Target geometric center point temperature measurement research method
By using the target geometric center point temperature measurement research method in the substation, using neural network and infrared images, combined with the rotation angle and aspect ratio of the rectangular frame, the problem of low accuracy of manual measurement temperature in the substation is solved, achieving more efficient and Accurate temperature detection.
Patent Information
- Application Number
- CN202411093771.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-09
- Publication Date
- 2024-11-01
AI Technical Summary
In substations, when manually measuring the temperature of a device, due to the long distance and environmental factors, the measurement accuracy and efficiency are low, especially when multiple target devices overlap.
The temperature measurement research method of target geometric center point is adopted, and the temperature of the device is determined by obtaining infrared images and inputting them into the trained neural network, and the rotation angle information and aspect ratio of the rectangular box are used to determine the temperature of the device, thereby improving the accuracy and efficiency of temperature detection.
The accuracy and efficiency of temperature detection of target devices are improved, especially when multiple devices overlap, the center position of the rectangular frame and the pixel temperature can more accurately reflect the temperature of the device, reducing the impact of non-target areas.
Smart Images

Figure CN118887393A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of temperature measurement and a method for studying temperature measurement of a target geometric center point. Background Art
[0002] When the temperature of various components in the substation, such as insulators, knife switches, and infrared clamps, is too high, the following hazards may occur: High temperature will accelerate the aging and wear of electrical equipment, and may cause failure or damage to components such as insulators and knife switches. For example, insulators may cause aging or breakdown of insulation materials under high temperatures, and knife switches may cause oxidation or melting of contacts due to high temperatures. Electrical equipment may cause fires under high temperature conditions. For example, electrical contacts or wires may ignite surrounding insulation materials under excessive temperatures, causing fires.
[0003] In the prior art, the temperature of each device is usually measured manually, for example, a relevant person can measure the temperature of the device by using a temperature measuring gun or other equipment. When the temperature of one or some devices is too high, corresponding measures are taken to reduce the temperature of the device.
[0004] However, since the devices in the substation are usually far from the ground, and manual measurement of the device temperature is usually affected by the surrounding environmental factors, the accuracy and efficiency of manual temperature measurement of the devices far from the ground are low. Summary of the invention
[0005] In order to solve the above-mentioned technical problem of low accuracy in temperature measurement of devices, the present invention proposes a research method for measuring temperature at a target geometric center point.
[0006] In a first aspect of an embodiment of the present invention, a method for measuring temperature at a target geometric center point is provided, comprising: Acquire infrared images of substation components; The infrared image is input into the trained neural network to obtain the positions of the four vertices of the rectangular box corresponding to the target device in the substation in the infrared image and the position of the center point of the rectangular box. If the aspect ratio of the rectangular box is within a preset range, the temperature corresponding to the center point position and the temperature of each pixel in the rectangular box are determined; The training process of the neural network is: Acquire multiple historical infrared images containing target devices, input the multiple historical infrared images into a neural network for training, and obtain a trained neural network, wherein each target device in the historical infrared image has a corresponding label, and the label includes position information of a rectangular frame surrounding the target device and a category of the target device; The position information of the rectangular frame includes the rotation angle information of the rectangular frame.
[0007] The present invention detects the temperature corresponding to the target device in the infrared image through the trained neural network. This process is automatic, thereby improving the efficiency of detecting the temperature of the target device.
[0008] Because at a certain shooting angle, multiple target devices of the same type may overlap in the image, that is, there is a "blocking" situation between the multiple target devices. For example, the category of these multiple target devices is insulator. At this time, in the process of determining the temperature corresponding to the target device, the rectangular frame corresponding to the target device is likely to surround these multiple target devices, which leads to the possibility that the center position of the rectangular frame is not the center position of the target device, and even the center position of the rectangular frame may be a position in the non-target device area, resulting in the temperature corresponding to the center position of the rectangular frame may be the temperature corresponding to a position in the non-target device area, resulting in the temperature corresponding to the center position of the rectangular frame reflecting the temperature of the target device with poor accuracy. When multiple target devices of the same type overlap in the image, it is more likely that the area of the non-target device area included in the rectangular frame is large. When multiple target devices of the same type overlap in the image, the aspect ratio of the rectangular frame may be too large or too small.
