[go: up one dir, main page]
More Web Proxy on the site http://driver.im/

WO2024027009A1 - 一种变电站绝缘子的红外热成像缺陷检测方法及装置 - Google Patents

一种变电站绝缘子的红外热成像缺陷检测方法及装置 Download PDF

Info

Publication number
WO2024027009A1
WO2024027009A1 PCT/CN2022/123250 CN2022123250W WO2024027009A1 WO 2024027009 A1 WO2024027009 A1 WO 2024027009A1 CN 2022123250 W CN2022123250 W CN 2022123250W WO 2024027009 A1 WO2024027009 A1 WO 2024027009A1
Authority
WO
WIPO (PCT)
Prior art keywords
infrared
temperature
module
defect
substation
Prior art date
Application number
PCT/CN2022/123250
Other languages
English (en)
French (fr)
Inventor
明若彤
杨帆
谭天
易弘睿
李志立
杨子康
姜慧
Original Assignee
重庆大学
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 重庆大学 filed Critical 重庆大学
Publication of WO2024027009A1 publication Critical patent/WO2024027009A1/zh

Links

Images

Classifications

    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N25/00—Investigating or analyzing materials by the use of thermal means
    • G01N25/72—Investigating presence of flaws
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01J—MEASUREMENT OF INTENSITY, VELOCITY, SPECTRAL CONTENT, POLARISATION, PHASE OR PULSE CHARACTERISTICS OF INFRARED, VISIBLE OR ULTRAVIOLET LIGHT; COLORIMETRY; RADIATION PYROMETRY
    • G01J5/00—Radiation pyrometry, e.g. infrared or optical thermometry
    • G01J5/0096—Radiation pyrometry, e.g. infrared or optical thermometry for measuring wires, electrical contacts or electronic systems
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01J—MEASUREMENT OF INTENSITY, VELOCITY, SPECTRAL CONTENT, POLARISATION, PHASE OR PULSE CHARACTERISTICS OF INFRARED, VISIBLE OR ULTRAVIOLET LIGHT; COLORIMETRY; RADIATION PYROMETRY
    • G01J5/00—Radiation pyrometry, e.g. infrared or optical thermometry
    • G01J5/48—Thermography; Techniques using wholly visual means
    • G01J5/485—Temperature profile
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
    • G01R31/00—Arrangements for testing electric properties; Arrangements for locating electric faults; Arrangements for electrical testing characterised by what is being tested not provided for elsewhere
    • G01R31/12—Testing dielectric strength or breakdown voltage ; Testing or monitoring effectiveness or level of insulation, e.g. of a cable or of an apparatus, for example using partial discharge measurements; Electrostatic testing
    • G01R31/1218—Testing dielectric strength or breakdown voltage ; Testing or monitoring effectiveness or level of insulation, e.g. of a cable or of an apparatus, for example using partial discharge measurements; Electrostatic testing using optical methods; using charged particle, e.g. electron, beams or X-rays
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
    • G01R31/00—Arrangements for testing electric properties; Arrangements for locating electric faults; Arrangements for electrical testing characterised by what is being tested not provided for elsewhere
    • G01R31/12—Testing dielectric strength or breakdown voltage ; Testing or monitoring effectiveness or level of insulation, e.g. of a cable or of an apparatus, for example using partial discharge measurements; Electrostatic testing
    • G01R31/1227—Testing dielectric strength or breakdown voltage ; Testing or monitoring effectiveness or level of insulation, e.g. of a cable or of an apparatus, for example using partial discharge measurements; Electrostatic testing of components, parts or materials
    • G01R31/1245—Testing dielectric strength or breakdown voltage ; Testing or monitoring effectiveness or level of insulation, e.g. of a cable or of an apparatus, for example using partial discharge measurements; Electrostatic testing of components, parts or materials of line insulators or spacers, e.g. ceramic overhead line cap insulators; of insulators in HV bushings
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/20—Image preprocessing
    • G06V10/26—Segmentation of patterns in the image field; Cutting or merging of image elements to establish the pattern region, e.g. clustering-based techniques; Detection of occlusion
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
    • G06V10/82—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using neural networks

