Intelligent temperature measurement system of temperature measurement type overhead transmission line image video monitoring device

Abstract

The invention provides an intelligent temperature measurement system of a temperature measurement type overhead transmission line image video monitoring device, and the system comprises a dual-light image fusion module which is used for generating a fusion image of an infrared image and a visible light image; the target detection module is used for identifying a target object in the image by adopting a target detection algorithm; the intelligent distance measuring module is used for calculating the actual distance between the device and the target; the environment parameter acquisition module is used for acquiring environment temperature data of an area where the cable terminal or the lightning arrester is located; the temperature measurement correction module is used for correcting an infrared temperature measurement error by using a multivariate nonlinear regression model based on Lasso regularization and outputting a corrected temperature value; and the image splicing module is used for generating a complete infrared image covering a plurality of cable terminals or lightning arresters. According to the invention, the problems of poor image information acquisition capability, insufficient target detection precision, large temperature measurement error, limited field angle coverage, single alarm mechanism, low system intelligence degree and the like in the prior art can be solved.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent monitoring and operation and maintenance of power systems, and specifically to an intelligent temperature measurement system for a temperature measurement type overhead transmission line image and video monitoring device, which is specifically applied to a transmission line image and video monitoring device that integrates infrared and visible light image processing, has target detection, intelligent ranging, temperature correction and image stitching functions, and supports local intelligent analysis and remote communication. Background Art

[0002] As a crucial component of the power system, the operational status of transmission lines is directly linked to the safety and stability of the power grid. Traditional transmission line monitoring methods rely primarily on regular manual inspections and fixed-point photography. While these methods can detect equipment hazards to a certain extent, they still suffer from issues such as slow response, limited coverage, low efficiency, and human misjudgment. Especially in complex terrain or large power grid areas, manual inspections are not only costly and time-consuming, but also pose significant safety risks, making them difficult to meet the "intelligent, real-time, and precise" operation and maintenance requirements of modern power grids.

[0003] With the continuous expansion of power systems and the increasingly complex operating environment of transmission lines, the safe and stable operation of overhead transmission lines faces severe challenges. To promptly detect temperature anomalies, equipment aging, poor contact, and other issues during transmission line operation, non-contact temperature measurement using infrared thermal imaging technology has become a key method of power line inspection.

[0004] Existing infrared monitoring systems for power transmission lines primarily use fixed or mobile infrared cameras to periodically measure the temperature of power equipment. Some systems incorporate artificial intelligence target detection and automatic alarm functions. However, existing technologies generally have the following shortcomings:

[0005] 1) Limited image information acquisition: Infrared images have low resolution and insufficient details. Target detection errors are prone to occur under low illumination or complex background conditions, affecting the accurate positioning of the temperature measurement area.

[0006] 2) Limited target detection accuracy: Traditional target detection methods are ineffective for identifying rotating or tilted equipment such as cable terminals and lightning arresters, and are difficult to adapt to target deformation under multi-angle shooting;

[0007] 3) Large temperature measurement errors: Affected by factors such as ranging errors, ambient temperature changes, and equipment model differences, infrared temperature measurement results have large deviations, making it difficult to meet the needs of precise monitoring of power equipment;

[0008] 4) Insufficient field of view: Infrared cameras typically have a small field of view and cannot cover multiple monitoring targets at once. Regional monitoring requires manual adjustment of the viewing angle or deployment of multiple devices.

[0009] 5) Imperfect alarm mechanism: Existing systems often use a single temperature threshold alarm, failing to make comprehensive judgments based on multiple factors such as device type, ambient temperature, and historical temperature rise trends, resulting in frequent false alarms or missed alarms.

[0010] 6) Limited intelligence: The system relies heavily on central servers for image processing and judgment analysis, and has problems such as long response delay, strong network dependence, and weak edge processing capabilities, which is not conducive to real-time monitoring needs in remote or edge areas.

[0011] In summary, there is an urgent need for an overhead transmission line image and video monitoring device with multimodal image fusion, precise target detection, non-contact ranging, high-precision temperature correction, image stitching and edge intelligent processing capabilities to improve the intelligence and practicality of the system and meet the temperature measurement needs of power equipment in complex scenarios. Summary of the Invention

[0012] In response to the problems existing in the prior art, such as poor image information acquisition capability, insufficient target detection accuracy, large temperature measurement errors, limited field of view coverage, single alarm mechanism, and low system intelligence, the present invention provides an intelligent temperature measurement system for a temperature measurement-type overhead transmission line image and video monitoring device, which is suitable for continuous, high-precision, and automated thermal status monitoring and alarming of overhead transmission lines and their key power equipment (such as cable terminals, lightning arresters, etc.).

[0013] The present invention achieves the above-mentioned purpose through the following technical solutions:

[0014] An intelligent temperature measurement system for a temperature measurement type overhead transmission line image and video monitoring device, comprising:

[0015] The dual-light image fusion module is used to collect infrared images and visible light images, align and fuse them, and generate a fused image;

[0016] The target detection module is used to perform image processing on the fused image, identify the target object in the image using a target detection algorithm, and output the position information and category information of the target object in the image;

[0017] The intelligent distance measurement module obtains the pixel size of the target in the image based on the target position information output by the target detection module, and calculates the actual distance between the device and the target in combination with the preset actual physical size of the target;

[0018] An environmental parameter acquisition module is used to obtain the ambient temperature data of the area where the cable terminal or lightning arrester is located, wherein the ambient temperature data is obtained by fusing the temperature measurement data with the temperature data of the non-hot spot area in the infrared image;

[0019] The temperature measurement correction module is used to take the original temperature measurement value, the actual distance of the target and the ambient temperature data as input, use the multivariate nonlinear regression model based on Lasso regularization to correct the infrared temperature measurement error, and output the corrected temperature value;

[0020] The image stitching module is used to control the pan-tilt head to rotate horizontally when the single-frame field of view of the infrared camera cannot cover all monitoring targets. The rotation angle range matches the field of view of the infrared camera, and infrared images covering the left, center and right directions are obtained in sequence. The three acquired infrared images are pre-processed and then stitched horizontally to generate a complete infrared image covering multiple cable terminals or lightning arresters.

[0021] According to an intelligent temperature measurement system of a temperature measurement type overhead transmission line image and video monitoring device provided by the present invention, the dual-light image fusion module includes:

[0022] The image enhancement submodule is used to enhance the brightness, contrast and noise of visible light images and infrared images respectively;

[0023] Image alignment submodule, used to spatially align visible light images with infrared images based on image feature points or geometric correction models;

[0024] The fusion processing submodule uses weighted average or maximum strategy to perform pixel-level fusion of the two images;

[0025] The color correction submodule is used to adjust the color temperature and balance the brightness of the fused image to ensure image quality consistency.

[0026] According to the intelligent temperature measurement system of the temperature measurement type overhead transmission line image and video monitoring device provided by the present invention, for the visible light image, the image enhancement submodule performs the following enhancement operations:

[0027] First, the image is grayscaled, which is expressed as the following formula:

[0028] I(x,y)=0.30R(x,y)+0.59G(x,y)+0.11B(x,y)

[0029] Where I(x,y) is the pixel value at the yth row and xth column of the grayscale image, R(x,y) is the pixel value at the yth row and xth column of the blue channel in the RGB image, G(x,y) is the pixel value at the yth row and xth column of the green channel in the RGB image, and B(x,y) is the pixel value at the yth row and xth column of the red channel in the RGB image.

