WO2020224458A1 - 一种基于图像处理的电晕检测方法 - Google Patents

一种基于图像处理的电晕检测方法 Download PDF

Info

Publication number
WO2020224458A1
WO2020224458A1 PCT/CN2020/086925 CN2020086925W WO2020224458A1 WO 2020224458 A1 WO2020224458 A1 WO 2020224458A1 CN 2020086925 W CN2020086925 W CN 2020086925W WO 2020224458 A1 WO2020224458 A1 WO 2020224458A1
Authority
WO
WIPO (PCT)
Prior art keywords
image
area
corona
quantization parameter
electrical signal
Prior art date
Application number
PCT/CN2020/086925
Other languages
English (en)
French (fr)
Inventor
陈玉峰
吴甜
刘云鹏
林颖
周加斌
张振军
孙运涛
秦佳峰
高鹏飞
刘海朋
蒲廷燕
吴丹
陈敏
裴少通
李泳霖
纪欣欣
Original Assignee
国网山东省电力公司电力科学研究院
北京瑞盈智拓科技发展有限公司
华北电力大学（保定）
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 国网山东省电力公司电力科学研究院, 北京瑞盈智拓科技发展有限公司, 华北电力大学（保定） filed Critical 国网山东省电力公司电力科学研究院
Priority to EP20802369.7A priority Critical patent/EP3783374A4/en
Publication of WO2020224458A1 publication Critical patent/WO2020224458A1/zh

Links

Images

Classifications

    • 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
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection
    • G06T7/11—Region-based segmentation
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection
    • G06T7/136—Segmentation; Edge detection involving thresholding
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection
    • G06T7/155—Segmentation; Edge detection involving morphological operators
    • 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