Definitions

  • the invention belongs to the technical field of insulator detection, and in particular relates to an infrared thermal imaging defect detection method and device for substation insulators.
  • Insulators in power systems not only insulate primary substation equipment from the ground, but also support and fix it.
  • the insulation strength of live equipment against the ground will be seriously affected due to the degradation of insulators leading to a decrease in insulation resistance, and discharge will occur, causing short circuit accidents, making the power grid unable to operate safely and stably.
  • Infrared diagnosis is a common equipment live diagnosis method in substations.
  • the principle is that when equipment in an abnormal state is powered on, it is often accompanied by changes in thermal effects.
  • Infrared temperature measurement can be used to determine problems such as poor contact, insulation deterioration, and abnormal eddy currents.
  • the infrared diagnosis method mainly involves operation and maintenance personnel using handheld infrared thermal imaging cameras to take thermal images of each equipment. After completing the data collection, the images are judged for defects and classified for backup.
  • the existing methods are highly repetitive and require a lot of manpower. For defect types with small heating temperature differences, high professionalism and work experience are required of inspectors.
  • the diagnostic results are subject to a certain degree, and the real-time performance of diagnostic analysis is also poor.
  • the technical problem to be solved by the present invention is to provide a method and device for infrared thermal imaging defect detection of substation insulators to solve the problem of the existing technology of using a handheld infrared thermal imaging camera to take thermal images of each equipment, and to collect the images after completing the data collection.
  • the method of defect judgment and classification backup is highly repetitive and requires a lot of manpower. For defect types with small heating and temperature differences, it requires high professionalism and work experience of the inspectors.
  • the diagnostic results are subject to a certain degree, and the diagnostic analysis is real-time. Poor technical issues.
  • An infrared thermal imaging defect detection device for substation insulators which includes: an infrared module and a core computing module.
  • the infrared module is bidirectionally connected to the core computing module, and the infrared temperature measurement module is also unidirectionally connected to a high-rate battery; the core computing module One-way connection with the touch screen; the infrared module is used to obtain infrared radiation information through the lens and calculate and generate a temperature distribution video.
  • the core computing module is used to control the operation and calculation of the infrared temperature measurement module and realize the adjustment of temperature calculation parameters; the core computing module realizes the infrared image recognition function of the substation insulator and the core computing function of detecting the key temperature information of the insulator area.
  • the substation insulator performs infrared heating. Imaging and fault detection functions.
  • the infrared image recognition of substation insulators adopts the improved Yolov5 algorithm model, and the calculation process is completed by the GPU of the core computing module; the defect diagnosis is implemented based on the adaptive characteristic temperature extraction algorithm, which is jointly completed by the CPU and GPU of the core computing module.
  • the specific process is in the core
  • the compiled Linux executable file is run in the calculation module.
  • the file will load the trained improved Yolov5 algorithm weight file.
  • the GPU performs a convolution operation on the captured video stream or picture.
  • the classifier is input and the pre-selected box Anchorboxes are used to classify the substation.
  • the insulators are identified.
  • the substation insulator defects are diagnosed according to the defect level judgment standards; the touch screen displays the infrared thermal image, the identified substation insulators and the defect levels in real time.
  • the infrared module uses an uncooled focal plane infrared temperature measurement module.
  • the infrared temperature image resolution is 640 ⁇ 480 and above.
  • the infrared module uses Ethernet to connect to the core computing module and uses the RTSP streaming protocol to communicate with the core computing module; the core
  • the computing module uses Nvidia's Jetson XavierNX; the touch screen is connected to the core computing module through the HDMI transmission protocol interface and USB transmission protocol interface, which are used to transmit image signals and control signals respectively.
  • the diagnostic development of the core computing module is based on Ubuntu ARM64; infrared image processing and recognition tasks are completed in the GPU of the core computing module.
  • the parallel computing architecture uses CUDA10.2+cuDnn8.0, and based on this, the Yolov5 convolutional neural network model is optimized.
  • the original RGB three-channel of the captured video stream or picture is changed into a single-channel grayscale image.
  • K-means clustering is performed based on the collected substation insulator data set; the data flow analysis tool DeepStream 5.0 is used to accelerate Real-time device identification is realized, which is used to realize real-time video data processing.
  • the detection method of the device includes:
  • Infrared video collection Turn on the device through the switch button. After the device is started, the infrared temperature measurement module will automatically process and generate infrared thermal images. The operator sets the environmental temperature parameters by touching the display screen and aims the device lens at the target device;
  • Hot zone segmentation Perform sliding window analysis on the thermal image obtained by the infrared module, and segment the square hot zone;
  • the display will display the captured infrared thermal image in real time and display the identified equipment type in text. If the diagnosis result is that there is a defect, the defect level will be displayed in text and the location of the hot spot will be marked in the form of a box. area;
  • the method of segmenting the square hot zone is: 1), set the size of the square window, and the side length value is the same as the infrared video height value; 2), calculate the average temperature under the square window; 3), slide the window several times, and select the average
  • the temperature maximum window segments the image and records the step size and the number of slides corresponding to the hot zone.
  • the identification of substation insulators is based on the improved Yolov5 model of the deep convolutional neural network model Yolov5.
  • the input object is the segmented square hot zone, and the output object is the substation insulator and its voltage level.
  • Key temperature information includes the maximum temperature value and its coordinates, median temperature value and average temperature in the divided square hot zone.
  • the methods for judging substation insulator type and displaying results include:
  • the coordinates of the hot spot in the thermal image before segmentation are calculated based on the number of hot zone slides and the step size;
  • the device of the invention is lightweight and convenient to use.
  • the infrared module imaging terminal based on the improved Yolov5 can greatly improve the efficiency of data processing and calculation.
  • the system has strong service capabilities: the embedded fault detection based on the semantic segmentation algorithm can make the identification more precise. ization, greatly reducing the workload of manual calculations in the later period; the system is convenient to use: the equipment has a real-time diagnosis function of substation insulator equipment defects, and users only need to confirm the automatic diagnosis results to quickly determine the defect level.
  • Figure 1 is a flow chart of image recognition of the infrared thermal imaging detection system for substation insulators
  • Figure 2 is a flow chart of defect diagnosis of the infrared thermal imaging detection system of substation insulators
  • Figure 3 is a schematic diagram of RGB three-channel conversion into a single-channel grayscale image
  • Figure 4 is a schematic diagram of Anchorboxes clustering
  • Figure 5 is a schematic diagram of the detection accuracy of the improved Yolov5 convolutional neural network.
  • An infrared thermal imaging defect detection device for substation insulators including: an infrared module, a touch screen, a core computing module, a USB flash drive or Micro SD card, power control, a high-rate battery, buttons and other peripheral circuits; the infrared module It is bidirectionally connected to the core computing module.
  • the infrared temperature measurement module i.e. infrared module
  • the core computing module is also connected to the 5V power supply control, USB flash drive or Micro SD card, and touch screen. , buttons and other peripheral circuits are connected, and the core computing module and the touch screen are connected in one direction; the infrared module is used to obtain infrared radiation information through the lens and calculate and generate a temperature distribution video.