[0030] Then, the image contrast is enhanced using a linear contrast stretching function, which is expressed as the following formula:

[0031] I enh(x,y)=α·I(x,y)+β

[0032] Among them, I enh (x, y) is the pixel value after image contrast enhancement, α is the contrast enhancement coefficient, and β is the brightness shift coefficient.

[0033] Next, a median filter is used to suppress image noise, which is expressed as the following formula:

[0034]

[0035] in, is a sliding window centered at (x,y).

[0036] According to the intelligent temperature measurement system of the temperature measurement type overhead transmission line image and video monitoring device provided by the present invention, for infrared image enhancement, the image enhancement submodule performs the following enhancement operations:

[0037] A linear contrast stretching function is used to perform linear contrast stretching, and adaptive normalization is performed according to the dynamic range of the image histogram to highlight the edge of the thermal distribution;

[0038] Wherein, the adaptive normalization includes: assuming the infrared image is I ir (x,y), the minimum pixel value is I min , the maximum value is I max , then the enhanced image is normalized according to the following formula:

[0039]

[0040] According to the intelligent temperature measurement system of the temperature measurement type overhead transmission line image video monitoring device provided by the present invention, after completing the spatial registration, the fusion processing submodule adopts a weighted average strategy to fuse the two images at the pixel level, which is expressed as the following formula:

[0041] I fused (x,y)=w1·I(x,y)+w2·I ir (x,y)

[0042] Among them, I fused (x, y) is the pixel value in the fused image, w1+w2=1.

[0043] According to the intelligent temperature measurement system of the temperature measurement type overhead transmission line image and video monitoring device provided by the present invention, after the fusion image is generated, the color correction submodule is used to perform brightness balance and color temperature adjustment:

[0044] Histogram equalization is used to enhance the consistency of brightness distribution across the entire image, which can be expressed as the following formula:

[0045]

[0046] Where h(i) is the number of pixels at the i-th grayscale level in the grayscale histogram of the image pixels, and L is the total number of grayscale levels.

[0047] According to the intelligent temperature measurement system of a temperature measurement overhead transmission line image and video monitoring device provided by the present invention, the target detection module uses a fused image as input and uses YOLOv11 as a deep learning detection network to build a target detection model. The target detection model is provided with an S-Ghost module in the backbone network to replace the first two general convolutional layers. Feature extraction is performed on the backbone network through the S-Ghost module. An RSK module with an attention mechanism is provided at the starting position of the neck network to realize the recognition of inclined targets such as cable terminals or lightning arresters. When performing image annotation, the rolabelimg annotation tool is used to annotate the cable terminals in the image with a rotated target box, and the annotated image is data enhanced to establish a sample data set.

[0048] The output of the target detection model includes: target category label, two-dimensional coordinates of the target in the fused image, rotation angle information and recognition confidence score.

[0049] According to the intelligent temperature measurement system of a temperature measurement-type overhead transmission line image and video monitoring device provided by the present invention, the intelligent ranging module adopts a distance measurement method based on the proportional relationship between the actual physical size of the target object and the pixel size in the image, and combines the focal length and pixel size of the camera to establish a functional relationship between the pixel size and the actual distance, which is used to realize non-contact target distance calculation under different shooting conditions.

[0050] According to an intelligent temperature measurement system of a temperature measurement type overhead transmission line image and video monitoring device provided by the present invention, the environmental parameter acquisition module includes:

[0051] Infrared image analysis submodule, used to extract temperature data of non-heating areas in infrared images;

[0052] The temperature sensor acquisition submodule is used to collect real-time ambient temperature data from temperature sensors around the monitoring device area;

[0053] The data fusion submodule is used to output the final ambient temperature by using the maximum fusion algorithm between the temperature data of the non-heating area in the infrared image and the ambient temperature data.

[0054] According to the intelligent temperature measurement system of the temperature measurement type overhead transmission line image and video monitoring device provided by the present invention, the infrared image analysis submodule performs the following temperature data extraction operations:

[0055] The temperature pixel values ​​of non-hot spot areas in the infrared image are called for statistical analysis. Target equipment, personnel, and heat source areas are excluded, and only the background area with lower temperature and smaller fluctuation in the image is retained for value calculation. It is expressed as the following formula:

[0056]

[0057] in, is the background area pixel set selected by image segmentation and temperature threshold filtering, and N is the total number of background pixels;

[0058] The temperature sensor acquisition submodule is used to collect real-time temperature data from temperature sensors around the monitoring device area;

[0059] The temperature sensor is used to collect the current ambient temperature value in real time, which is recorded as T sensor ;

[0060] The data fusion submodule uses the maximum fusion algorithm to fuse the two temperature data. The ambient temperature T env The expression is as follows:

[0061] T env =max(T ir-bg ,T sensor )

[0062] T sensor It is the real-time ambient temperature data of the temperature sensors around the monitoring device area.

[0063] According to the present invention, an intelligent temperature measurement system for a temperature-measuring overhead transmission line image and video monitoring device employs a temperature correction module whose multivariate nonlinear regression model is constructed using the Lasso regularization method. This model eliminates redundant influencing variables and improves model generalization by minimizing the linear combination of weighted squared error and the L1 norm. The temperature correction module uses the Lasso regularized multivariate nonlinear regression model to correct infrared temperature measurement errors.

[0064] According to an intelligent temperature measurement system of a temperature measurement type overhead transmission line image and video monitoring device provided by the present invention, the image stitching module controls the pan-tilt head to rotate to three preset angles in the horizontal direction through the main control module. The preset angles are set according to the field of view of the infrared camera to cover the complete area of ​​​​three monitoring targets. One frame of infrared image is taken at each angle, and the three frames of image are stitched in the horizontal direction to generate a complete scene image containing multiple cable terminals or lightning arresters.

[0065] Among them, the stitched image retains the temperature distribution information and pixel coordinate system of the original image, which is used for unified input calling of subsequent target detection, ranging and temperature correction modules to ensure data continuity and spatial consistency.

[0066] According to the intelligent temperature measurement system of a temperature-measuring overhead transmission line image and video monitoring device provided by the present invention, for infrared cameras of the same focal length but different models, the corresponding focal length model is evaluated using temperature measurement error samples of the model. If a stable offset exists, an average error inverse compensation method is used to add a constant offset correction value, which is expressed as the following formula:

[0067] T corr =T model +ΔT model-specific

[0068] Among them, T corr is the final corrected temperature value, ΔT model-specific This is a model correction item, which is determined by the systematic deviation of the device model under the standard test scenario.

[0069] A temperature measurement type overhead transmission line image video monitoring device, comprising:

[0070] A structural housing, used to encapsulate and protect internal electronic components and functional modules, the structural housing having waterproof, dustproof, corrosion-resistant and impact-resistant properties, and suitable for long-term outdoor operation;

[0071] The power module is used to provide power to the functional modules inside the device. The power module includes a solar power supply unit, a battery system and a power management circuit, and has the monitoring and management protection functions of voltage, current, power and temperature;

[0072] The communication module is used to realize data communication between the device and the remote management master station, supports 4G, 5G, and Wi-Fi wireless communication methods, and is used to upload image and video data, receive remote control commands, and send device status information at regular intervals;

[0073] A main control module, used for centralized control and task scheduling of the lens module, pan / tilt module, power module, communication module and alarm module in the device;

[0074] A lens module for collecting images and video information of the power transmission line and its surrounding environment, with automatic zoom, magnification, aperture adjustment, night vision imaging, image enhancement, and coding compression functions;

[0075] The pan / tilt module is used to drive the lens module to rotate horizontally and vertically, supporting multi-preset automatic cruise control, remote direction adjustment, wiper control, and fill light control to achieve all-round image acquisition;

[0076] The alarm module is used to comprehensively analyze the target temperature, ambient temperature and target detection information output by the temperature measurement system, and generate alarm information according to the set rules. The alarm rules include: global high and low temperature alarms, regional high and low temperature alarms, temperature rise alarms and three-phase temperature difference imbalance alarms.