Definitions

  • This application relates to the technical field of corona detection, for example, to a corona detection method based on image processing.
  • High-voltage electrical equipment running on site may experience insulation aging, deterioration, damage, cracking, loosening and other accidents under the action of long-term electric field, mechanical stress, and environmental factors.
  • electrical equipment is being designed, manufactured, installed, operated, and maintained. Improper handling of any of these links may cause equipment defects, leading to local electric field concentration and possibly corona discharge.
  • the above-mentioned defects pose a huge threat to the safe operation of power transmission and transformation equipment.
  • the ultraviolet imager began to be used in the discharge detection of electrical equipment.
  • the relevant research is mainly based on application research, qualitative research and model research. It is mainly used to judge the existence of discharge and the effect of discharge. It is unable to determine the relationship between the detection result and the traditional electrical pulse signal, and cannot quantitatively analyze the discharge.
  • This application provides a corona detection method based on image processing, which solves the problem that the ultraviolet imaging technology cannot determine the relationship between the detection result and the traditional electrical pulse signal, and cannot perform quantitative analysis of the discharge.
  • This application provides a corona detection method based on image processing, including:
  • FIG. 1 is a flowchart of a corona detection method based on image processing provided by an embodiment of the application
  • FIG. 2 is an image after grayscale processing provided by an embodiment of the application
  • FIG. 3 is an image after binarization processing provided by an embodiment of the application.
  • FIG. 4 is an image after opening and closing operation processing provided by an embodiment of the application.
  • FIG. 5 is an image after a small area is eliminated according to an embodiment of the application.
  • FIG. 6 is a schematic diagram of eight directions corresponding to a Freeman chain code provided by an embodiment of the application.
  • FIG. 7 is a schematic diagram of a corresponding relationship between photoelectric parameters provided by an embodiment of the application.
  • FIG. 8 is a schematic diagram of the imaging principle of a solar-blind ultraviolet imager provided by an embodiment of the application.
  • the embodiment of the present application provides a corona detection method based on image processing, as shown in FIG. 1, including steps S10-S60.
  • the solar-blind ultraviolet imager used in the examples of this application is South Africa’s CoroCAM504 ultraviolet imager.
  • computer image processing software is used to intercept each frame of the ultraviolet video, and then the original ultraviolet image is binarized to obtain a binary image , And then perform mathematical morphological opening and closing operations on the binary image, and use a small area area elimination algorithm for the image after the opening and closing operations, which can remove the noise regions that are difficult to filter out by morphology.
  • the original image output by the ultraviolet imager is a red, green, and blue (RGB) color digital image.
  • RGB red, green, and blue
  • Each pixel in the image is represented by three color components of red, green, and blue.
  • M and N are the number of rows and columns of the image matrix
  • 3 represents the three color components of red, green, and blue, that is, an image needs three matrices to save, so the image storage space is large.
  • Directly process the original image, and the amount of calculation is also large. Since the image of the discharge area is displayed as white, the color of the image does not affect the extraction of subsequent parameters.
  • the original image is converted into a grayscale image before image segmentation.
  • the grayscale transformation formula as follows:
  • Y is the gray value of the image pixel.
  • R, G, B are the three color component values of red, green and blue.
  • the range of Y is 0-255, and the gray value of all white pixels is "255" ,
  • the gray value of all black pixels is "0", as shown in FIG. 2, which is a gray-scale processed image provided by an embodiment of the application.
  • each discharge area image needs to be segmented from the ultraviolet image.
  • the main algorithms for image segmentation are: edge operator method, area growth method and threshold method.
  • the edges extracted by the edge operator method are often discontinuous, which is inconvenient for the extraction of subsequent parameters.
  • the region growing method needs to manually determine the center of the region. The degree of automation is low, and the edges of some discharges in the field are not continuous, resulting in the generated region The boundary cannot reflect the actual discharge area size.
  • a threshold segmentation algorithm that is, binarization processing
  • the mathematical model of the threshold method is:
  • t is the threshold
  • g(x,y) is the gray value after binarization
  • f(x,y) is the gray value before binarization
  • (x,y) is the pixel coordinate.
  • FIG. 3 is an image after binarization processing provided by an embodiment of the application.
  • this article adopts mathematical form
  • the opening and closing operation of learning has performed filtering processing on the binary image.
  • Corrosion operation and expansion operation refer to the convolution of the original image A and the core (structural element) B.
  • the erosion operation calculates the smallest pixel value in the area covered by the structural element B, and assigns this value to the reference point in the structural element B The specified pixel point;
  • the expansion operation is to calculate the maximum pixel value in the area covered by the structure element B, and also assign this value to the pixel point specified by the reference point in the structure element B. If the set and vector displacement operations are used to represent the corrosion operation and the expansion operation, it is:
  • Equation (3) indicates that the smallest pixel is calculated in the area where A is covered by structural element B, and the set of such pixels is the pixel after the corrosion operation.
  • Equation (4) indicates that the maximum pixel point is calculated in the area covered by the structural element B, and the set of such pixels is the pixel point after the expansion operation.
  • n 1 and n 2 are integers and are related to the size of B. Therefore, from the above equations (3) and (4), it can be seen that if the image is only used for corrosion or dilation, the subsequent parameter calculations will be affected. However, the two operations are not mutually inverse, so they can be used in cascade. First, perform the erosion operation and then the expansion operation as the open operation, and perform the expansion operation first and then the erosion operation as the closing operation.
  • the opening and closing operations are defined as follows:
  • the open operation can eliminate the isolated points higher than the adjacent points, and the thin points between the object boundaries can be smoothed without changing the area obviously; the closed operation can eliminate the isolated points lower than the adjacent points, and the area is not significantly changed.
  • the selected structural element is a circular structure.
  • the radius of the structure element needs to be determined according to the image characteristics. If the structure element is too small, it will not have an effective filtering effect. If the structure element is too large, the image will appear large distortion. In view of the fact that the image diameter of the noise area in the ultraviolet image is generally less than 10 pixels, the radius of the structural element in the embodiment of the present application is selected in the range of 2-5. When filtering, the initial radius is 2. If the filtering effect is not good, add 1 to the image and filter the image again. If the filtering effect is not satisfactory, continue to increase the radius for filtering. If the maximum radius reaches 5, there is still part The noise cannot be filtered out, and the radius will no longer be increased at this time. This indicates that there are some large noise interference points in the image, and the subsequent small area area elimination algorithm can be used to filter out. As shown in FIG. 4, FIG. 4 is an image after opening and closing operation processing provided by an embodiment of the application.