  • the core computing module is the control system of the device. It is used to control the operation and calculation of the infrared temperature measurement module and realize the adjustment of temperature calculation parameters. Among them, the identification function of substation insulators and the core computing function of detecting key temperature information in the insulator area.
  • the substation insulators conduct infrared Thermal imaging and fault detection functions are all performed in this module.
  • the infrared image recognition of substation insulators is an improved Yolov5 algorithm model, which has strong advantages in rapid deployment of the model.
  • the calculation process is completed by the GPU of the core computing module.
  • Defect diagnosis is based on an adaptive feature temperature extraction algorithm and is jointly completed by the CPU and GPU of the core computing module; the specific process is to run the compiled Linux executable file in the core computing module, which will load the trained improved Yolov5 Algorithm weight file, the GPU enhances the captured video stream or picture (480*640) through the input Mosaic data and calculates the image matrix adaptive anchor frame and adaptive picture scaling. Then through the Focus structure in the backbone part, the image is sliced and input into the subsequent CSP structure. The Neck part adopts the structure of FPN+PAN and draws on the CSP2 structure designed by CSPnet to enhance the feature fusion capability. Finally, the extracted features are finally input into the classifier and the substation insulators are identified by pre-selected Anchorboxes.
  • the substation insulator defects are diagnosed according to the defect level judgment standard;
  • the touch screen can display infrared thermal images, the identified substation insulators and defect levels in real time, and the touch screen can also Used to realize human-computer interaction;
  • the device can be inserted into a Micro SD card or USB flash drive to save image or video data, and can be hot-swapped without turning on video recording;
  • the device is powered by a high-rate battery, in which the infrared module It is directly powered by the high-magnification group, and the core computing module and display are controlled and powered by a 5V power supply.
  • the infrared module uses an uncooled focal plane infrared temperature measurement module.
  • the infrared temperature image resolution is 640 ⁇ 480 and above.
  • the infrared module uses Ethernet to connect to the core computing module and uses the RTSP streaming protocol to communicate with the core computing module.
  • the core computing module uses Nvidia's Jetson
  • the power chip is used to obtain 5V voltage to power the core module and touch display; the power chip provides 18V power supply voltage for JetsonXavierNX, and the infrared module and touch screen of the device require power supply voltages of 12V and 5V respectively, so TPS54540 and
  • the two LM2596 chips step down the input voltage to obtain the required voltage.
  • the power supply part of the circuit uses a 5200mah 25C model aircraft lithium battery as the power supply of the circuit, and uses an IMAX B6AC 80w model aircraft balance charger to charge the power supply.
  • the diagnostic software development of the core computing module is based on Ubuntu ARM64; for infrared image processing and recognition tasks, it is completed in the GPU of the core computing module.
  • the parallel computing architecture adopts CUDA10.2+cuDnn8.0, and based on this, Yolov5 convolution is performed
  • the neural network model is optimized. Since the gray value of the gray channel is linearly related to the temperature value corresponding to the pixel of the actual infrared image taken, it can completely express all the information that needs to be calculated in the image.
  • buttons and other peripheral circuits include switch buttons, shooting buttons and other functional buttons or switches.
  • the switch buttons are connected to the switch interface of the core module.
  • the shooting buttons and other functional buttons or switches are the general input and output of the core computing module. port connection.
  • the segmentation of the image hot zone specifically includes the following steps: 1) Set the size of the square window, whose side length value is the same as the infrared video height value; 2) Calculate the average temperature under the square window; 3) Several sliding windows times, select the window with the largest average temperature, segment the image, and record the step size and the number of slides corresponding to the hot zone.
  • the substation insulator is an improved Yolov5 model based on the deep convolutional neural network model Yolov5.
  • the input object is the square hot zone obtained by segmentation, and the output object is the substation insulator and its voltage level.
  • the extracted key temperature information includes: the highest temperature value and its coordinates, the median temperature value and the average temperature in the divided square hot zone.
  • a diagnostic method based on the device which includes the following steps:
  • S1 infrared video collection The operator first turns on the diagnostic device through the switch button. After the device is started, the infrared temperature measurement module will automatically process and generate infrared thermal images. The operator sets parameters such as ambient temperature through the touch screen (i.e. touch screen). , and just point the device lens at the target device.
  • S2 hot zone segmentation The device performs sliding window analysis on the thermal image obtained by the infrared module and segments the square hot zone.
  • S5 defect diagnosis Combined with the identified equipment type, temperature difference ⁇ T and relative temperature difference ⁇ , query the diagnostic criterion table to determine whether it is in the abnormal range and the corresponding defect nature; the diagnostic criterion table refers to the standard "DL/T 664-2016 Live Equipment” Appendix H and Appendix I of "Infrared Diagnostic Application Specifications" are implemented.
  • S6 result display The display will display the infrared thermal image captured by the lens in real time, and display the identified equipment type in text. If the diagnosis results indicate that there is a defect, the defect level will be displayed in text, and the location of the hot spot will be marked in the form of a box. area. The operator can set whether to display the maximum temperature T 1 , temperature difference ⁇ T and relative temperature difference ⁇ in real time.
  • S7 result saving The operator short-presses the shooting button to save the picture corresponding to the current frame, and generates a JSON file to record the picture number, substation insulator, temperature difference, relative temperature difference, hot spot location, and defect type information; the operator long-presses the shooting button to record infrared Video, touch the screen during the recording process to generate a JSON file to record the video number, time point, device type, temperature difference, relative temperature difference, hot spot location, defect type information.
  • the save function is only available when a Micro SD card or USB flash drive is inserted.
  • the flow chart of image recognition and defect diagnosis of the infrared thermal imaging inspection system for substation insulators includes the following steps:
  • the infrared module imaging end downloads the image of the insulator to be detected from the insulator image storage module.
  • the infrared module imaging end extracts the key parameters of the image through the image preprocessing module.
  • key parameters include infrared resolution, field of view, temperature range, measurement accuracy, etc.
  • the embedded fault detection terminal will display the data processed above in the initialized hot zone segmentation preview module, and finally display the entire interface in the given interface.
  • the Jetson core module is used to control the calculation accuracy of the infrared temperature measurement module and realize temperature calculation parameter adjustment.
  • defect diagnosis will be completed through an intelligent diagnosis algorithm based on multiple feature quantities, including temperature-based time-space gradient and temperature probability density.
  • Parallel acceleration can achieve grouping of pixels according to different semantic meanings expressed in insulator images, dense prediction and inference of labels for each pixel, and each pixel is labeled with the class identification object or region of its shell.
  • the infrared module imaging terminal decides whether to send the event task to the embedded fault detection terminal.
  • the infrared module imaging page can refresh the insulator image; otherwise, the infrared module imaging terminal can refresh the insulator image.
  • the module imaging end will send the event task and parameter data to the embedded fault detection end for processing. After the embedded fault detection end performs processing, it returns the new insulator image to the infrared module imaging end, and the infrared module imaging end re-executes the steps. 2) and step 3).