[0077] A further solution is that the main control module specifically includes:

[0078] Edge AI computing unit, used to perform artificial intelligence processing tasks locally, including image recognition and object detection, and to move some model inference functions locally to improve image processing efficiency and system response speed;

[0079] Local storage unit, used to cache image and video data, system operation logs, and device status information, with breakpoint resume, data recovery, and historical information query functions;

[0080] The scheduling control unit is used to receive and analyze control commands issued by the remote management platform, coordinate the linkage operation of the lens module, pan / tilt module, power module, and communication module, and realize unified task scheduling and status synchronization;

[0081] The main control module realizes data interaction with the lens module, pan / tilt module, power module and communication module through the internal bus, and dynamically allocates processing resources according to task priorities.

[0082] A further solution is that the alarm module specifically includes:

[0083] Global high and low temperature alarm: When the highest temperature in the infrared image is higher than the set global high temperature threshold or lower than the global low temperature threshold, a global high and low temperature abnormal alarm is triggered;

[0084] Regional high and low temperature alarm: When the maximum temperature in the area where any monitoring target is located is higher than the set regional high temperature threshold or lower than the regional low temperature threshold, the regional high and low temperature abnormal alarm is triggered;

[0085] Temperature rise alarm: When the difference between the maximum temperature of any monitoring target area and its surrounding temperature exceeds the preset temperature rise threshold, a temperature rise abnormality alarm is triggered;

[0086] Three-phase temperature imbalance alarm: When the temperature difference between three cable terminals or three lightning arresters on the same line exceeds the set temperature imbalance threshold, the cable terminal three-phase temperature difference abnormal alarm or the lightning arrester three-phase temperature difference abnormal alarm is triggered.

[0087] A further solution is that the target detection module, intelligent ranging module, and temperature measurement and correction module all perform local inference on the edge AI computing platform to achieve low-latency response and edge intelligent analysis.

[0088] A further solution is that the device supports automated periodic temperature measurement and image stitching processes, and based on the pan-tilt preset position setting and main control module scheduling, it can achieve continuous monitoring, intelligent analysis and synchronous image uploading.

[0089] A further solution is that the device supports data exchange with the remote monitoring platform through 4G, 5G or Wi-Fi communication modules, including fused images, temperature measurement data, target detection results and equipment operation status information.

[0090] It can be seen that compared with the prior art, the present invention has the following beneficial effects:

[0091] 1. The present invention uses infrared and visible light dual-light image fusion technology to effectively enhance image details and contrast, improve target recognizability in low-light or complex environments, and combine deep learning models with rotating target frame detection methods to accurately identify tilted equipment such as cable terminals and lightning arresters.

[0092] 2. The present invention proposes an ambient temperature acquisition method based on the fusion of infrared images and temperature sensors, which comprehensively considers the impact of environmental changes during the temperature measurement process, provides complete context information for subsequent correction algorithms, and significantly improves the stability and accuracy of temperature measurement data.

[0093] 3. The present invention combines the pan-tilt angle control and the camera field of view parameters to achieve continuous acquisition and stitching of multiple preset position images, effectively breaking through the single-frame viewing angle limitation of the infrared camera, meeting the unified imaging requirements of multiple cable terminals or lightning arresters, and improving the integrity and continuity of the temperature distribution analysis of the entire image.

[0094] 4. The present invention integrates multiple alarm rules, including global high and low temperatures, regional abnormal temperatures, abnormal temperature rise and three-phase imbalance, etc. It can combine target identification information and environmental parameters for comprehensive judgment, improve the accuracy and pertinence of alarms, and effectively serve the early warning and operation and maintenance response of transmission line equipment status.

[0095] 5. The core modules of the present invention can all be deployed and executed locally on the edge AI computing unit to achieve low-latency, high-efficiency intelligent reasoning and data analysis; at the same time, the wireless communication module supports data exchange and control command reception on the remote platform, which is suitable for distributed, multi-site, and remote power operation and maintenance scenarios.

[0096] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0097] Figure 1 This is a schematic diagram of an embodiment of an intelligent temperature measurement system of a temperature measurement type overhead transmission line image and video monitoring device of the present invention;

[0098] Figure 2This is a schematic diagram of image comparison before and after the fusion of infrared images and visible light images in an embodiment of an intelligent temperature measurement system of a temperature measurement type overhead transmission line image video monitoring device of the present invention;

[0099] Figure 3 It is a schematic diagram of a target detection module recognition result image and a rotating target frame in an embodiment of an intelligent temperature measurement system of a temperature measurement type overhead transmission line image video monitoring device of the present invention;

[0100] Figure 4 This is a schematic diagram of the results of non-contact ranging using infrared images in an embodiment of an intelligent temperature measurement system of a temperature measurement type overhead transmission line image and video monitoring device of the present invention;

[0101] Figure 5 This is a schematic diagram of the results of obtaining environmental parameters through infrared images in an embodiment of an intelligent temperature measurement system of a temperature measurement type overhead transmission line image and video monitoring device of the present invention;

[0102] Figure 6 This is a schematic diagram of temperature measurement error of a Lasso regularized multivariate nonlinear regression model in an embodiment of an intelligent temperature measurement system of a temperature measurement type overhead transmission line image and video monitoring device of the present invention;

[0103] Figure 7 This is a schematic diagram of image stitching results from multiple angles of a pan / tilt system in an embodiment of an intelligent temperature measurement system of a temperature measurement type overhead transmission line image and video monitoring device of the present invention;

[0104] Figure 8 The present invention is a block diagram of the principle of an intelligent temperature measurement system embodiment of a temperature measurement type overhead transmission line image and video monitoring device. DETAILED DESCRIPTION

[0105] To make the objectives, technical solutions, and advantages of the present invention more clear, the technical solutions of the present invention will be clearly and completely described below in conjunction with the accompanying drawings. Obviously, the embodiments described are only some of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts shall fall within the scope of protection of the present invention.