[0009] In order to improve the accuracy of measuring the temperature of the target device, the temperature corresponding to the center point of the rectangular frame and the temperature of each pixel in the rectangular frame are output only when the aspect ratio of the rectangular frame is within a preset range. Among them, the temperature corresponding to the center point of the rectangular frame can more accurately reflect the temperature of the center position of the target device, and the temperature of each pixel in the rectangular frame can more accurately reflect the temperature of the target device.
[0010] During the training process, since the label of each target device corresponds to the rotation angle information of the rectangular frame surrounding the target device, where the rotation angle information is obtained by manually marking the rectangular frame of the target device using relevant software, the difference between the area of the rectangular frame and the area of the corresponding target device in the infrared image is small. In other words, the rectangular frame "fits" the target device to a high degree, and the rectangular frame "surrounds" the target device better. As a result, the center position of the rectangular frame can better reflect the center position of the target device, and the position of the rectangular frame can better reflect the position of the target device. When measuring the temperature of the target device, the temperature corresponding to the center position of the rectangular frame is used as the temperature of the center position of the target device, and the temperature corresponding to all pixels in the rectangular frame area is used as the temperature of the target device, which can improve the accuracy of measuring the temperature of the target device.
[0011] Optionally, the condition for determining the temperature corresponding to the center point position and the temperature of each pixel in the rectangular frame also includes: the standard deviation of the temperature corresponding to all pixel points in the rectangular frame corresponding to the target device in the infrared image is less than a preset discrete threshold. The pixel points correspond to the temperature one by one.
[0012] When multiple target devices overlap in an image, it is highly likely that the non-target device region in the rectangular frame in the infrared image is larger in area.
[0013] The temperature distribution of the target device is relatively uniform, and there is a high possibility that the temperature of the target device is significantly different from the temperature of the surrounding non-target device area. In the infrared image, there is a high possibility that the temperature corresponding to the target device area is significantly different from the temperature corresponding to the surrounding non-target device area.
[0014] Based on this, if the standard deviation of the temperature corresponding to the pixel points in the rectangular frame corresponding to the target device in the infrared image is smaller, the possibility that the area of the non-target device region in the rectangular frame is smaller is greater, and the possibility of multiple target devices overlapping in the image in the rectangular frame is smaller.
[0015] Optionally, the position information of the rectangular frame also includes positions of four vertices of the rectangular frame, a width of the rectangular frame, a height of the rectangular frame and a center position of the rectangular frame.
[0016] Optionally, the positions of the four vertices of the rectangular frame are determined according to the width, height, rotation angle information and center position of the rectangular frame.
[0017] Optionally, the width, height, rotation angle information and center position of the rectangular frame are obtained by manually labeling the rectangular frame of the target device through rolabelImg.
[0018] The position and size of the rectangular frame correspond to the rotation angle, width, height and center position of the rectangular frame. Therefore, the positions of the four vertices of the rectangular frame can be determined according to the rotation angle, width, height and center position of the rectangular frame.
[0019] Optionally, the neural network is YOLOv5.
[0020] Optionally, the neural network includes YOLOv5s.
[0021] Optionally, the tag also includes a difficult parameter.
[0022] Optionally, the temperature is obtained according to an infrared temperature measurement algorithm.
[0023] In a second aspect of an embodiment of the present invention, a computer-readable storage medium is provided, in which program instructions are stored. When the program instructions are loaded and executed by a processor, the processor executes a target geometric center point temperature measurement research method according to any one of the first aspects of the embodiment of the present invention.
[0024] Beneficial effects of the present invention: The present invention detects the temperature of the target device through a trained neural network, thereby improving the detection efficiency. The influence of the situation where multiple target devices overlap in an image on the detection of the temperature of the target device is considered, thereby improving the accuracy of detecting the temperature of the target device. Moreover, in the process of training the neural network, since the label of the target device includes the rotation angle information of the rectangular frame surrounding the target device, the difference between the area of the rectangular frame and the area of the target device is small, and thus the difference between the center position of the rectangular frame and the center position of the target device is small. In the process of predicting the temperature of the target device, the temperature of the center position of the rectangular frame is used as the temperature of the center position of the target device, thereby improving the accuracy of measuring the temperature of the target device.