Landscapes

  • General Physics & Mathematics (AREA)
  • Physics & Mathematics (AREA)
  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Spectroscopy & Molecular Physics (AREA)
  • Health & Medical Sciences (AREA)
  • Evolutionary Computation (AREA)
  • Chemical & Material Sciences (AREA)
  • Multimedia (AREA)
  • General Health & Medical Sciences (AREA)
  • Ceramic Engineering (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Immunology (AREA)
  • Biochemistry (AREA)
  • Analytical Chemistry (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Artificial Intelligence (AREA)
  • Pathology (AREA)
  • Computing Systems (AREA)
  • Databases & Information Systems (AREA)
  • Medical Informatics (AREA)
  • Software Systems (AREA)
  • Radiation Pyrometers (AREA)
  • Investigating Or Analyzing Materials Using Thermal Means (AREA)

Abstract

一种变电站绝缘子的红外热成像缺陷检测方法及装置，装置包括：红外模组和核心计算模块，红外模组与核心计算模块双向连接，红外模组还与高倍率电池单向连接；核心计算模块和触控屏单向连接；红外模组用于通过镜头获取红外辐射信息并计算生成温度分布视频。方法包括红外视频采集、热区分割、设备类型判断、关键温度信息提取、缺陷诊断、结果显示及结果保存。该方法解决了现有技术中缺陷判断和分类备份的方法重复性高，需要花费大量的人力，对于发热温差小的缺陷类型，对检测人员专业性和工作经验要求高，诊断结果存在一定主观性，以及诊断分析实时性较差等技术问题。