[0106] References herein to "embodiments" mean that a particular feature, structure, or characteristic described in connection with the embodiments may be included in at least one embodiment of the present application. The appearance of this phrase in various places in the specification does not necessarily refer to the same embodiment, nor does it constitute an independent or alternative embodiment that is mutually exclusive of other embodiments. It is understood, both explicitly and implicitly, by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0107] See also Figure 1 This embodiment provides an intelligent temperature measurement system for a temperature measurement type overhead transmission line image and video monitoring device, comprising:

[0108] The dual-light image fusion module is used to collect infrared images and visible light images, align and fuse them, and generate a fused image;

[0109] The target detection module is used to perform image processing on the fused image, identify the target object in the image using a target detection algorithm, and output the position information and category information of the target object in the image;

[0110] The intelligent distance measurement module obtains the pixel size of the target in the image based on the target position information output by the target detection module, and calculates the actual distance between the device and the target in combination with the preset actual physical size of the target;

[0111] An environmental parameter acquisition module is used to obtain the ambient temperature data of the area where the cable terminal or lightning arrester is located, wherein the ambient temperature data is obtained by fusing the temperature measurement data with the temperature data of the non-hot spot area in the infrared image;

[0112] The temperature measurement correction module is used to take the original temperature measurement value, the actual distance of the target and the ambient temperature data as input, use the multivariate nonlinear regression model based on Lasso regularization to correct the infrared temperature measurement error, and output the corrected temperature value;

[0113] The image stitching module is used to control the pan-tilt head to rotate horizontally when the single-frame field of view of the infrared camera cannot cover all monitoring targets. The rotation angle range matches the field of view of the infrared camera, and infrared images covering the left, center and right directions are obtained in sequence. The three acquired infrared images are pre-processed and then stitched horizontally to generate a complete infrared image covering multiple cable terminals or lightning arresters.

[0114] The process of bi-optical image fusion specifically includes the following steps:

[0115] Using OpenCV image processing technology to preprocess, extract features, and fuse the bi-optical images to organize the dataset into a format suitable for object detection model training;

[0116] Use the PyTorch deep learning framework to build an object detection model, and use the functions and tools provided by the framework to implement model initialization, training, parameter tuning, and performance evaluation;

[0117] Conduct programming experiments using Python and Java. Write Python code to implement dual-light image fusion, target detection model building, and temperature measurement and correction model building. Write Java code to implement intelligent ranging, target detection model application, environmental parameter acquisition, temperature measurement and correction model application, and image stitching. Execute the code using the Python and Java runtime environments to complete the entire intelligent temperature measurement system construction process.

[0118] In practical applications, the process includes four modules: image enhancement, image alignment, fusion processing, and color correction. The bi-optical image fusion module specifically includes:

[0119] The image enhancement submodule is used to enhance the brightness, contrast and noise of visible light images and infrared images respectively;

[0120] Image alignment submodule, used to spatially align visible light images with infrared images based on image feature points or geometric correction models;

[0121] The fusion processing submodule uses weighted average or maximum strategy to perform pixel-level fusion of the two images;

[0122] The color correction submodule is used to adjust the color temperature and balance the brightness of the fused image to ensure image quality consistency.

[0123] For visible light images, the image enhancement submodule performs the following enhancement operations:

[0124] First, the image is grayscaled, which is expressed as the following formula:

[0125] I(x,y)=0.30R(x,y)+0.59G(x,y)+0.11B(x,y)

[0126] Where I(x,y) is the pixel value at the yth row and xth column of the grayscale image, R(x,y) is the pixel value at the yth row and xth column of the blue channel in the RGB image, G(x,y) is the pixel value at the yth row and xth column of the green channel in the RGB image, and B(x,y) is the pixel value at the yth row and xth column of the red channel in the RGB image.

[0127] Then, the image contrast is enhanced using a linear contrast stretching function, which is expressed as the following formula:

[0128] I enh (x,y)=α·I(x,y)+β

[0129] Among them, I enh (x, y) is the pixel value after image contrast enhancement, α is the contrast enhancement coefficient, and β is the brightness shift coefficient.

[0130] Next, a median filter is used to suppress image noise, which is expressed as the following formula:

[0131]

[0132] in, is a sliding window centered at (x,y).

[0133] For infrared image enhancement, the image enhancement submodule performs the following enhancement operations:

[0134] The same method (linear contrast stretching function) is used to perform linear contrast stretching and adaptive normalization is performed according to the dynamic range of the image histogram to highlight the edge of the heat distribution;

[0135] Wherein, the adaptive normalization includes: assuming the infrared image is I ir (x,y), the minimum pixel value is I min , the maximum value is I max , then the enhanced image is normalized according to the following formula:

[0136]

[0137] This method can effectively improve image contrast, highlight the temperature change boundary area in infrared images, and enhance the separation between thermal anomaly areas and background.

[0138] In this embodiment, the image alignment submodule is used to perform spatial registration on the infrared image and the visible light image by using a checkerboard template image in the system calibration phase to obtain the extrinsic parameter matrices and projection relationships of the two types of sensors.

[0139] The infrared image is aligned with the visible light image through the affine transformation matrix H, which is as follows:

[0140]

[0141] Among them, (x ir ,y ir ) is the pixel value at row y and column x in the infrared image, and (x,y) is the corresponding pixel value in the visible light image. The matrix H is fixed after initial calibration.

[0142] After completing the spatial registration, the fusion processing submodule uses a weighted average strategy to fuse the two images at the pixel level, which can be expressed as the following formula:

[0143] I fused (x,y)=w1·I(x,y)+w2·I ir (x,y)

[0144] Among them, I fused(x, y) is the pixel value in the fused image, w1+w2=1.

[0145] This fusion strategy ensures that the texture details of the visible light image are preserved as much as possible while retaining the temperature information.

[0146] After the fused image is generated, the color correction submodule is used to perform brightness balance and color temperature adjustment:

[0147] Histogram equalization is used to enhance the consistency of brightness distribution across the entire image, which can be expressed as the following formula:

[0148]

[0149] Where h(i) is the number of pixels at the i-th grayscale level in the grayscale histogram of the image pixels, and L is the total number of grayscale levels.

[0150] like Figure 2 As shown, Figure 2 (a) and Figure 2 (b) are visible light image and infrared image respectively, and the output fusion image is as follows Figure 2 As shown in (c), the temperature distribution level of the infrared image is retained, and the structural information and texture details of the visible light image are also included, which is suitable for subsequent target detection and temperature analysis.

[0151] In this embodiment, the target detection module is used to automatically identify targets on the fused image, identify key power equipment such as cable terminals and lightning arresters in the image, and output their categories, location information, and rotation angles.

[0152] Specifically, the target detection module takes the fused image as input and uses YOLOv11 as the deep learning detection network to build a target detection model. The target detection model is equipped with an S-Ghost module in the backbone network to replace the first two general convolutional layers. Feature extraction is performed on the backbone network through the S-Ghost module, and an RSK module with an attention mechanism is set at the starting position of the neck network to realize the recognition of inclined targets such as cable terminals or lightning arresters. When performing image annotation, the rolabelimg annotation tool is used to annotate the cable terminals in the image with rotated target boxes, and the annotated images are data enhanced to establish a sample dataset.

[0153] The output of the target detection model includes: target category label, two-dimensional coordinates of the target in the fused image, rotation angle information and recognition confidence score.

[0154] The target detection method used in this embodiment refers to the technical solution proposed in the authorized invention patent "A Cable Terminal Recognition Method and Device Based on Infrared Images and Improved YOLOv5" with authorization publication number CN117409083B. This patent, taking advantage of the low texture and low contrast characteristics of infrared images, effectively improves the recognition accuracy and robustness of rotating targets (such as cable terminals) in power equipment by constructing the feature enhancement module S-Ghost and the attention mechanism module RSK.

[0155] In this embodiment, the following expansions and adaptations are performed based on the above patented method:

[0156] 1) Input image type optimization

[0157] The original patent uses infrared images as detection input. This embodiment replaces the input image with a dual-light fusion image, that is, the infrared image and the visible light image after pixel-level registration and fusion processing are used as the input channels of the detection model, thereby improving the target contour extraction capability and enhancing the structural feature perception of the model.