[0025] It is to be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the invention. BRIEF DESCRIPTION OF THE DRAWINGS
[0026] By reading the following detailed description with reference to the accompanying drawings, the above and other objects, features and advantages of the exemplary embodiments of the present invention will become readily understood. In the accompanying drawings, several embodiments of the present invention are shown in an exemplary and non-limiting manner, and the same or corresponding reference numerals represent the same or corresponding parts, wherein: Figure 1 It schematically illustrates a target geometric center point temperature measurement research method according to an embodiment of the present invention; Figure 2 is a rectangular frame schematic diagram schematically illustrating an embodiment of the present invention; Figure 3 FIG. 4 is a schematic diagram schematically showing the overlapping of multiple insulators according to an embodiment of the present invention. DETAILED DESCRIPTION
[0027] In some of the processes described in the specification and claims of the present invention and the above-mentioned figures, multiple operations appearing in a specific order are included, but it should be clearly understood that these operations may not be executed in the order in which they appear in this article or executed in parallel. The serial numbers of the operations, such as 101, 102, etc., are only used to distinguish different operations, and the serial numbers themselves do not represent any execution order. In addition, these processes may include more or fewer operations, and these operations may be executed in sequence or in parallel. It should be noted that the descriptions of "first", "second", "A1", "B1", etc. in this article are used to distinguish different messages, devices, modules, etc., and do not represent the order of precedence, nor do they limit "A1", "B1", "first" and "second" to be different types; in other words, the serial numbers themselves for the components in this article, such as "A1", "B1", "first", "second", etc., are only used to distinguish the objects described, and do not have any order or technical meaning. For example: the first feature of an image is the feature of the image, where the first has no technical meaning and is only used to distinguish objects. The role of adding "first" includes: facilitating narration, understanding and explanation, and facilitating citing different objects, which can make the article more concise and clear. For example: the first feature and the second feature are different features.
[0028] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present invention.
[0029] The specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.
[0030] The present invention first trains the neural network to obtain a trained neural network, wherein each target device in the historical infrared image during the training process corresponds to a label, and the label includes the rotation angle information of the rectangular frame surrounding the target device. It should be noted that the rotation angle information makes the rectangular frame surround the corresponding target device better, that is, in the infrared image, the difference between the area of the rectangular frame and the area of the corresponding target device is small. Then the infrared image to be detected is input into the trained neural network to obtain the position information of the target device in the infrared image. If the position information and the temperature of the pixel point in the rectangular frame corresponding to the position information meet certain conditions, the temperature corresponding to the position information is output, and the temperature reflects the temperature of the corresponding target device.
[0031] Figure 1 FIG. 1 schematically illustrates a method for measuring temperature at a target geometric center point according to an embodiment of the present invention. Figure 1 The specific instructions are as follows.
[0032] S1: Train the neural network to obtain a trained neural network.
[0033] First, multiple historical infrared images containing target devices in the substation are acquired.
[0034] If the temperature of some components in the substation is too high, these components may be damaged or even cause a fire. In order to prevent the temperature of the components from being too high, the temperature of the components can be monitored. The present invention monitors the temperature of the components by collecting infrared images containing the components.
[0035] For example, devices include knife switches, insulators, and infrared clamps.
[0036] In one embodiment, the target device is a device whose temperature is to be detected. For example, the target device may be an insulator. It should be noted that, for ease of understanding, the target device is an insulator as an example for description in the context.
[0037] In one embodiment, a plurality of historical infrared images containing the target device are taken by means of a drone.
[0038] In one embodiment, the inspection route of the drone is pre-specified.
[0039] In one embodiment, the infrared camera on the drone automatically captures multiple infrared images containing the target device, which are recorded as historical infrared images, wherein the historical infrared images are used in the training process of the neural network, and the implementer can set the shooting frequency according to the specific scene. In another embodiment, the infrared camera on the drone is manually controlled to capture multiple historical infrared images containing the target device.
[0040] It should be noted that the historical infrared images are infrared images under multiple lighting conditions and multiple weather conditions. Using the historical infrared images can improve the generalization ability of the neural network model.
[0041] Secondly, the multiple historical infrared images are input into a neural network for training to obtain a trained neural network. Wherein, each target device in the multiple historical infrared images corresponds to a label, wherein each target device corresponds to a label, and the label includes the position information of the rectangular frame surrounding the target device and the category of the target device, wherein the category is the name of the target device. Wherein, the position information of the rectangular frame includes the rotation angle information of the rectangular frame.
[0042] For example, for a target device in a historical infrared image, there is a corresponding label, which includes the position information of a rectangular frame surrounding the target device and the category of the target device, and the position information includes the rotation angle information of the rectangular frame.
[0043] The position information of the rectangular frame of the target device can be obtained by manually marking the rectangular frame of the target device using marking software.
[0044] The rotation angle information of the rectangular frame reflects the tilt information of the rectangular frame. Figure 2 Schematic diagram of a rectangular frame according to an embodiment of the present invention. Figure 2 As shown, 201 is a non-inclined rectangular frame, 202 is an inclined rectangular frame, and 203 is another inclined rectangular frame. It should be noted that the rectangular frames in the context all have a rotation angle.
[0045] In one embodiment, during the training process, the position information of the rectangular box corresponding to the label includes the positions of the four vertices of the rectangular box, the width of the rectangular box, the height of the rectangular box, the rotation angle of the rectangular box, and the center position of the rectangular box. The positions of the four vertices of the rectangular box are determined according to the width of the rectangular box, the height of the rectangular box, the rotation angle of the rectangular box, and the center position of the rectangular box.
[0046] For example, for a target device in a historical infrared image, there is a corresponding label, which includes the position information of the rectangular frame surrounding the target device, and the position information includes the positions of the four vertices of the rectangular frame, the width of the rectangular frame, the height of the rectangular frame, the rotation angle of the rectangular frame, and the center position of the rectangular frame. The positions of the four vertices of the rectangular frame are determined according to the rotation angle, width, height, and center position of the rectangular frame.
[0047] For example, in a historical infrared image, for a target device, based on rolabelImg, the rectangular frame of the target device is manually labeled to determine the rotation angle, width, height and center position of the rectangular frame corresponding to the target device. For example, corresponding to a target device, a rectangular frame is manually drawn using rolabelImg to surround the target device, and then the position, size and rotation angle of the rectangular frame can be adjusted so that the rectangular frame surrounds the target device and the area of the rectangular frame in the infrared image is as small as possible from the area of the target device. It should be noted that the rotation angle, width, height and center position of the rectangular frame are automatically generated by rolabelImg.
[0048] Among them, rolabelimg is an image annotation tool developed based on Python and Qt, which is a prior art and will not be described in detail here.
[0049] The principle of obtaining the rotation angle, width, height and center position of the rectangular frame corresponding to other target devices is the same as the above-mentioned obtaining principle. Then, the rotation angle, width, height and center position of the rectangular frame corresponding to all target devices can be obtained.
[0050] It should be noted that the file after marking the position of the target device is an XML tag file, and the marked XML tag file is converted, that is, the rotation angle, width, height and center position of the rectangular box corresponding to each target device are first read, and then the rotation angle, width, height and center position of the rectangular box corresponding to each target device are converted into the positions of the four vertices of the rectangular box. Since the rotation angle, width, height and center position of the rectangular box corresponding to the target device can uniquely determine the position and size of the rectangular box, the rotation angle, width, height and center position of the rectangular box corresponding to the target device can be converted into the positions of the four vertices of the rectangular box by conventional technology.
[0051] In one embodiment, the position of each point in the infrared image can be represented by the row and column information of the pixel point at the position. For example, the position of a point in the infrared image is the 5th row and the 6th column. In another embodiment, the position of each point in the infrared image can be represented by the coordinates of the pixel point at the position. The pixel point corresponds to the position of each point in the infrared image one by one.
[0052] In one embodiment, the length unit may be millimeters. In another embodiment, the distance between the centers of two horizontally adjacent pixels in the infrared image is taken as unit 1, and other distances in the infrared image are based on the unit 1.
[0053] At this point, the labels of each target device in each historical infrared image are obtained. The labels include the positions of the four vertices of the rectangular box and the category of the target device. The historical infrared images containing the label data are input into the neural network for training to obtain a trained neural network.
[0054] S2: Obtain an infrared image of the substation device, input the infrared image into the trained neural network, and obtain the temperature corresponding to the location of the target device in the substation.
[0055] In one embodiment, an infrared image including substation components may be captured by an infrared camera on a drone, wherein the infrared image may be an infrared image at the current moment.
[0056] Secondly, the infrared image is input into the trained neural network to obtain the positions of the four vertices of the rectangular box corresponding to the target device in the substation in the infrared image and the position of the center point of the rectangular box; if the aspect ratio of the rectangular box is within the preset range, the temperature corresponding to the center point position and the temperature of each pixel in the rectangular box are determined. Among them, each target device corresponds to a rectangular box, and the rectangular box surrounds the corresponding target device. The aspect ratio is the ratio of the length of the long side of the rectangular box to the length of the short side of the rectangular box. The length of the long side of the rectangular box is the height of the rectangular box. The temperature corresponding to the center point position is used to represent the temperature of the center position of the target device, and the temperature of each pixel in the rectangular box is used to represent the temperature of the target device.