Description

一种变电站绝缘子的红外热成像缺陷检测方法及装置
本申请要求于2022年08月03日提交中国专利局、申请号为202210926177.9、发明名称为“一种变电站绝缘子的红外热成像缺陷检测方法及装置”的中国专利申请的优先权，其全部内容通过引用结合在本申请中。
技术领域
本发明属于绝缘子检测技术领域，尤其涉及一种变电站绝缘子的红外热成像缺陷检测方法及装置。
背景技术
绝缘子在电力系统中对一次变电设备不仅起到了带电设备对地之间的绝缘作用，也有着支撑固定的作用。带电设备的对地绝缘强度将因为绝缘子劣化导致绝缘电阻下降受到严重影响，并且产生放电，造成短路事故发生，从而使电网不能安全稳定运行。变电站中绝缘子的数量庞大，在日常运行维护过程中，如何有效、准确、快速地进行检测并加以判别是个关键问题。
红外诊断是变电站常见的设备带电诊断方法，其原理是存在异常状态的设备在带电工作时，往往伴随着热效应的变化，利用红外测温可以判断接触不良、绝缘劣化、异常涡流等问题。目前红外诊断方法主要为运维人员使用手持式红外热像仪拍摄各个设备的热像图，完成数据采集后对图像进行缺陷判断和分类备份。现有方法工作内容重复性高，需要花费大量的人力，对于发热温差小的缺陷类型，对检测人员专业性和工作经验要求高，诊断结果存在一定主观性，诊断分析实时性也较差。
发明内容
本发明要解决的技术问题是：提供一种变电站绝缘子的红外热成像缺陷检测方法及装置，以解决现有技术使用手持式红外热像仪拍摄各个设备的热像图，完成数据采集后对图像进行缺陷判断和分类备份的方法工作内容重复性高，需要花费大量的人力，对于发热温差小的缺陷类型，对检测人员专业性和工作经验要求高，诊断结果存在一定主观性，诊断分析实时性较差等技术问题。
本发明技术方案是：
一种变电站绝缘子的红外热成像缺陷检测装置，它包括：红外模组和核心计算模块，红外模组与核心计算模块双向连接，红外测温模组还与高倍率电池单向连接；核心计算模块和触控屏单向连接；红外模组用于通过镜头获取红外辐射信息并计算生成温度分布视频。
核心计算模块用于控制红外测温模组的运行与计算，实现温度计算参数调整；核心计算模块实现变电站绝缘子的红外图像识别功能，检测绝缘子区域关键温度信息的核心计算功能，变电站绝缘子进行红外热成像及故障检测功能。
变电站绝缘子红外图像识别采用改进的Yolov5算法模型，计算过程由核心计算模块的GPU完成；缺陷诊断基于自适应的特征温度提取算法实现，由核心计算模块的CPU和GPU共同完成，具体过程为在核心计算模块中运行编译好的linux可执行文件，该文件会加载训练好的改进Yolov5算法权重文件，GPU对拍摄好的视频流或图片进行卷积操作，最后输入分类器并由预选框Anchorboxes对变电站绝缘子进行识别，检测出变电站绝缘子后，基于自适应的特征温度提取算法，根据缺陷等级判断标准对变电站绝缘子缺陷进行诊断；触摸显示屏实时显示红外热影像、识别到的变电站绝缘子和缺陷等级。
红外模组采用非制冷焦平面红外测温模组，红外温度图像分辨率为640×480及以上，红外模组使用以太网与核心计算模块连接，使用RTSP流传输协议与核心计算模块通信；核心计算模块采用Nvidia公司的Jetson XavierNX；触控屏通过HDMI传输协议接口和USB传输协议接口与核心计算模块连接，分别用于传输图像信号和控制信号。
核心计算模块的诊断开发基于UbuntuARM64；红外图像处理和识别任务在核心计算模块的GPU中完成，并行计算架构采用CUDA10.2+cuDnn8.0，并基于此对Yolov5卷积神经网络模型进行优化，首先将拍摄提取的视频流或图片本来的RGB三通道变为灰度图像单通道，其次对于预选框Anchorboxes，根据采集到的变电站绝缘子数据集进行K-means聚类；采用数据流分析工具DeepStream 5.0加速实现设备识别实时化，用于实现视频数据处理的实时化。
所述的装置的检测方法，它包括：
S1、红外视频采集：通过开关按键开启装置，待装置启动完成后红外测温模组将自动处理生成红外热影像，操作人员通过触摸显示屏设置环境温度参数，并将装置镜头对准目标设备；
S2、热区分割：将红外模组得到的热影像进行滑窗分析，分割得到正方形热区；
S3、设备类型判断：将正方形热区图像输入设备识别卷积神经网络模型，模型将返回识别到的变电站绝缘子，若未检测到可识别的设备类型则返回“无设备”；
S4、关键温度信息提取：计算正方形热区中的最高温度值T 1及其坐标、温度值中位数T 2和平均温度T 0，并计算温差ΔT＝T 1-T 2，相对温差δ＝(T 1-T 2)/(T 1-T 0)；
S5、缺陷诊断：结合识别到的设备类型、温差ΔT和相对温差δ，查询诊断判据表判断是否处于异常区间及对应的缺陷性质；
S6、结果显示：显示屏将实时显示拍摄的红外热影像，并以文字形式显示识别到的设备类型，若诊断结果为存在缺陷，则以文字显示缺陷等级，以方框的形式标记热点所在位置区域；
S7、结果保存：保存当前帧对应的图片，并生成JSON文件记录图片编号、变电站绝缘子、温差、相对温差、热点位置和缺陷类型信息。
分割得到正方形热区的方法为：1)、设定正方形窗口尺寸，边长值与红外视频高度值相同；2)、计算正方形窗口下的温度平均值；3)、滑动窗口若干次，选取平均温度最大窗口对图像进行分割，并记录步长和热区对应的滑动次数。
变电站绝缘子的识别是基于深度卷积神经网络模型Yolov5的改进Yolov5模型，输入对象为分割得到的正方形热区，输出对象为变电站绝缘子及其电压等级。
关键温度信息包括分割后正方形热区内的最高温度值及其坐标、温度值中位数和平均温度。
变电站绝缘子类型判断和结果显示的方法具体包括：
1)、将最高温视为疑似发热点温度，将温度中位数视作正常温度值， 将平均温度值视作环境温度，并计算温差和相对温差，计算正方形热区中的最高温度值T 1及坐标、温度值中位数T 2和平均温度T 0，并计算温差ΔT＝T 1-T 2，相对温差δ＝(T 1-T 2)/(T 1-T 0)；
2)、根据识别到的电力设备类型，自动查询温差和相对温差值是否处于缺陷范围；
3)、若热点符合缺陷判据，则根据热区滑动次数和步长计算热点在分割前热像图中的坐标；
4)、显示原始的热像数据流，实时显示识别的设备类型，若存在缺陷则用方框标记热点位置。
本发明的有益效果：
本发明装置使用轻量便捷，基于改进的Yolov5的红外模组成像端，可以大大提高数据处理与计算的工作效率；系统服务能力强：基于语义分割算法的嵌入式故障检测，可以使得识别更加精细化，大大减少后期手动计算的工作量；系统使用方式便捷：设备具有变电站绝缘子设备缺陷实时诊断功能，用户仅需对自动诊断结果进行确认，即可快速确定缺陷等级。
解决了现有技术使用手持式红外热像仪拍摄各个设备的热像图，完成数据采集后对图像进行缺陷判断和分类备份的方法工作内容重复性高，需要花费大量的人力，对于发热温差小的缺陷类型，对检测人员专业性和工作经验要求高，诊断结果存在一定主观性，诊断分析实时性较差等技术问题。
说明书附图
图1为变电站绝缘子的红外热成像检测系统图像识别的流程图；
图2为变电站绝缘子的红外热成像检测系统缺陷诊断的流程图；
图3为RGB三通道变为灰度图像单通道示意图；
图4为Anchorboxes聚类示意图；
图5为提升型Yolov5卷积神经网络检测精度示意图。
具体实施方式
一种变电站绝缘子的红外热成像缺陷检测装置，包括：红外模组、触控屏、核心计算模块、优盘或Micro SD卡、电源控制、高倍率电池、按键及其它外围电路；所述红外模组与核心计算模块双向连接，所述红外测 温模组(即红外模组)还与高倍率电池单向连接，所述核心计算模块还分别与5V电源控制、优盘或Micro SD卡、触控屏、按键及其它外围电路相连接，核心计算模块和触控屏单向连接；所述红外模组用于通过镜头获取红外辐射信息并计算生成温度分布视频。
核心计算模块是装置的控制系统，用于控制红外测温模组的运行与计算，实现温度计算参数调整；其中变电站绝缘子的识别功能，检测绝缘子区域关键温度信息的核心计算功能，变电站绝缘子进行红外热成像及故障检测功能均在该模块进行。变电站绝缘子红外图像识别为改进的Yolov5算法模型，在模型的快速部署上具有极强优势。计算过程由核心计算模块的GPU完成。
