL) technology (rotational target detection technology) is used to classify the angles of the orientation boxes. The angle regression methods that can be used in this process typically include the OpenCV definition method, the long-side definition method, and the ordered quadrilateral definition method. In one embodiment of this invention, the long-side representation method is used to represent the orientation box information of the insulator. The long-side representation method is used to represent the insulator orientation box information (c_x, c_y, longside, shortside, theta); where c_x and c_y represent the center coordinates of the rotated rectangle; longside represents the longest side; shortside represents the shortest side; and theta represents the angle between the longest side and the x-axis rotated counterclockwise, with the counterclockwise angle being negative, theta ∈ [-180, 0].

[0066] As an optional embodiment of the present invention, the filled temperature visual matrix is ​​used as input to train the basic model of the preset bounding box detection model, resulting in a power equipment bounding box detection model including:

[0067] The dataset of power equipment temperature values ​​corresponding to the filled temperature visual matrix was divided into training and testing sets in an 8:2 ratio.

[0068] The basic model is trained using the training set to obtain a primary training model;

[0069] The initial training model is tested using a test set. When the loss curve of the initial training model meets the preset model training loss requirements, the power equipment bounding box detection model is obtained.

[0070] Specifically, the dataset of power equipment temperature values ​​corresponding to the filled temperature visual matrix is ​​divided into a training set and a test set in an 8:2 ratio. In a preferred embodiment of the invention, the specific parameter settings of the network are shown in Table 1.

[0071]

[0072] Table 1

[0073] Training loss curve as shown Figure 4As shown, the initial training loss decreases rapidly with the increase of the number of periods, and the rate of decrease of the training loss slows down in subsequent periods but still decreases steadily, indicating that the parameter settings are reasonable.

[0074] As an optional embodiment of the present invention, the evaluation method for the power equipment bounding box detection model includes: precision and recall; wherein,

[0075] The formula for calculating accuracy is: P = TP / (TP + FP);

[0076] The recall rate is calculated using the formula: R = TP / (TP + FN); where,

[0077] TP represents the number of devices correctly identified as electrical equipment, FP represents the number of devices incorrectly identified as electrical equipment, and FN represents the number of devices incorrectly identified as non-electrical equipment.

[0078] Specifically, taking insulators as an example, the test results of the power equipment directional frame detection model are as follows: Figure 3 As shown, the insulator orientation identification results are marked with solid-line orientation rectangles, while the dashed-line orientation rectangles represent the original manual marking results. The fact that the heights of the light-colored and dark-colored boxes overlap in the figure indicates that the algorithm's orientation identification results are quite satisfactory.

[0079] Step S140: By inputting the insulator temperature value matrix data of the target device to be detected into the power equipment orientation frame detection model, the power equipment orientation frame of the target device is obtained.

[0080] Specifically, infrared equipment is used to acquire insulator information data of the target equipment to be detected. Then, insulator temperature value matrix data is extracted from the insulator information data of the target equipment, and the matrix data is input into the power equipment orientation frame detection model to obtain the power equipment orientation frame of the target equipment.

[0081] Step S150: Based on the relative position information in the power equipment orientation frame of the target device, determine the component frame of the target device according to the maximum connected component extracted from the power equipment orientation frame of the target device, so as to complete the temperature visual positioning of the target device.

[0082] Specifically, the relative position information in the power equipment orientation frame of the target device, combined with the largest connected component presented in the target device image, can determine the component frame of the target device, thereby achieving temperature visual localization of the target device, that is, locating each component in the target device.

[0083] As an optional embodiment of the present invention, based on the relative position information in the power equipment orientation frame of the target device, and according to the maximum connected component extracted from the power equipment orientation frame of the target device, the component frame of the target device is determined to complete the temperature visual positioning of the target device, including:

[0084] Set the temperature value of the insulator outside the power equipment orientation frame of the target device to 0, and binarize the image corresponding to the insulator temperature value matrix of the target device to obtain the binarized image of the target device.

[0085] Obtain the largest connected component within the bounding box of electrical equipment in a binarized image;

[0086] Based on the relative position information in the power equipment orientation frame of the target device, the component frame of the target device is determined according to the maximum connected component to complete the temperature visual positioning of the target device.