[0158] 2) Target detection model structure upgrade

[0159] While maintaining the same structure as the feature enhancement modules S-Ghost and RSK, the overall object detection framework has been upgraded from YOLOv5 to YOLOv11. YOLOv11 boasts a deeper network structure, improved training strategies, and higher inference efficiency, further improving detection accuracy and real-time performance.

[0160] 3) Rotate target frame output

[0161] The detection model still uses the Rotated Bounding Box (RBox) format for output, such as Figure 3 As shown, Figure 3 (a) is the cable terminal target detection result after dual-light fusion. Figure 3 (b) is the result of infrared light cable terminal target detection. Figure 3 (c) is the visible light cable terminal target detection result image, which is suitable for the cable terminal and lightning arrester targets photographed at an angle in the image. The output results include target category, two-dimensional coordinates, rotation angle and confidence.

[0162] Through the above improvements, this embodiment introduces dual-light fusion images and a high-performance model architecture, which significantly improves the adaptability and accuracy of the target detection module in complex environments.

[0163] In this embodiment, the intelligent ranging module adopts a distance measurement method based on the proportional relationship between the actual physical size of the target object and the pixel size in the image, and combines the focal length and pixel size of the camera to establish a functional relationship between the pixel size and the actual distance, which is used to realize non-contact target distance calculation under different shooting conditions.

[0164] Specifically, the intelligent ranging module is used to calculate the distance between the target and the device based on the target information output by the target detection module, combined with the pixel size in the image and the actual physical size, through the principle of monocular vision to achieve non-contact intelligent ranging, such as Figure 4 shown.

[0165] The ranging process of the intelligent ranging module specifically includes the following two aspects:

[0166] 1) Distance estimation based on image geometric features

[0167] The monocular vision similar triangle principle is used to calculate the target distance based on the pixel width and pixel height of the target. The calculation formula is as follows:

[0168] Distance calculation based on width:

[0169]

[0170] Distance calculation based on height:

[0171]

[0172] Among them, D w and D h Represents the ranging results based on width and height respectively, f is the focal length, s is the pixel size, W r and H r are the actual physical width and height of the target, W p and H p are the pixel width and height of the object in the image, respectively.

[0173] The final distance D is obtained by taking the arithmetic average of the results of the above two calculation methods:

[0174]

[0175] 2) Dynamic ranging update mechanism

[0176] In order to improve the stability and robustness of the system in continuous detection scenarios, the ranging module is provided with a dynamic update mechanism.

[0177] In the first detected image, a confidence check is performed on all identified objects, and only those with a confidence score above a set threshold are retained as valid distance measurement objects. Their pixel locations and category labels are cached in system memory along with the initial distance measurement results for reference in subsequent images.

[0178] If the target detection module fails to identify any target in a certain detection cycle, or the confidence of the identified targets is lower than the set threshold, the system calls the cached target information of the last detected image and uses it as the ranging output of the current image to achieve fault-tolerant filling of the data.

[0179] When detecting a new image, if a new target is identified, the system will match it with the cached target in the previous frame. The intersection over union (IoU) is used as the matching strategy, and the specific calculation method is as follows:

[0180]

[0181] Among them, A inter is the intersection area of ​​the current detection box and the cached target box, A union =The union area of ​​the two. If IoU ≥ θ (threshold, such as 0.5) and the category information is consistent, they are considered the same target, and the system updates its pixel position and ranging value. Otherwise, it is determined to be a new target, reassigned a number, and enters the ranging process. All ranging targets are assigned a unique number. The system carries the target number, category label, and current estimated distance in the ranging result output, which is used for the linkage call of the target detection and temperature correction modules.

[0182] In this embodiment, the environmental parameter acquisition module includes:

[0183] Infrared image analysis submodule, used to extract temperature data of non-heating areas in infrared images;

[0184] The temperature sensor acquisition submodule is used to collect real-time ambient temperature data from temperature sensors around the monitoring device area;

[0185] The data fusion submodule is used to output the final ambient temperature by using the maximum fusion algorithm between the temperature data of the non-heating area in the infrared image and the ambient temperature data.

[0186] Specifically, the infrared image analysis submodule performs the following temperature data extraction operations:

[0187] The temperature pixel values ​​of the non-heat point area (background area) in the infrared image are called for statistical analysis, excluding the target equipment, personnel, and heat source areas, and only the background area with lower temperature and smaller fluctuation in the image is retained for value calculation, which is expressed as the following formula:

[0188]

[0189] Among them, B is the background area pixel set selected by image segmentation and temperature threshold filtering, and N is the total number of background pixels.

[0190] This method can effectively capture the temperature of areas in the image scene that are not affected by thermal interference and serve as an objective reference.

[0191] The temperature sensor acquisition submodule is used to collect real-time temperature data from temperature sensors around the monitoring device area;

[0192] The device is equipped with a temperature sensor to collect the current ambient temperature in real time, which is recorded as T sensor ;

[0193] This sensor can reflect temperature changes in the device's microenvironment, offering fast response and strong stability. However, because the temperature sensor may be affected by the device's own heat dissipation, its measured values ​​may be higher during high-load operation, necessitating fusion with image temperature information.

[0194] The data fusion submodule is used to output the final ambient temperature using the maximum fusion algorithm among the above two temperature values ​​for subsequent temperature measurement correction.

[0195] In order to avoid the sensor being interfered with by the heat of the equipment and causing it to be too high, and to avoid the image processing being too low due to segmentation errors, this embodiment uses the maximum fusion algorithm to fuse the two temperature data. The ambient temperature T env The expression is as follows:

[0196] T env =max(T ir-bg ,T sensor )

[0197] T sensor It is the real-time ambient temperature data of the temperature sensors around the monitoring device area.

[0198] This fusion strategy has the following advantages:

[0199] When the sensor measurement is high due to the heat dissipation of the device, the image temperature is used as compensation;

[0200] When image analysis results in a low background due to occlusion or misjudgment, sensor data is used as compensation;

[0201] The maximum value fusion method has a greater safety margin in infrared temperature measurement applications and is suitable for abnormal monitoring scenarios of power equipment.

[0202] Through the above-mentioned environmental parameter acquisition method, the system realizes the real-time, stable and spatially fitting acquisition of temperature external variables. The environmental data provided by this module will play a key role in the temperature measurement error compensation model, and will help to build an intelligent temperature measurement system with high environmental adaptability, such as Figure 5 shown.

[0203] In this embodiment, the multivariate nonlinear regression model of the temperature measurement correction module is constructed based on the Lasso regularization method. By minimizing the linear combination of the weighted square error and the L1 norm, redundant influencing variables are eliminated and the generalization ability of the model is improved. The temperature measurement correction module calls the multivariate nonlinear regression model based on Lasso regularization to correct the infrared temperature measurement error.

[0204] Specifically, the temperature measurement correction module is used to correct the error of the original temperature value output by the infrared detector. The original infrared detector temperature value, target distance, and ambient temperature are used as input factors to comprehensively correct the temperature measurement deviation caused by the device model, shooting distance, and environmental fluctuations. The corrected temperature measurement error is as follows: Figure 6 shown.