表1装置性能指标总览表
规格 参数
红外分辨率 640×480
可识别设备 绝缘子
异常识别 准确率90％
检测速率 29帧/秒
测温范围 -20～150℃
测温精度 ±2℃
典型功耗 15W@25℃
尺寸 240×200×135mm
续航时间 5.5小时
缺陷诊断基于自适应的特征温度提取算法实现，由核心计算模块的CPU和GPU共同完成；其具体过程为在核心计算模块中运行编译好的linux可执行文件，该文件会加载训练好的改进Yolov5算法权重文件，GPU对拍摄好的视频流或图片(480*640)经过输入端Mosaic数据增强并将图像矩阵自适应锚框计算、自适应图片缩放。然后在backbone部分经过Focus结构，将图像切片并输入到后续的CSP结构中。Neck部分采用 FPN+PAN的结构并借鉴CSPnet设计的CSP2结构，加强特征融合的能力。最后，将提取好的特征，最后输入分类器并由预选框Anchorboxes对变电站绝缘子进行识别。检测出变电站绝缘子后，基于自适应的特征温度提取算法，根据缺陷等级判断标准对变电站绝缘子缺陷进行诊断；触摸显示屏可实时显示红外热影像、识别到的变电站绝缘子和缺陷等级，触控屏也用于实现人机交互；装置可以插入Micro SD卡或优盘实现图像或视频数据的保存，在不开启视频录制的情况下，可以实现热插拔；本装置由高倍率电池供电，其中红外模组由高倍率组直接供电，核心计算模块和显示屏通过5V电源控制供电。
所述红外模组采用非制冷焦平面红外测温模组，红外温度图像分辨率为640×480及以上，红外模组使用以太网与核心计算模块连接，使用RTSP流传输协议与核心计算模块通信；核心计算模块采用Nvidia公司的Jetson XavierNX；触控屏通过HDMI传输协议接口和USB传输协议接口与核心计算模块连接，分别用于传输图像信号和控制信号，使用高倍率电池为红外模组供电，利用电源芯片得到5V电压为核心模块和触摸显示屏供电；电源芯片为JetsonXavierNX提供18V的供电电压，而装置的红外模组及触控屏需要的供电电压分别为12V、5V，故分别使用TPS54540以及LM2596两种芯片对输入电压进行降压处理得到所需电压。电路的供电部分使用一块5200mah 25C的航模锂电池作为电路的供电电源，使用IMAX B6AC 80w的航模平衡充电器对电源充电。
进一步的，所述核心计算模块的诊断软件开发基于UbuntuARM64；对于红外图像处理和识别任务在核心计算模块的GPU中完成，并行计算架构采用CUDA10.2+cuDnn8.0，并基于此对Yolov5卷积神经网络模型进行优化，由于灰度通道的灰度值与实际拍摄红外图片像素对应的温度值为线性相关关系，可以完整的表达图像所有需要进行运算的信息，故如图3所示，首先将拍摄提取的视频流或图片本来的RGB三通道变为灰度图像单通道，对应的神经网络输入变为原来的1/3，由此节省Yolov5卷积神经网络运算量2/3，最终经过训练后的权重文件大小减少2/5，更轻量化，更有利于搭载在嵌入式设备上。其次如图4所示，对于预选框Anchor boxes，根据采集到的变电站绝缘子数据集进行K-means聚类，最终聚类 为大尺寸、中等尺寸、小尺寸三个大类，大中小三个小类共9个Anchor boxes，更适应绝缘子在不同场景，不同距离的检测，提升了Yolov5卷积神经网络检测精度，其中提升型Yolov5卷积神经网络检测精度示意图参见图5；采用数据流分析工具DeepStream 5.0加速实现设备识别实时化，用于实现视频数据处理的实时化。
进一步的，所述按键及其它外围电路包括开关按键、拍摄按键和其它功能性按键或开关，开关按键连接核心模块的开关接口，拍摄按键和其它功能性按键或开关与核心计算模块的通用输入输出端口连接。
进一步的，所述图像热区的分割具体包括以下步骤：1)设定正方形窗口尺寸，其边长值与红外视频高度值相同；2)计算正方形窗口下的温度平均值；3)滑动窗口若干次，选取平均温度最大窗口，对图像进行分割，并记录步长和热区对应的滑动次数。
进一步的，所述变电站绝缘子是基于深度卷积神经网络模型Yolov5改进的改进Yolov5模型，输入对象为分割得到的正方形热区，输出对象为变电站绝缘子及其电压等级。
进一步的，所述提取的关键温度信息包括：分割后正方形热区内的最高温度值及其坐标、温度值中位数和平均温度。
进一步的，所述变电站绝缘子异常状态的实时判断和结果显示具体包括：1)将最高温视为疑似发热点温度，将温度中位数视作正常温度值，将平均温度值视作环境温度，并按照行业标准DL/T 664中的公式计算温差和相对温差，计算正方形热区中的最高温度值T 1及其坐标、温度值中位数T 2和平均温度T 0，并计算温差ΔT＝T 1-T 2，相对温差δ＝(T 1-T 2)/(T 1-T 0)；2)根据识别到的电力设备类型，自动查询温差和相对温差值是否处于缺陷范围，缺陷界定方法按照行业标准DL/T 664执行；3)若热点符合缺陷判据，则根据热区滑动次数和步长计算热点在分割前热像图中的坐标；4)显示原始的热像数据流，实时显示识别的设备类型，若存在缺陷则用方框标记热点位置。
一种基于所述装置的诊断方法，其包括以下步骤：
S1红外视频采集：操作人员首先通过开关按键开启诊断装置，待装置启动完成后红外测温模组将自动处理生成红外热影像，操作人员通过触 摸显示屏(即触控屏)设置环境温度等参数，并将装置镜头对准目标设备即可。
S2热区分割：装置将红外模组得到的热影像进行滑窗分析，分割得到正方形热区。
S3设备类型判断：将正方形热区图像输入设备识别卷积神经网络模型，模型将返回识别到变电站绝缘子，若未检测到可识别的设备类型则返回“无设备”。
S4关键温度信息提取：计算正方形热区中的最高温度值T 1及其坐标、温度值中位数T 2和平均温度T 0，并计算温差ΔT＝T 1-T 2，相对温差δ＝(T 1-T 2)/(T 1-T 0)。
S5缺陷诊断：结合识别到的设备类型、温差ΔT和相对温差δ，查询诊断判据表判断是否处于异常区间及其对应的缺陷性质；诊断判据表参照标准《DL/T 664-2016带电设备红外诊断应用规范》附录H和附录I执行。
S6结果显示：显示屏将实时显示镜头拍摄的红外热影像，并以文字形式显示识别到的设备类型，若诊断结果认为存在缺陷，则以文字显示缺陷等级，以方框的形式标记热点所在位置区域。操作人员可以设置是否实时显示最高温T 1、温差ΔT和相对温差δ。
S7结果保存：操作人员短按拍摄按键可保存当前帧对应的图片，并生成JSON文件记录图片编号、变电站绝缘子、温差、相对温差、热点位置、缺陷类型信息；操作人员长按拍摄按键可录制红外视频，录制过程触摸屏幕可生成JSON文件记录视频编号、时间点、设备类型、温差、相对温差、热点位置、缺陷类型信息，保存功能仅在插入Micro SD卡或优盘时可用。
如图1及图2所示，变电站绝缘子的红外热成像检测系统图像识别及缺陷诊断的流程图，包括以下步骤：
1)红外模组成像端从绝缘子图像储存模块下载待检测绝缘子图像。
2)图像下载成功之后，红外模组成像端通过图像预处理模块提取图像关键参数，常见的关键参数有红外分辨率、视场角、温度量程、测量精度等。
3)关键参数提取之后，嵌入式故障检测端将经过上述处理好后的数 据在初始化好的热区分割预览模块中显示，最后将整个界面在给定的界面中显示。
4)将上述参数导入改进的Yolov5模型中，训练权重文件后，可进行绝缘子识别。
5)通过核心计算模块，Jetson核心模块，用于控制红外测温模组的计算精度，实现温度计算参数调整。
6)判断识别准确与否。
7)若识别准确，则通过基于多特征量，包括基于温度的时间-空间梯度、温度概率密度的智能诊断算法来完成缺陷诊断。
8)并行加速可以实现根据绝缘子图像中表达的不同语义意义对像素进行分组，对每个像素进行密集的预测和推断标签，每个像素都被标记为其外壳的类识别对象或区域。
9)当交互事件触发时，红外模组成像端决定是否发送事件任务到嵌入式故障检测端，当不需要红外模组成像端的支持时，红外模组成像页面重新刷新绝缘子图像即可；反之红外模组成像端会将事件任务及参数数据发送到嵌入式故障检测端进行处理，嵌入式故障检测端进行处理之后，返回新的绝缘子图像至红外模组成像端，红外模组成像端重新执行步骤2)和步骤3)。