  • the image quantization parameters include the number of photons, the spot area, the perimeter of the area boundary, the long axis and the short axis.
  • Spot area refers to the number of pixels contained in the discharge spot area.
  • Perimeter of regional boundary Similar to the concept of continuous curve length in geometry, the perimeter value is the sum of the number of consecutive pixels on the boundary point.
  • Long axis and short axis The discharge on the surface of the external insulation equipment sometimes appears to be slender. At this time, only the area and circumference parameters cannot reflect the discharge characteristics.
  • two parameters of the long axis and the short axis are introduced.
  • the long axis is defined as the longest distance between the centroid point of the discharge area and two points on the edge.
  • the short axis is defined as the shortest distance between two points on the edge through the centroid point of the discharge area.
  • the approximate method of retrieving the boundary uses the Freeman chain code to access the contour.
  • the Freeman chain code generally uses 8 directions, namely 0, 1, 2, 3, 4, 5, 6, 7 for contour detection.
  • the embodiment of the application A schematic diagram of eight directions provided is shown in Figure 6.
  • the coordinate values of multiple pixels on the boundary of each area can be obtained.
  • the coordinates of the extracted boundary points are connected in sequence to form a closed curve, and then the closed curve is superimposed
  • the label of each area is displayed at the center point of each area, which is convenient for evaluating the extraction effect of the extracted discharge area. If the closed curve matches the boundary of the original image, the extraction effect is good. The curve does not match the boundary of the original image and needs to be extracted again.
  • the coordinate values of the boundary points of the four discharge regions in Figure 5 can be obtained.
  • the multiple coordinate values are connected to form a closed curve and superimposed on the original image.
  • the boundary of the image almost coincides with the original image, that is,
  • the above-mentioned image processing algorithm can effectively segment the discharge light spot while filtering out the interference points, and has little effect on the distortion of the light spot area.
  • a chain code table for each area will be generated, and the chain code table can be linearly transformed to obtain a line segment table.
  • the line segment table contains point coordinates on the boundary of the discharge area. Therefore, the area, circumference, equivalent long axis, equivalent short axis and other parameters of the discharge area can be obtained through the chain code table and the line segment table.
  • the area area is calculated by the total number of pixels in the specified area. If the pixel with the gray value of 255 in the discharge area in the binary image is defined as 1, then the area is the total number of 1 in the area.
  • the calculation formula is as follows:
  • Calculating the perimeter of the spot is to calculate the continuous length of the boundary point of the spot.
  • the distance between adjacent bright spots on the boundary is only 1 or It can be seen from Figure 6 that relative to the current boundary point, the chain code value of the adjacent point with a distance of 1 is 0 or an even number, and the distance is The chain code value of the neighboring points of is odd, if the number and distance of boundary points with a distance of 1 on the entire boundary of the specified area are counted as The number of boundary points and the perimeter calculation formula is:
  • P is the number of boundary points with a distance of 1
  • Q is the distance The number of boundary points.
  • the equivalent long axis calculation is to calculate the longest distance between the two ends of the discharge area through the center of the area.
  • the equivalent minor axis is the shortest distance in the discharge area connecting the two ends through the center of the area.
  • the electrical signal quantization parameters of the corona include discharge amount, current signal peak value, current signal average value, current signal effective value, and current pulse number.
  • the light-electric signal relationship model is a mapping relationship between image quantization parameters and corona electrical signal quantization parameters.
  • the steps of establishing the mapping relationship between the image quantization parameter and the corona electrical signal quantization parameter are: pre-testing a large number of sample data, extracting the image quantization parameter and the corona electrical signal quantization parameter from the sample data respectively, and quantizing the image according to the extracted image
  • the parameters and the electrical signal quantization parameters of the corona establish a photoelectric relationship curve between the photoelectric parameters, and the mapping relationship between the image quantization parameter and the electrical signal quantization parameter of the corona is determined according to the photoelectric relationship curve.
  • the discharge of external insulation equipment has strong randomness, the electrical signal collected by the data acquisition system has high resolution to time, and the ultraviolet imager is limited by the working principle, as shown in Figure 8 for the imaging of the solar-blind ultraviolet imager
  • the principle is to first convert the ultraviolet image into an electronic image by using an ultraviolet photocathode, and then gain and amplify the electronic image through the Microchannel Plate (MCP). It is converted into a visible light image, and then collected and imaged by a Charge Coupled Device (CCD), so the resolution of time is low, which makes the relationship between the electrical signal quantization parameter and the optical signal quantization parameter more complicated. Therefore, the mapping relationship between the electrical signal quantization parameters and the image quantization parameters is a complex non-linear relationship.
  • regression analysis is used to establish a quantitative relationship between the image quantization parameters and the electrical signal quantization parameters.
  • regression analysis refers to a statistical analysis method to determine the quantitative relationship between two or more variables.
  • regression analysis is a predictive modeling technique. What is studied is the relationship between the dependent variable (target) and the independent variable (predictor). This technique is commonly used in predictive analysis, time series models, and discovering causal relationships between variables.
  • the regression analysis method used in the embodiments of the present application may be least squares regression, neural network regression, support vector machine regression, and nuclear regression. In this way, after a large amount of sample data, the optical-electric signal relationship model can be established using regression analysis.
  • the light-electric signal relationship model can be used in corona detection, as long as the image in the ultraviolet video output by the solar-blind ultraviolet imager is image processed to obtain the image quantization parameters, and then According to the image quantization parameters and the light-electric signal relationship model, the electrical signal quantization parameters (ie discharge volume, current signal peak value, current signal average value, current signal effective value, number of current pulses) can be obtained, thus realizing judgment and detection The relationship between the result and the traditional electric pulse signal, and realizes the quantitative analysis of the discharge.
  • the electrical signal quantization parameters ie discharge volume, current signal peak value, current signal average value, current signal effective value, number of current pulses