[0087] Specifically, the target device is first marked with a power equipment orientation box detection model. Then, the temperature values ​​of the insulators outside the power equipment orientation box in the image of the target device are set to 0. The image corresponding to the insulator temperature value matrix of the target device is binarized to obtain a binarized image of the target device. The largest connected component within the power equipment orientation box in the binarized image is obtained; that is, adjacent pixels with the same pixel value are found and marked, and then connected component analysis is performed to determine the component box of the target device, thereby completing the visual temperature localization of the target device.

[0088] like Figure 2 The diagram shown is a functional block diagram of a power equipment temperature visual positioning device according to an embodiment of the present invention.

[0089] The power equipment temperature visual positioning device 200 of this invention can be installed in an electronic device. Depending on the functions implemented, the power equipment temperature visual positioning device may include a matrix extraction module 210, a zero-filling module 220, a model training module 230, a fixed-frame detection module 240, and a temperature visual positioning module 250. The module described in this invention can also be called 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 are stored in the memory of the electronic device.

[0090] In this embodiment, the functions of each module / unit are as follows:

[0091] The matrix extraction module 210 is used to extract a temperature visual matrix from the collected insulator information data; wherein, in the temperature visual matrix, the value of each element corresponds to the actual temperature value of a region in the environment.

[0092] As an optional embodiment of the present invention, the matrix extraction module 210 further includes an acquisition unit and an extraction unit (not shown in the figure).

[0093] The acquisition unit is used to acquire information data of the insulator through an infrared thermal imager and obtain an image in the infrared universal data file storage format;

[0094] The extraction unit is used to extract a temperature visual matrix with a size of 640*480*1 from an image stored in the infrared general data file format.

[0095] The zero-fill module 220 is used to perform zero-fill processing on the top and bottom of the short side of the temperature vision matrix, so that the length of the short side of the temperature vision matrix is ​​equal to that of the long side, thus obtaining the filled temperature vision matrix.

[0096] As an optional embodiment of the present invention, the zero-fill module 220 further includes a filling unit and a shrinking unit (not shown in the figure).

[0097] The filling unit is used to perform zero-filling on the top and bottom of the short side of the temperature vision matrix, so that the size of the temperature vision matrix is ​​filled from 640*480*1 to 640*640*1, thus obtaining the primary filled temperature vision matrix.

[0098] The shrinking unit is used to shrink the range of numerical values ​​of the primary filled temperature visual matrix to 0~255, so as to obtain the filled temperature visual matrix.

[0099] The model training module 230 is used to train the basic model of the preset bounding box detection model by taking the filled temperature visual matrix as input, so as to obtain the bounding box detection model of power equipment; wherein, in the basic model, the rotation target detection technology is used to classify the angle of the bounding box.

[0100] As an optional embodiment of the present invention, the basic model uses the RetinaNet network as the framework and ResNet50 and FPN as the backbone feature extraction networks.

[0101] Specifically, the RetinaNet network structure in object detection algorithms uses ResNet as the backbone and FPN as the neck neural network, which provides good control over the number of training parameters. Overall, the RetinaNet network structure is simple and clear.

[0102] As an optional embodiment of the present invention, in the basic model, the classification of the angle of the bounding box using rotational target detection technology includes:

[0103] The orientation frame information of the insulator is represented by the long side representation method, where c_x and c_y represent the center coordinates of the rotating rectangle; longside represents the longest side; shortside represents the short side; theta represents the angle between the longest side and the x-axis in a counterclockwise rotation. The counterclockwise angle is negative, and theta ranges from -180° to 0°.

[0104] As an optional embodiment of the present invention, the model training module 230 further includes: a data allocation unit, a training unit, and a testing unit (not shown in the figure).

[0105] The data allocation unit is used to divide the dataset of power equipment temperature values ​​corresponding to the filled temperature visual matrix into a training set and a test set in an 8:2 ratio;

[0106] The training unit is used to train the base model using the training set to obtain a primary training model;

[0107] The test unit is used to test the primary training model using a test set. When the loss curve of the primary training model meets the preset model training loss requirements, the power equipment orientation box detection model is obtained.

[0108] As an optional embodiment of the present invention, the evaluation method for the power equipment bounding box detection model includes: precision and recall; wherein,

[0109] The formula for calculating accuracy is: P = TP / (TP + FP);

[0110] The recall rate is calculated using the formula: R = TP / (TP + FN); where,

[0111] TP represents the number of devices correctly identified as electrical equipment, FP represents the number of devices incorrectly identified as electrical equipment, and FN represents the number of devices incorrectly identified as non-electrical equipment.