[0057] Since the value of the preset range needs to be selected according to the actual situation, the specific range is not given here. The implementer can set the preset range according to the specific implementation scenario.
[0058] For ease of understanding, for example, if the target device is an insulator, the preset range is (3,6), and the aspect ratio of the rectangular frame is greater than 3 and less than 6, then the temperature corresponding to the center point position and the temperature of each pixel in the rectangular frame are determined.
[0059] It should be noted that by using an infrared temperature measurement algorithm on an infrared image, the temperature corresponding to the position of each point in the infrared image can be obtained, that is, the temperature corresponding to each pixel in the infrared image can be obtained.
[0060] For example, for an infrared image to be monitored, it is input into a trained neural network to obtain the rotation angle, width, height and center position of the rectangular box of each target device in the infrared image. The rotation angle, width, height and center position of the rectangular box can be converted into the positions of the four vertices of the rectangular box.
[0061] The following is an explanation. For a target device, if the aspect ratio of the rectangular frame corresponding to the target device is within a preset range, it is highly likely that the target device does not overlap with other devices of the same type in the infrared image, that is, it is highly likely that there is no "occlusion" between the target device and other devices of the same type, and the temperature corresponding to the center point of the rectangular frame reflects the temperature of the center position of the target device with greater accuracy, and the temperature of each pixel point in the rectangular frame reflects the temperature of the target device with greater accuracy.
[0062] In one embodiment, for any target device, if the aspect ratio of the rectangular box corresponding to the target device in the infrared image is within a preset range, and the standard deviation of the temperature corresponding to the pixel points in the rectangular box is less than a preset discrete threshold, then the temperature corresponding to the center point of the rectangular box and the temperature of each pixel point in the rectangular box are determined.
[0063] Since the value of the preset discrete threshold needs to be selected according to the actual situation, the specific value of the preset discrete threshold is not given here. The implementer can set the preset discrete threshold according to the specific implementation scenario.
[0064] The explanation is as follows: For a target device, if the standard deviation of the temperature of all pixel points in the rectangular box corresponding to the target device is less than the preset discrete threshold, where the pixel points correspond to the temperature one by one, it means that the area of the non-target device region in the rectangular box is likely to be small, the temperature corresponding to the center point of the rectangular box reflects the temperature of the center position of the target device with greater accuracy, and the temperature of each pixel point in the rectangular box reflects the temperature of the target device with greater accuracy.
[0065] For example, the target device may be an insulator, Figure 3 FIG. 2 is a schematic diagram schematically showing the overlapping of multiple insulators according to an embodiment of the present invention. Figure 3 As shown, 301 is a first insulator, 302 is a second insulator, 303 is a rectangular frame, and 304 is another rectangular frame.
[0066] There are a variety of infrared temperature measurement algorithms in the prior art, and the infrared temperature measurement algorithm is not limited here. The implementer can select the infrared temperature measurement algorithm according to the specific implementation scenario.
[0067] Implementers can choose the neural network model according to the specific scenario.
[0068] In one embodiment, the neural network is YOLOv5.
[0069] In one embodiment, if the background in the infrared image is relatively simple, in order to pursue detection efficiency, YOLOv5s can be selected for the neural network.
[0070] Implementers can choose the type of neural network based on specific scenarios.
[0071] In one embodiment, the tag further includes a difficulty parameter, wherein the difficulty parameter is a tool for indicating the difficulty level of a sample in the target detection task.
[0072] In one embodiment, some historical infrared images are selected as training sets, and other historical infrared images are selected as verification sets, wherein each target device in the historical infrared image is marked with a rectangular box, and the ratio of the total number of historical infrared images in the training set to the total number of historical infrared images in the verification set is 4:1.
[0073] In one embodiment, during the process of training the neural network model, the model is continuously updated, that is, the optimal model in the training process is used as the updated model.
[0074] In another embodiment, the implementer may determine whether it is necessary to supplement the data used for training or to increase the number of training iterations based on the total number of data used for training and the actual test situation. For example, if the total number of historical infrared images used for training is too small, the implementer may supplement the historical infrared images and increase the number of training iterations; if the actual test situation does not meet expectations, the implementer may supplement the historical infrared images and increase the number of training iterations.