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Theoretical Computer Science (AREA)
  • Image Analysis (AREA)
  • Image Processing (AREA)

Abstract

本文公开了一种基于图像处理的电晕检测方法，所述方法包括：截取日盲紫外成像仪输出的紫外视频中的每一帧图像；对截取的图像进行二值化处理；对二值化处理后的图像进行数学形态学的开运算和闭运算；对开运算和闭运算后的图像利用小区域面积消除算法，以消除所述开运算和闭运算后的图像中的噪声区域；对消除噪声后的图像提取图像量化参数；根据所述图像量化参数以及预先建立的光-电信号关系模型计算电晕的电信号量化参数。

Description

一种基于图像处理的电晕检测方法
本申请要求在2019年05月07日提交中国专利局、申请号为201910376702.2的中国专利申请的优先权，该申请的全部内容通过引用结合在本申请中。
技术领域
本申请涉及电晕检测技术领域，例如涉及一种基于图像处理的电晕检测方法。
背景技术
运行在现场的高压电气设备在长期的电场、机械应力、环境因素的作用下，可能会发生绝缘老化、劣化、破损、开裂、松动等事故；同时电气设备在设计、制造、安装、运行、维护中任一环节若处理不当都有可能造成设备缺陷，导致局部电场集中进而可能形成电晕放电，上述缺陷对输变电设备的安全运行造成巨大威胁。根据放电中伴随有紫外光信号辐射的特点，在电力系统中，开始将紫外成像仪用于电气设备的放电检测中。
由于紫外成像技术在电力系统中的应用时间还不长，综合公开的研究资料来看，相关研究主要以应用研究、定性研究和模型研究为主，主要用于判断放电的有无以及对放电的定位，而不能判断检测结果与传统的电脉冲信号的关系，无法对放电进行定量化分析。
发明内容
本申请提供了一种基于图像处理的电晕检测方法，解决紫外成像技术不能判断检测结果与传统的电脉冲信号之间的关系，无法对放电进行定量化分析的问题。
本申请提供了一种基于图像处理的电晕检测方法，包括：
截取日盲紫外成像仪输出的紫外视频中的每一帧图像；
对截取的图像进行二值化处理；
对二值化处理后的图像进行数学形态学的开运算和闭运算；
对开运算和闭运算后的图像利用小区域面积消除算法，以消除所述开运算和闭运算后的图像中的噪声区域；
对消除噪声后的图像提取图像量化参数；
根据所述图像量化参数以及预先建立的光-电信号关系模型计算电晕的电信号量化参数。
附图说明
图1为本申请实施例提供的一种基于图像处理的电晕检测方法的流程图；
图2为本申请实施例提供的一种灰度处理后的图像；
图3为本申请实施例提供的一种二值化处理后的图像；
图4为本申请实施例提供的一种开闭运算处理后的图像；
图5为本申请实施例提供的一种小面积消除后的图像；
图6为本申请实施例提供的一种Freeman链码对应的八个方向的示意图；
图7为本申请实施例提供的一种光电参数之间的对应关系的示意图；
图8为本申请实施例提供的一种日盲紫外成像仪的成像原理的示意图。
具体实施方式
以下通过具体实施例说明本申请的实施方式。本申请还可以通过另外不同的具体实施方式加以实施或应用。
以下实施例中所提供的图示仅以示意方式说明本申请的基本构想，遂图式中仅显示与本申请中有关的组件而非按照实际实施时的组件数目、形状及尺寸绘制，其实际实施时每个组件的型态、数量及比例可为一种随意的改变，且其组件布局型态也可能更为复杂。
本申请实施例提供了一种基于图像处理的电晕检测方法，如图1所示，包括步骤S10-S60。
S10.截取日盲紫外成像仪输出的紫外视频中的每一帧图像。
S20.对截取的图像进行二值化处理。
S30.对二值化处理后的图像进行数学形态学的开运算和闭运算。
S40.对开运算和闭运算后的图像利用小区域面积消除算法，以消除开运算和闭运算后的图像中的噪声区域。
S50.对消除噪声后的图像提取图像量化参数。
S60.根据所述图像量化参数以及预先建立的光-电信号关系模型计算电晕的电信号量化参数。
本申请实施例所使用的日盲紫外成像仪为南非CoroCAM504紫外成像仪， 首先，利用计算机图像处理软件从紫外视频中截取每一帧图像，然后对原始紫外图像进行二值化处理得到二值图像，然后将二值图像进行数学形态学的开运算和闭运算，并对开运算和闭运算后的图像采用小区域面积消除算法，可以去掉形态学难以滤除的噪声区域。
紫外成像仪输出的原始图像为红绿蓝(Red Green Blue，RGB)彩色数字图像，图像中的每一个像素由红、绿、蓝三个颜色分量表示，在计算机种以M×N×3的形式保存，M和N分别为图像矩阵的行数和列数，3表示红、绿、蓝三个颜色分量，也即一幅图像需要三个矩阵进行保存，因而占用的图像存储空间较大，直接对原始图像进行处理，计算量也较大。由于放电区域的图像显示为白色，因而图像的颜色不影响后续参数的提取，为便于后续处理，本申请实施例中，在进行图像分割之前将原始图像转换为灰度图像，灰度变换的公式如下：
Y＝0.299R+0.587G+0.114B     (1)
Y的物理意义就是图像像素点的灰度值，R，G，B为红、绿、蓝三个颜色分量值，Y的范围为0-255，全白像素点的灰度值为“255”，全黑像素点的灰度值为“0”，如图2所示，图2为本申请实施例提供的一种灰度处理后的图像。
为提取图像量化参数需将每个放电区域图像从紫外图像中分割出来，对图像进行分割的主要算法有：边缘算子法、区域生长法和阈值法。但边缘算子法提取的边缘往往不连续，不便于后续参数的提取，区域生长法需要人为确定区域的中心，提取的自动化程度低，且现场有些放电的边缘本身就不连续，导致生成的区域边界不能反映真实的放电区域大小。鉴于放电区域灰度值明显高于背景图像的灰度值，本申请实施例采用了阈值分割算法(即二值化处理)。
阈值法的原理为将灰度图的每个像素在一特定值范围内赋予白色(Y＝255)或黑色(Y＝0)，阈值法的数学模型为：
Figure PCTCN2020086925-appb-000001
式(2)中t为阈值；g(x,y)为二值化后的灰度值；f(x,y)为二值化前的灰度值，(x,y)为像素坐标。
鉴于紫外图像中的放电区域的图像较白，而背景图像的灰度值一般远低于放电区域的灰度值，从大量的测试灰度化图像的直方图中发现在灰度值为220左右存在一个明显的波谷，因而先选择220作为默认阈值，若处理效果不好，则可以改变阈值对图像再次进行阈值分割，直到取得较好的效果为止。如图3所示，图3为本申请实施例提供的一种二值化处理后的图像。
图像经过二值化后虽然可以达到一定的去噪效果，但对于一些灰度值接近 于放电区域的像素点经过二值化后仍无法去除噪声，为了更准确地提取放电区域，本文采用数学形态学的开闭运算对二值图像进行了滤波处理。
形态学基本运算包括腐蚀运算与膨胀运算。腐蚀运算与膨胀运算是指将原图像A与核(结构元素)B进行卷积，腐蚀运算是计算结构元素B所覆盖区域中最小像素值，并将该值赋给结构元素B中参考点所指定的像素点；膨胀运算是计算结构元素B所覆盖区域中最大像素值，同样将该值赋给结构元素B中参考点所指定的像素点。若用集合和向量位移运算表示腐蚀运算与膨胀运算即为：
Figure PCTCN2020086925-appb-000002
Figure PCTCN2020086925-appb-000003
式(3)表示在A被结构元素B覆盖的区域中计算最小像素点，则这样的像素点所构成的集合为腐蚀运算后的像素点。式(4)表示在A被结构元素B覆盖的区域中计算最大像素点，则这样的像素点所构成的集合为膨胀运算后的像素点。n 1，n 2为整数，与B的尺寸有关。所以由上式(3)和(4)可知若对图像单纯运用腐蚀运算或膨胀运算会影响后续的参数计算。但是两种运算并不互逆，因此可以级联运用，先进行腐蚀运算再进行膨胀运算为开运算，先进行膨胀运算再进行腐蚀运算为闭运算，对开运算和闭运算定义如下：
Figure PCTCN2020086925-appb-000004
Figure PCTCN2020086925-appb-000005
开运算可以消除高于邻近点的孤立点，在不明显改变面积的情况下消除物体边界间的纤细点同时起到平滑作用；闭运算可以消除低于邻近点的孤立点，在不明显改变面积的情况下填充并平滑邻近物体的边界。考虑到放电区域通常呈圆形，因此选择的结构元素为圆形结构。