Claims (10)

  1. 一种变电站绝缘子的红外热成像缺陷检测装置，它包括：红外模组和核心计算模块，其特征在于：红外模组与核心计算模块双向连接，红外测温模组还与高倍率电池单向连接；核心计算模块和触控屏单向连接；红外模组用于通过镜头获取红外辐射信息并计算生成温度分布视频。
  2. 根据权利要求1所述的一种变电站绝缘子的红外热成像缺陷检测装置，其特征在于：核心计算模块用于控制红外测温模组的运行与计算，实现温度计算参数调整；核心计算模块实现变电站绝缘子的红外图像识别功能，检测绝缘子区域关键温度信息的核心计算功能，变电站绝缘子进行红外热成像及故障检测功能。
  3. 根据权利要求2所述的一种变电站绝缘子的红外热成像缺陷检测装置，其特征在于：变电站绝缘子红外图像识别采用改进的Yolov5算法模型，计算过程由核心计算模块的GPU完成；缺陷诊断基于自适应的特征温度提取算法实现，由核心计算模块的CPU和GPU共同完成，具体过程为在核心计算模块中运行编译好的linux可执行文件，该文件会加载训练好的改进Yolov5算法权重文件，GPU对拍摄好的视频流或图片进行卷积操作，最后输入分类器并由预选框Anchor boxes对变电站绝缘子进行识别，检测出变电站绝缘子后，基于自适应的特征温度提取算法，根据缺陷等级判断标准对变电站绝缘子缺陷进行诊断；触摸显示屏实时显示红外热影像、识别到的变电站绝缘子和缺陷等级。
  4. 根据权利要求1所述的一种变电站绝缘子的红外热成像缺陷检测装置，其特征在于：红外模组采用非制冷焦平面红外测温模组，红外温度图像分辨率为640×480及以上，红外模组使用以太网与核心计算模块连接，使用RTSP流传输协议与核心计算模块通信；核心计算模块采用Nvidia公司的Jetson XavierNX；触控屏通过HDMI传输协议接口和USB传输协议接口与核心计算模块连接，分别用于传输图像信号和控制信号。
  5. 根据权利要求1所述的一种变电站绝缘子的红外热成像缺陷检测装置，其特征在于：核心计算模块的诊断开发基于Ubuntu ARM64；红外图像处理和识别任务在核心计算模块的GPU中完成，并行计算架构采用CUDA10.2+cuDnn8.0，并基于此对Yolov5卷积神经网络模型进行优化，首先将拍摄提取的视频流或图片本来的RGB三通道变为灰度图像单通 道，其次对于预选框Anchor boxes，根据采集到的变电站绝缘子数据集进行K-means聚类；采用数据流分析工具DeepStream 5.0加速实现设备识别实时化，用于实现视频数据处理的实时化。
  6. 如权利要求1所述的装置的检测方法，其特征在于：它包括：
    S1、红外视频采集：通过开关按键开启装置，待装置启动完成后红外测温模组将自动处理生成红外热影像，操作人员通过触摸显示屏设置环境温度参数，并将装置镜头对准目标设备；
    S2、热区分割：将红外模组得到的热影像进行滑窗分析，分割得到正方形热区；
    S3、设备类型判断：将正方形热区图像输入设备识别卷积神经网络模型，模型将返回识别到的变电站绝缘子，若未检测到可识别的设备类型则返回“无设备”；
    S4、关键温度信息提取：计算正方形热区中的最高温度值T 1及其坐标、温度值中位数T 2和平均温度T 0，并计算温差ΔT＝T 1-T 2，相对温差δ＝(T 1-T 2)/(T 1-T 0)；
    S5、缺陷诊断：结合识别到的设备类型、温差ΔT和相对温差δ，查询诊断判据表判断是否处于异常区间及对应的缺陷性质；
    S6、结果显示：显示屏将实时显示拍摄的红外热影像，并以文字形式显示识别到的设备类型，若诊断结果为存在缺陷，则以文字显示缺陷等级，以方框的形式标记热点所在位置区域；
    S7、结果保存：保存当前帧对应的图片，并生成JSON文件记录图片编号、变电站绝缘子、温差、相对温差、热点位置和缺陷类型信息。
  7. 根据权利要求6所述的检测方法，其特征在于：分割得到正方形热区的方法为：1)、设定正方形窗口尺寸，边长值与红外视频高度值相同；2)、计算正方形窗口下的温度平均值；3)、滑动窗口若干次，选取平均温度最大窗口对图像进行分割，并记录步长和热区对应的滑动次数。
  8. 根据权利要求6所述的检测方法，其特征在于：变电站绝缘子的识别是基于深度卷积神经网络模型Yolov5的改进Yolov5模型，输入对象为分割得到的正方形热区，输出对象为变电站绝缘子及其电压等级。
  9. 根据权利要求6所述的检测方法，其特征在于：关键温度信息包括 分割后正方形热区内的最高温度值及其坐标、温度值中位数和平均温度。
  10. 根据权利要求6所述的检测方法，其特征在于：变电站绝缘子类型判断和结果显示的方法具体包括：
    1)、将最高温视为疑似发热点温度，将温度中位数视作正常温度值，将平均温度值视作环境温度，并计算温差和相对温差，计算正方形热区中的最高温度值T 1及坐标、温度值中位数T 2和平均温度T 0，并计算温差ΔT＝T 1-T 2，相对温差δ＝(T 1-T 2)/(T 1-T 0)；
    2)、根据识别到的电力设备类型，自动查询温差和相对温差值是否处于缺陷范围；
    3)、若热点符合缺陷判据，则根据热区滑动次数和步长计算热点在分割前热像图中的坐标；
    4)、显示原始的热像数据流，实时显示识别的设备类型，若存在缺陷则用方框标记热点位置。
PCT/CN2022/123250 2022-08-03 2022-09-30 一种变电站绝缘子的红外热成像缺陷检测方法及装置 WO2024027009A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202210926177.9 2022-08-03
CN202210926177.9A CN115290696A (zh) 2022-08-03 2022-08-03 一种变电站绝缘子的红外热成像缺陷检测方法及装置

Publications (1)

Publication Number Publication Date
WO2024027009A1 true WO2024027009A1 (zh) 2024-02-08

Family

ID=83826690

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2022/123250 WO2024027009A1 (zh) 2022-08-03 2022-09-30 一种变电站绝缘子的红外热成像缺陷检测方法及装置

Country Status (2)

Country Link
CN (1) CN115290696A (zh)
WO (1) WO2024027009A1 (zh)

Cited By (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117783793A (zh) * 2024-02-23 2024-03-29 泸州老窖股份有限公司 一种开关柜的故障监测方法及系统
CN117953401A (zh) * 2024-03-26 2024-04-30 国网浙江省电力有限公司电力科学研究院 一种复杂背景红外图像复合绝缘子发热缺陷自动识别方法及系统
CN117991061A (zh) * 2024-04-01 2024-05-07 星智元(厦门)科技有限公司 一种基于声纹和红外热成像的绝缘子污闪位置识别方法
CN118237558A (zh) * 2024-05-23 2024-06-25 广州市型腔模具制造有限公司 一种轻量化车身结构件的压铸模具
CN118379602A (zh) * 2024-06-25 2024-07-23 浙江大学 一种用语义和视觉解释增强半导体缺陷分析的方法和系统
CN118470051A (zh) * 2024-07-12 2024-08-09 国网江西省电力有限公司电力科学研究院 基于实例分割的电力设备绝缘子串红外温度异常检测方法
CN118644482A (zh) * 2024-08-14 2024-09-13 广东电网有限责任公司东莞供电局 一种基于改进的一维残差网络的绝缘子发热缺陷智能检测方法、设备及存储介质
CN118799763A (zh) * 2024-09-11 2024-10-18 天津航天中为数据系统科技有限公司 一种绝缘子缺陷识别方法、装置及相关设备

Families Citing this family (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117405239A (zh) * 2023-12-14 2024-01-16 武汉亚为电子科技有限公司 一种基于Modbus协议的红外热像仪及其系统成像方法

Citations (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2005265492A (ja) * 2004-03-17 2005-09-29 Meidensha Corp 電気機器の絶縁劣化診断方法と診断後の電気機器運転方法
US20130136151A1 (en) * 2011-11-24 2013-05-30 National Central University Highly adaptive thermal properties measurement system and measuring method thereof
CN107607207A (zh) * 2017-08-22 2018-01-19 中国科学院深圳先进技术研究院 一种电力设备热故障诊断方法、系统及电子设备
CN110598736A (zh) * 2019-08-06 2019-12-20 西安理工大学 一种电力设备红外图像故障定位、识别与预测方法
CN111679142A (zh) * 2020-06-17 2020-09-18 国网山西省电力公司电力科学研究院 一种便携式输变电设备红外智能诊断装置及方法
CN112697798A (zh) * 2020-12-07 2021-04-23 国网信息通信产业集团有限公司 面向红外图像的变电设备电流致热型缺陷的诊断方法和装置
CN113160062A (zh) * 2021-05-25 2021-07-23 烟台艾睿光电科技有限公司 一种红外图像目标检测方法、装置、设备及存储介质
CN113378856A (zh) * 2021-06-24 2021-09-10 重庆大学 一种基于人工智能的嵌入式输变电装备红外检测方法
CN114494462A (zh) * 2022-02-10 2022-05-13 西安电子科技大学芜湖研究院 一种基于Yolov5和改进跟踪算法的双目相机测距方法