[0075] In one embodiment, a preset number of epochs (training cycles) is set, for example, 10,000 epochs can be set. During the training process, the accuracy of the test is verified using the verification set in each epoch. If the accuracy corresponding to a continuous number of epochs reaches a preset accuracy threshold, and the accuracy corresponding to the continuous number of epochs is less than the accuracy corresponding to the epochs before the continuous number of epochs, the iteration is stopped to obtain a trained neural network model. The implementer can set the preset accuracy threshold according to the specific implementation scenario.
[0076] In another embodiment, in order to ensure the accuracy of the network model, the richness of the samples is first ensured, and multiple historical infrared images corresponding to each photo point are collected at multiple time points, and then a labeling training test is performed. If the test result does not meet the requirements, the historical infrared images continue to be collected, wherein the historical infrared images of multiple photo points are respectively input into the neural network model to obtain output results, and targeted images are collected for the photo points corresponding to the output results with poor accuracy.
[0077] In one embodiment, during the training of the neural network model, if a single category of target devices is trained, the None category is added during the training process to improve the generalization ability of the model.
[0078] In order to more accurately determine the center position of each target device and the position of each target device area in the infrared image, the present invention takes into account the rotation angle of the rectangular frame during the neural network training process, and also takes into account interference factors such as the occlusion between multiple target devices of the same type. Based on the center position of each target device and the rectangular frame corresponding to each target device, the temperature corresponding to the center position and the temperature of the pixel point in the rectangular frame are obtained.
[0079] In the description of this specification, "plurality" or "several" means at least two, such as two, three or more, etc., unless otherwise clearly defined.
[0080] As used herein, the term "or" is intended to mean an inclusive "or" rather than an exclusive "or." Additionally, the articles "a" and "an" as used in this application and the appended claims are generally understood to mean "one or more" unless specified otherwise or clear from context to be directed to a singular form.
[0081] Where the terms "includes," "having," "having," "having," or variations thereof are used in the detailed description or claims, such terms are intended to be inclusive in a manner similar to the term "comprising."
[0082] Although this specification has shown and described a number of embodiments of the present invention, it will be apparent to those skilled in the art that such embodiments are provided by way of example only. Those skilled in the art will conceive of many modifications, changes and alternatives without departing from the ideas and spirit of the present invention. It should be understood that in the practice of the present invention, various alternatives to the embodiments of the present invention described herein may be employed.
Claims
1. A research method for measuring temperature at the target geometric center point. It is characterized in that include: Acquire infrared images of substation components; The infrared image is input into the trained neural network to obtain the positions of the four vertices of the rectangular box corresponding to the target device in the substation in the infrared image and the position of the center point of the rectangular box. If the aspect ratio of the rectangular box is within a preset range, the temperature corresponding to the center point position and the temperature of each pixel in the rectangular box are determined; The training process of the neural network is: Acquire multiple historical infrared images containing target devices, input the multiple historical infrared images into a neural network for training, and obtain a trained neural network, wherein each target device in the historical infrared image has a corresponding label, and the label includes position information of a rectangular frame surrounding the target device and a category of the target device; The position information of the rectangular frame includes the rotation angle information of the rectangular frame.
2. A target geometric center point temperature measurement research method according to claim 1, It is characterized in that The condition for determining the temperature corresponding to the center point position and the temperature of each pixel point in the rectangular frame also includes: a standard deviation of the temperature corresponding to the pixel points in the rectangular frame corresponding to the target device in the infrared image is less than a preset discrete threshold.
3. A target geometric center point temperature measurement research method according to claim 1, It is characterized in that The position information of the rectangular frame also includes the positions of the four vertices of the rectangular frame, the width, height and center position of the rectangular frame.
4. A target geometric center point temperature measurement research method according to claim 3, It is characterized in that The positions of the four vertices of the rectangular frame are determined according to the width, height, rotation angle information and center position of the rectangular frame.
5. A target geometric center point temperature measurement research method according to claim 3, It is characterized in that The width, height, rotation angle information and center position of the rectangular frame are obtained by manually labeling the rectangular frame of the target device through rolabelImg.
6. A target geometric center point temperature measurement research method according to claim 1, It is characterized in that The tag also includes a difficult parameter.
7. A target geometric center point temperature measurement research method according to claim 1, It is characterized in that The temperature is obtained according to an infrared temperature measurement algorithm.
8. A computer-readable storage medium, It is characterized in that Program instructions are stored therein, and when the program instructions are loaded and executed by a processor, the processor executes a target geometric center point temperature measurement research method according to any one of claims 1 to 7.