结构元素的半径需要根据图像特征来确定，结构元素过小，起不到有效的滤波效果，结构元素过大，会导致图像出现较大的畸变。鉴于紫外图像中噪声区域的图像直径一般在10个像素点以下，本申请实施例中结构元素的半径选择2～5的范围。滤波时初始半径为2，若滤波效果不好，则将半径加1，对图像再次进行滤波，若滤波效果还不理想，则继续增加半径进行滤波，若最大的半径达到5后，仍然有部分噪声不能滤除，此时将不再继续增加半径，这说明图像中存在部分较大的噪声干扰点，可采用后续的小区域面积消除算法进行滤除。如图4所示，图4为本申请实施例提供的一种开闭运算处理后的图像。
经过数学形态滤波处理后，图4中仍然有大量的干扰点，但与电光斑相比干扰点的面积小得多，为此本文采用了二值图像的小区域面积消除算法去除图4中的干扰点，该算法的基本步骤如下：
a)检测图4中每一个高亮区域，获取每一个区域的边界信息。
b)通过边界信息获取每个连通区域的面积大小。统计可知图中共有12个连通区域，每个区域区域包含的像素点的个数如表1所示。
表1每个区域包含的像素点的个数
Figure PCTCN2020086925-appb-000006
c)设定阈值为100，依次将每个区域面积与该阈值进行对比，大于该阈值的区域保留下来，而小于该阈值的区域会被消除，得到如图5所示的图像。在滤波后，得到的放电区域个数为4。
从图5可知，上述算法有效地滤除了噪声区域并保存了放电的区域，同时该滤波方法对放电区域自身的图像大小和形状无任何影响。
在进行小区域面积消除之后，提取图像量化参数，所述的图像量化参数包括光子数、光斑面积、区域边界周长、长轴和短轴。
光子数：由紫外成像仪直接输出得到。
光斑面积：是指放电光斑区域内所包含的像素点的个数。
区域边界周长：类似于几何学中的连续曲线长度的概念，该周长值即为边界点上的连续像素点的个数之和。
长轴和短轴：外绝缘设备表面的放电有时表现为细长型，此时仅利用面积和周长参数还不能较好地反映放电特征，在此引入了长轴和短轴两个参数，长轴，定义为通过放电区域的形心点，边缘上两点之间的最长距离。短轴，定义为通过放电区域的形心点，边缘上两点之间的最短距离。
要获得放电区域图像的量化参数需提取每个放电区域的轮廓边界点，由于紫外图像中的放电点往往不止一个，因此多区域边界轮廓点的坐标提取是本算法中的一个核心环节，本申请实施例采用了多区域边界跟踪算法，基本原理和步骤如下：
a)检测图5中每一个高亮区域，获取每一个区域的边界信息。
b)通过边界信息获取每个连通区域的面积大小。
c)设定一个阈值，依次将每个区域面积与该阈值进行对比，区域面积大于该阈值的区域保留下来，区域面积小于该阈值的区域会被消除。
采用检索边界的近似方法使用Freeman链码存取轮廓，Freeman链码一般是采用8个方向，即0、1、2、3、4、5、6、7的方式进行轮廓检测，本申请实施例提供的一种八个方向的示意图如图6所示。
以初始点为中心，以逆时针或顺时针的方式检测八个方向的像素值，并按照从上到下，从左到右的方式检测整个图，当检测到边界点时，则定义该点为检测边缘起始点，并保存在链码内，以此继续向下检测，直到回到边缘起始点，则完成了一个轮廓的检测，并将表示轮廓的八个方向的编码存储于链码内。
至此可以获得每个区域的边界上多个像素点的坐标值，为了直观地显示出所提取的边界是否满足需要，将提取的边界点的坐标依次连线构成一条封闭曲线，然后将该封闭曲线叠加到了原始图像之上，在每个区域的中心点显示出该区域的标号，这样便于评估所提取放电区域的提取效果，若封闭曲线与原始图像的边界较吻合，说明提取的效果好，若封闭曲线与原始图像的边界不吻合，需要重新进行提取。
采用多区域轮廓跟踪算法后，可得到图5中四个放电区域边界点的坐标值，将多个坐标值连接起来构成封闭曲线叠加到原始图像中，图像的边界与原图几乎重合，也即采用上述图像处理算法在滤除干扰点的同时，还能有效的分割出放电光斑，并且对光斑区域的畸变作用很小。
经过轮廓提取后，会产生每个区域的链码表，对链码表进行线性变换可得到线段表，线段表内存有放电区域边界上的点坐标。因此通过链码表和线段表可以求出放电区域的面积、周长、等效长轴、等效短轴等参数。
(1)光斑面积S
区域面积的计算方法是指定区域内像素点的总数和。若将二值图像中放电区域中灰度值为255的像素点定义为1，则面积为该区域中1的总数。计算式如下：
Figure PCTCN2020086925-appb-000007
(2)区域边界周长
计算光斑周长即计算光斑的边界点的连续长度综合，边界上相邻亮点的距离只有1或
Figure PCTCN2020086925-appb-000008
由图6可以看出相对于当前边界点，距离为1的相邻点的链码 值为0或偶数，距离为
Figure PCTCN2020086925-appb-000009
的相邻点的链码值为奇数，若统计出指定区域整个边界上距离为1的边界点的个数及距离为
Figure PCTCN2020086925-appb-000010
的边界点的个数则周长计算公式为：
Figure PCTCN2020086925-appb-000011
P为距离为1的边界点的个数，Q为距离为
Figure PCTCN2020086925-appb-000012
的边界点的个数。
(3)等效直径、等效长轴短轴计算
由于放电区域近似圆形，所以计算放电光斑的直径时，利用圆面积计算公式得出等效直径。
等效长轴计算即计算放电区域内通过区域中心连接两端的最长距离。等效短轴为放电区域内通过区域中心连接两端的最短距离。
基于上述参数的定义方法，对图5的四个放电区域计算得到的量化参数表如表2所示：
表2提取紫外检测放电量化参数
放电点编号 光斑面积 光斑周长 光斑等效长轴 光斑等效短轴
1 350.5 87.3 27.65 13.58
2 717 122.33 43.57 22.85
3 738 105.25 34.4 26.92
4 1249.5 182.88 71.98 24.51
所述电晕的电信号量化参数包括放电量、电流信号峰值、电流信号平均值、电流信号有效值、电流脉冲个数。
所述光-电信号关系模型为图像量化参数与电晕的电信号量化参数的映射关系。
所述图像量化参数与电晕的电信号量化参数的映射关系建立的步骤为：预先试验大量的样本数据，对样本数据分别提取图像量化参数与电晕的电信号量化参数，根据提取的图像量化参数与电晕的电信号量化参数建立光电参数之间的光电关系曲线，根据所述光电关系曲线确定图像量化参数与电晕的电信号量化参数的映射关系。
鉴于放电具有一定的随机性，通过试验获得大量的样本数据，对样本数据分别提取电信号量化参数和图像量化参数，依次建立如图7中光电参数之间的光电关系曲线，可以获得电信号与光信号具有强关联关系的量化参数。
外绝缘设备的放电具有较强的随机性，数据采集系统采集到的电信号对时间的分辨率高，而紫外成像仪受到工作原理的限制，如图8所示为日盲紫外成像仪的成像原理，先利用紫外光电阴极将紫外图像转换为电子图像，然后经微通道板(Microchannel Plate，MCP)对电子图像增益放大，MCP输出的电子流高速轰击到MCP后部的荧光屏上，将电子图像又转换为可见光图像，然后经电荷耦合器件(Charge Coupled Device，CCD)采集成像，因此对时间的分辨率较低，这就使得电信号量化参数与光信号量化参数之间的关系必然比较复杂，所以电信号量化参数与图像量化参数之间映射关系是复杂的非线性关系，利用回归分析建立图像量化参数与电信号量化参数之间定量的关联关系。在统计学中，回归分析指的是确定两种或两种以上变量间相互依赖的定量关系的一种统计分析方法，在大数据分析中，回归分析是一种预测性的建模技术，它研究的是因变量(目标)和自变量(预测器)之间的关系。这种技术通常用于预测分析，时间序列模型以及发现变量之间的因果关系。本申请实施例所使用的回归分析方法可以为最小二乘回归、神经网络回归、支持向量机回归、核回归，这样通过大量的样本数据后，利用回归分析就可建立光-电信号关系模型。
在建立好光-电信号关系模型之后就可将光-电信号关系模型用于电晕检测中，只要对日盲紫外成像仪输出的紫外视频中的图像进行图像处理后得到图像量化参数，然后根据图像量化参数以及利用光-电信号关系模型就可以获得电信号量化参数(即放电量、电流信号峰值、电流信号平均值、电流信号有效值、电流脉冲个数)，这样就实现了判断检测结果与传统的电脉冲信号之间的关系，并实现了对放电进行定量化分析。