[0205] The temperature measurement and correction process of the temperature measurement and correction module specifically includes the following three aspects:

[0206] 1) Applying the multi-focal length error correction model

[0207] For infrared cameras with different focal lengths (such as 9.1mm and 13mm), independent temperature correction models are established, using a polynomial regression based on Lasso to express the nonlinear relationship between temperature measurement error and multiple factors. The correction output is as follows:

[0208] For an infrared camera with a focal length of 9.1 mm, the error comparison before and after correction is as follows: Figure 2 As shown, the basic correction model is:

[0209]

[0210] Among them, T CW91001 model It represents the temperature value of the model with focal length of 9.1mm and model 001, T CW91001de v represents the temperature value measured by the original infrared detector with a focal length of 9.1 mm and model 001.

[0211] For an infrared camera with a focal length of 13mm, the error comparison before and after correction is as follows: Figure 3 As shown, the basic correction model is:

[0212]

[0213] Among them, T CW13001 modelIt represents the temperature value of the model with focal length of 13mm and model 001, T CW13001 dev Indicates the temperature value measured by the original infrared detector with a focal length of 13mm and model 001.

[0214] 2) Compensation strategy for cameras of different models with the same focal length

[0215] For infrared cameras with the same focal length but different models, the system uses the temperature measurement error samples of the model to evaluate the corresponding focal length model. If there is a stable offset, the system uses the "average error reverse compensation" method to add a constant offset correction value:

[0216] T corr =T model +ΔT model-specific

[0217] Among them, T corr is the final corrected temperature value, ΔT model-specific This is a model correction item, which is determined by the systematic deviation of the device model under the standard test scenario.

[0218] 3) Model application and edge deployment

[0219] The above-mentioned temperature measurement error correction model is deployed in the edge AI computing platform, and the inference process is automatically executed after each image acquisition and temperature measurement analysis.

[0220] Input variables include: original infrared detector temperature value, current target distance, and current ambient temperature.

[0221] The model output is: the corrected target temperature, with the error controlled within the range of ±2°C; it is bound to the target detection ID for subsequent alarm module calls.

[0222] In this embodiment, the image stitching module controls the pan-tilt head to rotate to three preset angles in the horizontal direction through the main control module. The preset angles are set according to the field of view of the infrared camera to cover the complete area of ​​the three monitoring targets. A frame of infrared image is taken at each angle, and the three frames of image are stitched in the horizontal direction to generate a complete scene image containing multiple cable terminals or lightning arresters.

[0223] Among them, the stitched image retains the temperature distribution information and pixel coordinate system of the original image, which is used for unified input calling of subsequent target detection, ranging and temperature correction modules to ensure data continuity and spatial consistency.

[0224] Specifically, during the image stitching process, the field of view (FOV) of an infrared camera with a long focal length is relatively small, and a single frame image may not be able to simultaneously cover all temperature measurement targets (such as the left, center, and right three-phase cable terminals). Therefore, it is necessary to introduce a pan-tilt control mechanism and an image stitching algorithm to achieve scene expansion. By controlling the pan-tilt to shoot at fixed angles and horizontally stitching multiple images, a complete infrared image covering multiple targets is generated. The specific steps include:

[0225] 1) PTZ viewing angle distribution strategy

[0226] Due to the large focal length of the infrared camera, the field of view of a single frame image is narrow and cannot cover multiple cable terminals or lightning arrester targets at the same time. Therefore, the PTZ is set to rotate in sections according to a fixed field of view. Three PTZ preset positions are set in the main control module, corresponding to the left, center, and right viewing angles, forming horizontal section coverage:

[0227] The camera's horizontal field of view angle is recorded as θ;

[0228] The three preset angles of the PTZ are set as follows:

[0229] Middle viewing angle (preset position 1): θ0 = 0°;

[0230] Left viewing angle (preset position 2): θ0 = -θ;

[0231] Right viewing angle (preset position 3): θ0 = +θ;

[0232] The rotation angle θ is the horizontal field of view of the infrared camera.

[0233] In this way, each time the gimbal rotates by angle θ, it just spans the width of one frame of field of view, so that there is no overlapping area between images, achieving end-to-end docking.

[0234] 2) Image acquisition and naming cache

[0235] The main control module controls the pan / tilt to rotate to the center, left, and right preset positions in sequence, and triggers infrared image acquisition to obtain:

[0236] I C (x,y): intermediate image;

[0237] I L (x,y): left image;

[0238] I R (x,y): right image;

[0239] The three images are cached in the order they were taken, and the shooting time, gimbal angle, and image number are recorded.

[0240] 3) Image stitching algorithm

[0241] The stitching process uses direct stitching, which does not require alignment and registration of overlapping image areas. The stitching method is as follows:

[0242] I full (x,y)=I L (x,y)⊕I C (x,y)⊕I R (x,y)

[0243] Among them, ⊕ represents the non-overlapping splicing operation of the image in the horizontal direction.

[0244] The horizontal width of the stitched image is three times the width of a single-frame image, and the coverage angle is 3θ, which can fully display multiple adjacent targets. It is suitable for scenarios such as three-phase cable terminals and lightning arresters arranged side by side.

[0245] 4) Image stitching data structure preservation

[0246] After the image stitching is completed, in order to ensure the unified processing of the stitched image by modules such as target detection, intelligent ranging and temperature correction, the present invention reconstructs and uniformly maps the data structure and regional coordinate information of the stitched image.

[0247] In this embodiment, an infrared camera is used to capture images in the center, left, and right directions, respectively, and the resolution of each image is W×H. During stitching, the three images are stitched together horizontally without overlap, and the stitched image size is 3W×H. The pixel coordinate system is continuously expanded along the x-axis. During the generation of the stitched image, the lateral offsets of the three images are recorded separately according to the shooting order of the original images and the preset position settings of the pan / tilt head. The corresponding position intervals of the image areas of the left, center, and right images in the stitched image are [0, W), [W, 2W), and [2W, 3W], respectively. The coordinates of all identified targets in the stitched image are mapped and converted by the lateral offset. The specific conversion relationship is:

[0248] x concat =x raw +δ,y concat =y raw

[0249] Among them, δ=0 (left picture), δ=W (middle picture), and δ=2W (right picture) are used to relocate the position of each target area in the stitched image.

[0250] In order to ensure the integrity of the image visual elements and the non-redundancy of information, the present invention adopts a differentiated OSD drawing strategy. In a single-frame image, the following information is drawn respectively: the left image (preset position 2) contains the time information and custom text in the upper left corner, the channel number information in the lower left corner, and the number of each target area and the highest temperature value of the area; the middle image (preset position 1) does not draw any OSD elements; the right image (preset position 3) draws the color palette and the temperature range information of the current thermal image. After the spliced ​​image is generated, the contents of the three frames are unified and fused, and all the temperature measurement area information (including numbering), the highest temperature mark of the temperature measurement area, and the global highest temperature mark are centrally drawn on the spliced ​​image. At the same time, the time information, custom text, channel information, area information of the left image and the palette content of the right image are retained to form an information image with a unified coordinate system, a complete OSD structure and a standard temperature measurement display, such as Figure 7 shown.

[0251] Each temperature measurement target in the mosaic is associated with its region number, identification category, temperature value, region boundary, and identification confidence level, all stored in the mosaic's metadata table. The mosaic serves as the input for subsequent recognition analysis, distance calculation, and temperature correction, ensuring the algorithm call interface remains consistent with the single-frame image.

[0252] In addition, the global hot spot information in the mosaic map is selected by comparing the maximum temperature values ​​in all temperature measurement areas. Its coordinates are accurately drawn in the mosaic map after offset correction, which is used to trigger functional modules such as the global high temperature alarm.