Family Cites Families (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113569956B (zh) * 2021-07-29 2024-09-06 广西电网有限责任公司电力科学研究院 一种基于ai算法的山火灾害勘查识别方法
CN114018983A (zh) * 2021-11-03 2022-02-08 国网福建省电力有限公司莆田供电公司 一种变电设备发热缺陷预测方法、设备和可存储介质

Patent Citations (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2005265492A (ja) * 2004-03-17 2005-09-29 Meidensha Corp 電気機器の絶縁劣化診断方法と診断後の電気機器運転方法
US20130136151A1 (en) * 2011-11-24 2013-05-30 National Central University Highly adaptive thermal properties measurement system and measuring method thereof
CN107607207A (zh) * 2017-08-22 2018-01-19 中国科学院深圳先进技术研究院 一种电力设备热故障诊断方法、系统及电子设备
CN110598736A (zh) * 2019-08-06 2019-12-20 西安理工大学 一种电力设备红外图像故障定位、识别与预测方法
CN111679142A (zh) * 2020-06-17 2020-09-18 国网山西省电力公司电力科学研究院 一种便携式输变电设备红外智能诊断装置及方法
CN112697798A (zh) * 2020-12-07 2021-04-23 国网信息通信产业集团有限公司 面向红外图像的变电设备电流致热型缺陷的诊断方法和装置
CN113160062A (zh) * 2021-05-25 2021-07-23 烟台艾睿光电科技有限公司 一种红外图像目标检测方法、装置、设备及存储介质
CN113378856A (zh) * 2021-06-24 2021-09-10 重庆大学 一种基于人工智能的嵌入式输变电装备红外检测方法
CN114494462A (zh) * 2022-02-10 2022-05-13 西安电子科技大学芜湖研究院 一种基于Yolov5和改进跟踪算法的双目相机测距方法

Cited By (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117783793A (zh) * 2024-02-23 2024-03-29 泸州老窖股份有限公司 一种开关柜的故障监测方法及系统
CN117783793B (zh) * 2024-02-23 2024-05-07 泸州老窖股份有限公司 一种开关柜的故障监测方法及系统
CN117953401A (zh) * 2024-03-26 2024-04-30 国网浙江省电力有限公司电力科学研究院 一种复杂背景红外图像复合绝缘子发热缺陷自动识别方法及系统
CN117991061A (zh) * 2024-04-01 2024-05-07 星智元(厦门)科技有限公司 一种基于声纹和红外热成像的绝缘子污闪位置识别方法
CN117991061B (zh) * 2024-04-01 2024-06-07 星智元(厦门)科技有限公司 一种基于声纹和红外热成像的绝缘子污闪位置识别方法
CN118237558A (zh) * 2024-05-23 2024-06-25 广州市型腔模具制造有限公司 一种轻量化车身结构件的压铸模具
CN118379602A (zh) * 2024-06-25 2024-07-23 浙江大学 一种用语义和视觉解释增强半导体缺陷分析的方法和系统
CN118470051A (zh) * 2024-07-12 2024-08-09 国网江西省电力有限公司电力科学研究院 基于实例分割的电力设备绝缘子串红外温度异常检测方法
CN118644482A (zh) * 2024-08-14 2024-09-13 广东电网有限责任公司东莞供电局 一种基于改进的一维残差网络的绝缘子发热缺陷智能检测方法、设备及存储介质
CN118799763A (zh) * 2024-09-11 2024-10-18 天津航天中为数据系统科技有限公司 一种绝缘子缺陷识别方法、装置及相关设备

Also Published As

Publication number Publication date
CN115290696A (zh) 2022-11-04

Similar Documents

Publication Publication Date Title
WO2024027009A1 (zh) 一种变电站绝缘子的红外热成像缺陷检测方法及装置
CN110598736B (zh) 一种电力设备红外图像故障定位、识别与预测方法
Davari et al. Intelligent diagnosis of incipient fault in power distribution lines based on corona detection in UV-visible videos
Liu et al. Box-point detector: A diagnosis method for insulator faults in power lines using aerial images and convolutional neural networks
WO2020199538A1 (zh) 基于图像监控数据的桥梁关键构件病害预警系统及方法
CN105956618B (zh) 基于图像动静态特征的转炉炼钢吹炼状态识别系统及方法
WO2024000372A1 (zh) 一种缺陷检测的方法和装置
CN109300114A (zh) 高铁接触网支撑装置极小目标零部件顶紧缺失检测方法
CN110493574B (zh) 基于流媒体和ai技术的安监可视化系统
CN113947731A (zh) 一种基于接触网安全巡检的异物识别方法及系统
CN107633266B (zh) 一种电力机车接触网受电弓电弧检测方法
CN116797977A (zh) 巡检机器人动态目标识别与测温方法、装置和存储介质
CN111561967A (zh) 弓网运行状态实时在线检测方法及系统
CN116310274A (zh) 一种输变电设备的状态评估方法
CN112595936A (zh) 基于无人机巡检的输电线复合绝缘子智能检测系统
CN117423018A (zh) 基于无人机采集多模态源融合的电力设备检测系统及方法
CN108596196A (zh) 一种基于绝缘子图像特征词典的污秽状态评估方法
CN115147591A (zh) 一种变电设备红外图像电压致热型缺陷诊断方法及系统
CN115393747A (zh) 一种基于深度学习的光伏故障检测方法
CN112241691B (zh) 基于无人机巡检与图像特征的渠道冰情智能识别方法
Wang et al. Insulator defect detection based on improved YOLOv5 algorithm
CN112132088A (zh) 一种巡检点位漏巡识别方法
CN117095316B (zh) 路面巡检方法、装置、设备及可读存储介质
CN108470141B (zh) 一种基于统计特征与机器学习的配电线路中绝缘子识别方法
CN115049623A (zh) 一种视觉分割分析受电弓轮廓的装置

Legal Events

Date Code Title Description
WWE Wipo information: entry into national phase

Ref document number: 18274448

Country of ref document: US

121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 22953802

Country of ref document: EP

Kind code of ref document: A1