[0112] The phase frame detection module 240 is used to obtain the power equipment orientation frame of the target equipment by inputting the insulator temperature value matrix data of the target equipment to be detected into the power equipment orientation frame detection model.

[0113] The temperature visual positioning module 250 is used to determine the component frame of the target device based on the relative position information in the power equipment orientation frame of the target device and the maximum connected component extracted from the power equipment orientation frame of the target device, so as to complete the temperature visual positioning of the target device.

[0114] As an optional embodiment of the present invention, the temperature visual positioning module 250 further includes an image processing unit, a connected component acquisition unit, and a component box determination unit (not shown in the figure).

[0115] The image processing unit is used to set the temperature value of the insulator outside the orientation frame of the power equipment located in the target device to 0, and to perform binarization processing on the image corresponding to the insulator temperature value matrix of the target device to obtain a binarized image of the target device.

[0116] The connected component acquisition unit is used to acquire the largest connected component within the bounding box of the power equipment in the binarized image.

[0117] The component frame determination unit is used to determine the component frame of the target device based on the relative position information in the power equipment orientation frame of the target device and the maximum connected component, so as to complete the temperature visual positioning of the target device.

[0118] It should be understood that the embodiments described are for illustrative purposes only and are not limited to this structure in the scope of the patent application.

[0119] The power equipment temperature visual positioning program stored in the memory of the electronic device is a combination of multiple instructions, which, when run in the processor, can achieve the following:

[0120] A temperature visual matrix is ​​extracted from the collected insulator information data; in the temperature visual matrix, the value of each element corresponds to the actual temperature value of a region in the environment.

[0121] The short side of the temperature visual matrix is ​​zero-filled both above and below, so that the length of the short side is equal to that of the long side, resulting in the filled temperature visual matrix.

[0122] The filled temperature visual matrix is ​​used as input to train the base model of the preset bounding box detection model, thus obtaining the power equipment bounding box detection model. In the base model, rotation target detection technology is used to classify the angle of the bounding box.

[0123] The insulator temperature value matrix data of the target device to be detected is input into the power equipment orientation frame detection model to obtain the power equipment orientation frame of the target device;

[0124] Based on the relative position information in the power equipment orientation frame of the target device, and according to the maximum connected component extracted from the power equipment orientation frame of the target device, the component frame of the target device is determined to complete the temperature visual positioning of the target device.

[0125] Specifically, the processor's implementation method for the above instructions can be found in [reference needed]. Figure 1 The descriptions of the relevant steps in the corresponding embodiments are not repeated here.

[0126] Furthermore, if the modules / units integrated into the electronic device 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 medium may include: any entity or device 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).

[0127] In the several embodiments provided by this invention, it should be understood that the disclosed devices, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of modules is only a logical functional division, and other division methods may be used in actual implementation.

[0128] 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.

[0129] 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.

[0130] 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.

[0131] Therefore, the embodiments should be considered exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of equivalents of the claims are intended to be embraced within the invention. No appended diagram markings in the claims should be construed as limiting the scope of the claims.

[0132] Furthermore, it is clear that the word "comprising" does not exclude other units or steps, and the singular does not exclude the plural. Multiple units or devices recited in a system claim may also be implemented by a single unit or device through software or hardware. The term "second class" is used to indicate names and does not indicate any specific order.