[0253] Through the above design, the present invention takes into account the scalability and information integrity of the image structure when performing image stitching operations, so that the stitched image can be used as a unified image source equivalent to a single-frame input, meeting the data continuity and spatial consistency requirements in the scenario of simultaneous temperature measurement of multiple targets of power equipment.

[0254] 5)Automation triggering and control logic

[0255] The system supports scheduled image stitching from the master control module, and can also be triggered in real time by a remote platform. After local processing, the stitched image is uploaded to the backend platform and synchronized with the recognition results and temperature measurement data, enabling batch visual analysis and alarm judgment.

[0256] Through the above method, the image stitching module proposed in the present invention achieves efficient infrared image coverage of multiple target areas by relying on precise control of the pan-tilt perspective without the need for image overlapping areas, and is suitable for high-precision temperature measurement applications of power equipment arranged linearly in space.

[0257] This embodiment also provides a temperature measurement type overhead transmission line image video monitoring device, the structural block diagram of which is as follows: Figure 8As shown, the process of developing a temperature measurement type overhead transmission line image video monitoring device specifically includes:

[0258] Combining Java, C, and C++ programming languages, we developed the integrated hardware and software design for a temperature-measuring overhead transmission line image and video monitoring device. By executing code within the Java, C, and C++ runtime environments, we completed the entire development process of the temperature-measuring overhead transmission line image and video monitoring device, deploying and debugging core functional modules such as the device's control logic, image acquisition and processing, task scheduling, edge reasoning, and data communication, completing the system build.

[0259] The device includes:

[0260] The structural housing is used to encapsulate and protect the internal electronic components and functional modules. The structural housing is waterproof, dustproof, corrosion-resistant, and impact-resistant, making it suitable for long-term outdoor operation. The structural housing also integrates solar panel mounting positions, wiring ports, and standard mounting bracket interfaces for secure deployment of the equipment on a tower or substation.

[0261] The power module is used to provide power to the various functional modules within the device. The power module includes a solar power supply unit, a battery system, and a power management circuit. It has the function of real-time detection and control of voltage, current, power consumption, power consumption, and device temperature, and supports efficient charge and discharge management and overload or undervoltage protection strategies;

[0262] The communication module is used to realize data communication between the device and the remote management master station. It supports multiple network transmission methods such as 4G, 5G, and Wi-Fi. It is used to upload image and video data, receive remote control commands, and send device status information at regular intervals. At the same time, it receives control commands, parameter configurations, and remote maintenance commands issued by the platform to ensure that the data link between the device and the platform is unobstructed in real time;

[0263] The main control module, as the core control center of the device, is responsible for the coordinated scheduling, data aggregation and intelligent decision-making of various functional modules. For example, it is used to centrally control and schedule tasks for the lens module, pan / tilt module, power module, communication module and alarm module in the device;

[0264] The lens module is used to collect image and video information of the transmission line itself, tower structure, insulators, conductors and the surrounding operating environment. It has autofocus, automatic aperture control, night vision enhancement, high dynamic range imaging, coding compression and wide temperature operation capabilities;

[0265] The pan / tilt module is used to drive the lens module's horizontal and vertical rotation. It supports multiple preset position settings, multi-angle timed inspections, and remote zoom adjustment. It can quickly switch shooting angles according to task configurations, enabling multi-point cruise image acquisition and continuous tracking shooting.

[0266] The alarm module is used to comprehensively analyze the target temperature, ambient temperature and target detection information output by the temperature measurement system, and generate alarm information according to the set rules. The alarm rules include: global high and low temperature alarms, regional high and low temperature alarms, temperature rise alarms and three-phase temperature difference imbalance alarms.

[0267] In this embodiment, the main control module specifically includes:

[0268] Edge AI computing units are used to perform artificial intelligence processing tasks locally, including image recognition, target detection, temperature correction, and alarm judgment. By deploying models forward, they enable local reasoning, reduce cloud dependency, and improve response efficiency.

[0269] Local storage unit, used to cache image and video data, system operation logs, and device status information, with breakpoint resume, data recovery, and historical information query functions;

[0270] The scheduling control unit is used to receive and analyze control commands issued by the remote management platform, coordinate the linkage operation of the lens module, pan / tilt module, power module, and communication module, and realize multi-threaded task scheduling, priority management, and synchronous status update;

[0271] The main control module exchanges data with the lens module, gimbal module, power module and communication module through an internal high-speed bus, and can dynamically allocate resources according to task requirements to achieve device-level edge intelligent operation.

[0272] In this embodiment, the alarm module specifically includes:

[0273] Global high and low temperature alarm: When the highest temperature in the infrared image is higher than the set global high temperature threshold or lower than the global low temperature threshold, a global high and low temperature abnormal alarm is triggered;

[0274] Regional high and low temperature alarm: When the maximum temperature in the area where any monitoring target is located is higher than the set regional high temperature threshold or lower than the regional low temperature threshold, the regional high and low temperature abnormal alarm is triggered;

[0275] Temperature rise alarm: When the difference between the maximum temperature of any monitoring target area and its surrounding temperature exceeds the preset temperature rise threshold, a temperature rise abnormality alarm is triggered;

[0276] Three-phase temperature imbalance alarm: When the temperature difference between three cable terminals or three lightning arresters on the same line exceeds the set temperature imbalance threshold, the cable terminal three-phase temperature difference abnormal alarm or the lightning arrester three-phase temperature difference abnormal alarm is triggered.

[0277] The alarm module can upload the alarm information to the main station through the communication module, including the number, location, time, temperature value, alarm type and image information of the alarm target, for remote analysis and early warning linkage response.

[0278] Through the collaborative work of the above modules, the overhead transmission line image and video monitoring device (temperature measurement type) described in the present invention can realize automated monitoring of the entire process of image acquisition, temperature perception, target analysis and intelligent alarm of power equipment. It has the technical advantages of flexible deployment, stable operation and high degree of intelligence, and is suitable for long-term non-contact thermal imaging monitoring scenarios of key parts such as cable terminals, lightning arresters, and wire joints.

[0279] In summary, the present invention realizes closed-loop processing of the entire process from image acquisition, target detection, non-contact ranging, intelligent temperature measurement to automatic alarm by constructing a multi-module integrated overhead transmission line image and video monitoring device (temperature measurement type), significantly improving the intelligence, automation and real-time level of transmission line operation and maintenance. In traditional power equipment temperature measurement systems, there are common problems such as a single temperature measurement method, weak environmental adaptability, delayed data update, and inaccurate alarm response. The present invention solves the technical problems of insufficient temperature measurement accuracy and system coordination ability in complex scenarios by introducing dual-light image fusion, deep learning target detection and intelligent alarm strategy.

[0280] Furthermore, the present invention adopts a target detection algorithm based on dual-light fusion images, which effectively improves the recognition accuracy under harsh conditions such as low illumination and hot spot interference; combined with image geometric ranging and environmental perception modules, it realizes temperature correction and error compensation for different devices and different distances, and the temperature measurement error can be stably controlled within ±2°C; combined with pan-tilt control and image stitching strategies, it ensures the camera's complete coverage of multiple temperature measurement targets under conditions of limited field of view, and adapts to detection needs in complex wiring and irregular installation scenarios.

[0281] Furthermore, the present invention integrates an edge AI computing platform, a solar power supply system, a remote communication module and a master control scheduling mechanism in the system architecture. It has powerful edge reasoning capabilities and local task scheduling capabilities, can independently complete temperature measurement analysis and preliminary alarm judgment in an offline state, and supports uploading images and alarm information to the remote monitoring platform, realizing the distributed architecture advantage of "front-end intelligence + back-end supervision".