Claims (7)

  1. 一种基于图像处理的电晕检测方法，包括：
    截取日盲紫外成像仪输出的紫外视频中的每一帧图像；
    对截取的图像进行二值化处理；
    对二值化处理后的图像进行数学形态学的开运算和闭运算；
    对开运算和闭运算后的图像利用小区域面积消除算法，以消除所述开运算和闭运算后的图像中的噪声区域；
    对消除噪声后的图像提取图像量化参数；
    根据所述图像量化参数以及预先建立的光-电信号关系模型计算电晕的电信号量化参数。
  2. 根据权利要求1所述的方法，其中，所述图像量化参数包括光子数、光斑面积、区域边界周长、长轴和短轴。
  3. 根据权利要求2所述的方法，其中，所述光斑面积表示放电光斑区域内所包含的像素点的个数；所述区域边界周长表示放电光斑区域内边界点上的连续像素点的个数之和；所述长轴表示放电光斑区域边缘上两点之间的最长距离；所述短轴表示放电光斑区域边缘上两点之间的最短距离。
  4. 根据权利要求1-3任一项所述的方法，其中，所述电晕的电信号量化参数包括放电量、电流信号峰值、电流信号平均值、电流信号有效值、电流脉冲个数。
  5. 根据权利要求4所述的方法，其中，所述光-电信号关系模型为图像量化参数与电晕的电信号量化参数的映射关系。
  6. 根据权利要求5所述的方法，还包括：
    预先试验多个样本数据，对所述多个样本数据分别提取图像量化参数与电晕的电信号量化参数；
    根据提取的图像量化参数与电晕的电信号量化参数建立光电参数之间的光电关系曲线；
    根据所述光电关系曲线确定所述图像量化参数与所述电晕的电信号量化参数的映射关系，以得到所述光-电信号关系模型。
  7. 根据权利要求6所述的方法，其中，根据所述光电关系曲线确定所述图像量化参数与所述电晕的电信号量化参数的映射关系的方法为以下至少之一：最小二乘回归、神经网络回归、支持向量机回归、核回归。
PCT/CN2020/086925 2019-05-07 2020-04-26 一种基于图像处理的电晕检测方法 WO2020224458A1 (zh)