[0133] 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 visually locating the temperature of power equipment, characterized in that, The method includes: A temperature visual matrix is ​​extracted from the collected insulator information data; wherein, in the temperature visual matrix, the value of each element corresponds to the actual temperature value of a region in the environment. The extraction of the temperature visual matrix from the collected insulator information data includes: The information data of the insulator is collected by an infrared thermal imager, and an image in the infrared universal data file storage format is obtained. Extract a temperature visual matrix with a size of 640*480*1 from the image in the infrared universal data file storage format; The short side of the temperature visual matrix is ​​zero-filled both above and below, so that the length of the short side is equal to that of the long side, thus obtaining the filled temperature visual matrix. The step of zero-filling the short side of the temperature visual matrix above and below, so that the length of the short side is equal to the length of the long side, results in a filled temperature visual matrix including: The short side of the temperature visual matrix is ​​zero-filled at the top and bottom, so that the size of the temperature visual matrix is ​​filled from 640*480*1 to 640*640*1, thus obtaining the primary filled temperature visual matrix. The numerical value range of the primary filled temperature visual matrix is ​​narrowed to 0~255 to obtain the filled temperature visual matrix; The filled temperature visual matrix is ​​used as input to train the base model of the preset bounding box detection model to obtain the power equipment bounding box detection model; wherein, in the base model, the angle of the bounding box is classified by rotation target detection technology; The classification of the bounding box angle using rotational target detection technology in the basic model includes: The orientation frame information of the insulator is represented by the long side representation method, where c_x and c_y represent the center coordinates of the rotated rectangle; longside represents the longest side; shortside represents the short side; theta represents the angle between the longest side and the x-axis in a counterclockwise rotation, with the counterclockwise angle being negative, and theta range being -180° to 0°. The insulator temperature value matrix data of the target device to be detected is input into the power equipment orientation frame detection model to obtain the power equipment orientation frame of the target device; Based on the relative position information in the power equipment orientation frame of the target device, and according to the maximum connected component extracted from the power equipment orientation frame of the target device, the component frame of the target device is determined to complete the temperature visual positioning of the target device; The step of determining the component frame of the target device based on the relative position information in the power equipment orientation frame of the target device, and according to the maximum connected component extracted from the power equipment orientation frame of the target device, to complete the temperature visual localization of the target device includes: Set the temperature value of the insulator outside the power equipment orientation frame of the target device to 0, and binarize the image corresponding to the insulator temperature value matrix of the target device to obtain a binarized image of the target device. Obtain the largest connected component within the bounding box of the power equipment in the binarized image; Based on the relative position information in the power equipment orientation frame of the target device, the component frame of the target device is determined according to the maximum connected component to complete the temperature visual positioning of the target device.

2. The method for visually locating the temperature of power equipment according to claim 1, characterized in that, The basic model uses RetinaNet as its framework and ResNet50 and FPN as its backbone feature extraction networks.

3. The method for visually locating the temperature of power equipment according to claim 1, characterized in that, The step of using the filled temperature visual matrix as input to train the basic model of the preset bounding box detection model to obtain the power equipment bounding box detection model includes: The dataset of power equipment temperature values ​​corresponding to the filled temperature visual matrix is ​​divided into a training set and a test set in an 8:2 ratio; The basic model is trained using the training set to obtain a primary training model; The primary training model is tested using the test set. When the loss curve of the primary training model meets the preset model training loss requirements, a power equipment bounding box detection model is obtained.

4. The method for visually locating the temperature of power equipment according to claim 1, characterized in that, The evaluation method for the power equipment bounding box detection model includes: precision and recall; wherein, The formula for calculating the accuracy rate is: P = TP / (TP + FP); The recall rate is calculated using the formula: R = TP / (TP + FN); where, TP represents the number of devices correctly identified as electrical equipment, FP represents the number of devices incorrectly identified as electrical equipment, and FN represents the number of devices incorrectly identified as non-electrical equipment.

5. A power equipment temperature visual positioning device, used to implement the power equipment temperature visual positioning method according to any one of claims 1-4, characterized in that, The device includes: The matrix extraction module is used to extract a temperature visual matrix from the collected insulator information data; wherein, in the temperature visual matrix, the value of each element corresponds to the actual temperature value of a region in the environment. The zero-fill module is used to perform zero-fill processing on the top and bottom of the short side of the temperature visual matrix, so that the length of the short side of the temperature visual matrix is ​​equal to that of the long side, thus obtaining the filled temperature visual matrix. The model training module is used to train the base model of the preset bounding box detection model by taking the filled temperature visual matrix as input, so as to obtain the bounding box detection model of power equipment; wherein, in the base model, the angle of the bounding box is classified by rotating target detection technology; The phase frame detection module is used to obtain the power equipment orientation frame of the target equipment by inputting the insulator temperature value matrix data of the target equipment to be detected into the power equipment orientation frame detection model; The temperature visual positioning module is used to determine the component frame of the target device based on the relative position information in the power equipment orientation frame of the target device and according to the maximum connected component extracted from the power equipment orientation frame of the target device, so as to complete the temperature visual positioning of the target device.

6. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the power equipment temperature visual positioning method as described in any one of claims 1 to 4.

Images