[0282] Furthermore, the alarm module proposed in the present invention integrates multiple types of alarm strategies such as global high and low temperatures, regional high and low temperatures, temperature rise and three-phase imbalance, forming a multi-target, multi-scenario, and multi-dimensional alarm judgment system. It can quickly respond to thermal anomalies in key parts such as cable terminals and lightning arresters, assist the power system in early intervention and precise maintenance, and ensure the stability and safety of transmission line operation.

[0283] The technical features of the above embodiments can be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0284] The above embodiments are only preferred embodiments of the present invention and cannot be used to limit the scope of protection of the present invention. Any non-substantial changes and replacements made by technicians in this field on the basis of the present invention fall within the scope of protection required by the present invention.

Claims

1. An intelligent temperature measurement system for a temperature measurement type overhead transmission line image and video monitoring device, characterized in that: include: The dual-light image fusion module is used to collect infrared images and visible light images, align and fuse them, and generate a fused image; The target detection module is used to perform image processing on the fused image, identify the target object in the image using a target detection algorithm, and output the position information and category information of the target object in the image; The intelligent distance measurement module obtains the pixel size of the target in the image based on the target position information output by the target detection module, and calculates the actual distance between the device and the target in combination with the preset actual physical size of the target; An environmental parameter acquisition module is used to obtain the ambient temperature data of the area where the cable terminal or lightning arrester is located, wherein the ambient temperature data is obtained by fusing the temperature measurement data with the temperature data of the non-hot spot area in the infrared image; The temperature measurement correction module is used to take the original temperature measurement value, the actual distance of the target and the ambient temperature data as input, use the multivariate nonlinear regression model based on Lasso regularization to correct the infrared temperature measurement error, and output the corrected temperature value; The image stitching module is used to control the pan-tilt head to rotate horizontally when the single-frame field of view of the infrared camera cannot cover all monitoring targets. The rotation angle range matches the field of view of the infrared camera, and infrared images covering the left, center and right directions are obtained in sequence. The three acquired infrared images are pre-processed and then stitched horizontally to generate a complete infrared image covering multiple cable terminals or lightning arresters.

2. The system according to claim 1, wherein: The dual-light image fusion module includes: The image enhancement submodule is used to enhance the brightness, contrast and noise of visible light images and infrared images respectively; Image alignment submodule, used to spatially align visible light images with infrared images based on image feature points or geometric correction models; The fusion processing submodule uses weighted average or maximum strategy to perform pixel-level fusion of the two images; The color correction submodule is used to adjust the color temperature and balance the brightness of the fused image to ensure image quality consistency.

3. The system according to claim 2, characterized in that: For visible light images, the image enhancement submodule performs the following enhancement operations: First, the image is grayscaled, which is expressed as the following formula: I(x,y)=0.30R(x,y)+0.59G(x,y)+0.11B(x,y) Where I(x,y) is the pixel value at the yth row and xth column of the grayscale image, R(x,y) is the pixel value at the yth row and xth column of the blue channel in the RGB image, G(x,y) is the pixel value at the yth row and xth column of the green channel in the RGB image, and B(x,y) is the pixel value at the yth row and xth column of the red channel in the RGB image. Then, the image contrast is enhanced using a linear contrast stretching function, which is expressed as the following formula: I enh (x,y)=α·I(x,y)+β Among them, I enh (x, y) is the pixel value after image contrast enhancement, α is the contrast enhancement coefficient, and β is the brightness shift coefficient. Next, a median filter is used to suppress image noise, which is expressed as the following formula: in, is a sliding window centered at (x,y).

4. The system according to claim 3, wherein: For infrared image enhancement, the image enhancement submodule performs the following enhancement operations: A linear contrast stretching function is used to perform linear contrast stretching, and adaptive normalization is performed according to the dynamic range of the image histogram to highlight the edge of the thermal distribution; Wherein, the adaptive normalization includes: assuming the infrared image is I ir (x,y), the minimum pixel value is I min , the maximum value is I max , then the enhanced image is normalized according to the following formula:

5. The system according to claim 4, characterized in that: After completing the spatial registration, the fusion processing submodule uses a weighted average strategy to fuse the two images at the pixel level, which can be expressed as the following formula: I fused (x,y)=w1·I(x,y)+w2·I ir (x,y) Among them, I fused (x, y) is the pixel value in the fused image, w1+w2=1.

6. The system according to claim 5, characterized in that: After the fused image is generated, the color correction submodule is used to perform brightness balance and color temperature adjustment: Histogram equalization is used to enhance the consistency of brightness distribution across the entire image, which can be expressed as the following formula: Where h(i) is the number of pixels at the i-th grayscale level in the grayscale histogram of the image pixels, and L is the total number of grayscale levels.

7. The system according to any one of claims 1 to 6, characterized in that: The target detection module takes the fused image as input and uses YOLOv11 as a deep learning detection network to build a target detection model. The target detection model is equipped with an S-Ghost module in the backbone network to replace the first two general convolutional layers. Feature extraction is performed on the backbone network through the S-Ghost module. An RSK module with an attention mechanism is provided at the starting position of the neck network to realize the recognition of inclined targets such as cable terminals or lightning arresters. When performing image annotation, the rolabelimg annotation tool is used to annotate the cable terminals in the image with a rotated target box, and the annotated image is data enhanced to establish a sample data set. The output of the target detection model includes: target category label, two-dimensional coordinates of the target in the fused image, rotation angle information and recognition confidence score.

8. The system according to any one of claims 1 to 6, characterized in that The environmental parameter acquisition module includes: Infrared image analysis submodule, used to extract temperature data of non-heating areas in infrared images; The temperature sensor acquisition submodule is used to collect real-time ambient temperature data from temperature sensors around the monitoring device area; The data fusion submodule is used to output the final ambient temperature by using the maximum fusion algorithm between the temperature data of the non-heating area in the infrared image and the ambient temperature data.

9. The system according to claim 8, characterized in that: The infrared image analysis submodule performs the following temperature data extraction operations: The temperature pixel values ​​of non-hot spot areas in the infrared image are called for statistical analysis. Target equipment, personnel, and heat source areas are excluded, and only the background area with lower temperature and smaller fluctuation in the image is retained for value calculation. It is expressed as the following formula: in, is the background area pixel set selected by image segmentation and temperature threshold filtering, and N is the total number of background pixels; The temperature sensor acquisition submodule is used to collect real-time temperature data from temperature sensors around the monitoring device area; The temperature sensor is used to collect the current ambient temperature value in real time, which is recorded as T sensor ; The data fusion submodule uses the maximum fusion algorithm to fuse the two temperature data. The ambient temperature T env The expression is as follows: T env =max(T ir-bg ,T sensor ) T sensor It is the real-time ambient temperature data of the temperature sensors around the monitoring device area.

10. The system according to any one of claims 1 to 6, characterized in that: For infrared cameras with the same focal length but different models, the corresponding focal length model is evaluated using the temperature measurement error samples of that model. If a stable offset exists, the average error inverse compensation method is used to add a constant offset correction value, which is expressed as the following formula: T corr =T model +ΔT model-specific Among them, T corr is the final corrected temperature value, ΔT model-specific This is a model correction item, which is determined by the systematic deviation of the device model under the standard test scenario.

Images