Priority Applications (1)

Application Number Priority Date Filing Date Title
EP20802369.7A EP3783374A4 (en) 2019-05-07 2020-04-26 METHOD OF DETECTING CORONA DISCHARGE USING IMAGE PROCESSING

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201910376702.2 2019-05-07
CN201910376702.2A CN110146791A (zh) 2019-05-07 2019-05-07 一种基于图像处理的电晕检测方法

Publications (1)

Publication Number Publication Date
WO2020224458A1 true WO2020224458A1 (zh) 2020-11-12

Family

ID=67594958

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2020/086925 WO2020224458A1 (zh) 2019-05-07 2020-04-26 一种基于图像处理的电晕检测方法

Country Status (3)

Country Link
EP (1) EP3783374A4 (zh)
CN (1) CN110146791A (zh)
WO (1) WO2020224458A1 (zh)

Cited By (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113298832A (zh) * 2021-07-02 2021-08-24 华北电力大学 一种放电紫外视频量化参数提取和显示方法及其应用
CN113406448A (zh) * 2021-06-15 2021-09-17 中国铁道科学研究院集团有限公司基础设施检测研究所 铁路绝缘子电气状态检测方法及装置
CN113538351A (zh) * 2021-06-30 2021-10-22 国网山东省电力公司电力科学研究院 一种融合多参数电信号的外绝缘设备缺陷程度评估方法
CN113933563A (zh) * 2021-09-29 2022-01-14 国电南瑞科技股份有限公司 基于自适应迭代运算数学形态法的采样异常大值滤除方法、装置及系统
CN113962136A (zh) * 2021-12-22 2022-01-21 广东工业大学 一种基于有限元的焊接后工件应力重构方法及系统
CN114166852A (zh) * 2021-12-06 2022-03-11 国网宁夏电力有限公司超高压公司 基于多光谱检测的平波电抗器在线监测方法及系统
CN114779031A (zh) * 2022-06-21 2022-07-22 国网山东省电力公司电力科学研究院 一种数字化电力设备紫外成像放电异常检测方法及系统
CN115375022A (zh) * 2022-08-18 2022-11-22 中国南方电网有限责任公司超高压输电公司广州局 电极脉冲电流数据的处理方法、装置、计算机设备

Families Citing this family (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110146791A (zh) * 2019-05-07 2019-08-20 国网山东省电力公司电力科学研究院 一种基于图像处理的电晕检测方法
CN111493853A (zh) * 2020-04-24 2020-08-07 天津恒宇医疗科技有限公司 一种针对血管性皮肤病的血管参数评价方法及系统
CN112857751B (zh) * 2021-01-14 2023-06-06 北方夜视技术股份有限公司 一种数字化像增强器暗计数测试装置、方法及存储介质
CN114113947A (zh) * 2021-11-30 2022-03-01 国网辽宁省电力有限公司铁岭供电公司 一种基于紫外成像法的开关柜及其放电状态感知方法
CN115205246B (zh) * 2022-07-14 2024-04-09 中国南方电网有限责任公司超高压输电公司广州局 换流阀电晕放电紫外图像特征提取方法和装置

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103018640A (zh) * 2012-11-27 2013-04-03 华北电力大学（保定） 高压绝缘子表面电晕放电强度测试方法
CN103543394A (zh) * 2013-10-27 2014-01-29 华北电力大学（保定） 一种高压电气设备放电紫外成像量化参数提取方法
CN204479836U (zh) * 2014-01-10 2015-07-15 奥费有限公司 观察装置
CN105004972A (zh) * 2015-06-25 2015-10-28 华北电力大学（保定） 基于日盲紫外成像图像特征的瓷绝缘子绝缘状态评估方法
CN106054032A (zh) * 2016-03-08 2016-10-26 华北电力大学（保定） 一种高压绝缘子沿面放电脉冲峰值的非接触式测量方法
CN110146791A (zh) * 2019-05-07 2019-08-20 国网山东省电力公司电力科学研究院 一种基于图像处理的电晕检测方法

Family Cites Families (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106124942B (zh) * 2016-06-27 2019-06-04 华北电力大学（保定） 一种基于红外和紫外成像法的零值绝缘子检测方法
CN106940886B (zh) * 2017-03-08 2019-11-22 贵州众创巨电力科技有限公司 一种基于灰度的电气设备放电紫外成像量化参数提取方法
CN107505546A (zh) * 2017-08-25 2017-12-22 国家电网公司 一种利用紫外成像仪监测电晕放电的方法

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103018640A (zh) * 2012-11-27 2013-04-03 华北电力大学（保定） 高压绝缘子表面电晕放电强度测试方法
CN103543394A (zh) * 2013-10-27 2014-01-29 华北电力大学（保定） 一种高压电气设备放电紫外成像量化参数提取方法
CN204479836U (zh) * 2014-01-10 2015-07-15 奥费有限公司 观察装置
CN105004972A (zh) * 2015-06-25 2015-10-28 华北电力大学（保定） 基于日盲紫外成像图像特征的瓷绝缘子绝缘状态评估方法
CN106054032A (zh) * 2016-03-08 2016-10-26 华北电力大学（保定） 一种高压绝缘子沿面放电脉冲峰值的非接触式测量方法
CN110146791A (zh) * 2019-05-07 2019-08-20 国网山东省电力公司电力科学研究院 一种基于图像处理的电晕检测方法

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
See also references of EP3783374A4

Cited By (13)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113406448A (zh) * 2021-06-15 2021-09-17 中国铁道科学研究院集团有限公司基础设施检测研究所 铁路绝缘子电气状态检测方法及装置
CN113538351B (zh) * 2021-06-30 2024-01-19 国网山东省电力公司电力科学研究院 一种融合多参数电信号的外绝缘设备缺陷程度评估方法
CN113538351A (zh) * 2021-06-30 2021-10-22 国网山东省电力公司电力科学研究院 一种融合多参数电信号的外绝缘设备缺陷程度评估方法
CN113298832A (zh) * 2021-07-02 2021-08-24 华北电力大学 一种放电紫外视频量化参数提取和显示方法及其应用
CN113298832B (zh) * 2021-07-02 2024-02-02 华北电力大学 一种放电紫外视频量化参数提取和显示方法及其应用
CN113933563A (zh) * 2021-09-29 2022-01-14 国电南瑞科技股份有限公司 基于自适应迭代运算数学形态法的采样异常大值滤除方法、装置及系统
CN113933563B (zh) * 2021-09-29 2024-04-26 国电南瑞科技股份有限公司 基于自适应迭代运算数学形态法的采样异常大值滤除方法、装置及系统
CN114166852A (zh) * 2021-12-06 2022-03-11 国网宁夏电力有限公司超高压公司 基于多光谱检测的平波电抗器在线监测方法及系统
CN114166852B (zh) * 2021-12-06 2024-03-15 国网宁夏电力有限公司超高压公司 基于多光谱检测的平波电抗器在线监测方法及系统
CN113962136A (zh) * 2021-12-22 2022-01-21 广东工业大学 一种基于有限元的焊接后工件应力重构方法及系统
CN114779031A (zh) * 2022-06-21 2022-07-22 国网山东省电力公司电力科学研究院 一种数字化电力设备紫外成像放电异常检测方法及系统
CN115375022A (zh) * 2022-08-18 2022-11-22 中国南方电网有限责任公司超高压输电公司广州局 电极脉冲电流数据的处理方法、装置、计算机设备
CN115375022B (zh) * 2022-08-18 2024-04-05 中国南方电网有限责任公司超高压输电公司广州局 电极脉冲电流数据的处理方法、装置、计算机设备

Also Published As

Publication number Publication date
EP3783374A4 (en) 2022-02-23
CN110146791A (zh) 2019-08-20
EP3783374A1 (en) 2021-02-24

Similar Documents

Publication Publication Date Title
WO2020224458A1 (zh) 一种基于图像处理的电晕检测方法
CN112419250B (zh) 路面裂缝数字图像提取、裂缝修补及裂缝参数计算方法
CN111260616A (zh) 一种基于Canny算子二维阈值分割优化的绝缘子裂纹检测方法
CN112614062B (zh) 菌落计数方法、装置及计算机存储介质
CN107784669A (zh) 一种光斑提取及其质心确定的方法
CN110309806B (zh) 一种基于视频图像处理的手势识别系统及其方法
CN112149543B (zh) 一种基于计算机视觉的建筑扬尘识别系统与方法
CN109410205B (zh) 一种复杂路面背景下的裂缝提取方法
KR101549495B1 (ko) 문자 추출 장치 및 그 방법
CN105701491A (zh) 固定格式文档图像模版的制作方法及其应用
CN115100077B (zh) 一种图像增强方法与装置
CN111539293A (zh) 一种果树病害诊断方法及系统
CN111008967B (zh) 一种绝缘子rtv涂层缺陷识别方法
Arunachalam et al. Identification of defects in fruits using digital image processing
CN115601379A (zh) 一种基于数字图像处理的表面裂纹精确检测技术
CN115272362A (zh) 一种数字病理全场图像有效区域分割方法、装置
CN109682821B (zh) 一种基于多尺度高斯函数的柑橘表面缺陷检测方法
CN111027564A (zh) 基于深度学习一体化的低照度成像车牌识别方法及装置
CN110288616B (zh) 一种基于分形和rpca分割眼底图像中硬性渗出的方法
CN112509026A (zh) 一种绝缘子裂缝长度识别方法
CN110264463A (zh) 一种基于matlab图像处理的物料清点方法
Tabatabaei et al. A novel method for binarization of badly illuminated document images
CN112465817B (zh) 一种基于方向滤波器的路面裂缝检测方法
Khan et al. Shadow removal from digital images using multi-channel binarization and shadow matting
CN113516193B (zh) 基于图像处理的红枣缺陷识别分类方法及装置

Legal Events

Date Code Title Description
ENP Entry into the national phase

Ref document number: 2020802369

Country of ref document: EP

Effective date: 20201119

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

Ref document number: 20802369

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE