A real-time monitoring method and system for substation condensation based on image recognition

Abstract

The present invention discloses a real-time monitoring method and system for condensation in a substation based on image recognition, which relates to the field of image recognition technology, including an image acquisition unit, a wireless communication unit, a cloud computing unit, etc. Through perspective transformation based on a tangent function, the condensation distribution image on the condensation condensation surface is quickly obtained, and the response speed is fast; the nonlinear stretching based on a cosine function is adopted for the normalized image, which can effectively improve the contrast between the pixels of the condensation boundary contour and the pixels of the rest of the image, and helps the morphological operation to more accurately find the condensation contour in the image; by generating binary image data of the condensation contour, the total number of condensed dewdrops can be quickly estimated, taking into account both real-time and economic performance; the camera is arranged obliquely, which, on the one hand, has a wider shooting field of view and a large coverage range, and on the other hand, reduces the risk of condensation on the camera lens, and has a strong anti-interference ability.

Description

Technical Field

[0001] The present invention relates to the technical field of image recognition, and in particular to a method and system for real-time monitoring of condensation in a substation based on image recognition. Background Art

[0002] Substations are the core hubs of the power system, and the operating environment of their equipment directly affects the reliability of power supply. Condensation is a common environmental hazard in substations, mainly due to sudden changes in humidity or temperature differences that cause water vapor to condense on the surface of the equipment. Long-term condensation will accelerate the corrosion of metal parts, cause the degradation of insulation performance, and even lead to serious accidents such as short circuits and discharges. The risk of condensation is particularly prominent in local spaces of equipment such as closed cabinets and terminal boxes. Traditional manual inspection methods rely on periodic inspections, which make it difficult to detect sudden condensation in a timely manner. There is an urgent need to develop real-time, automated monitoring methods.

[0003] The current mainstream condensation monitoring methods mainly include: (1) Temperature and humidity sensor method: by arranging sensors to collect temperature and humidity data, and combining the dew point model to estimate the condensation risk. However, the sensors need to be densely arranged and are susceptible to electromagnetic interference, and cannot directly observe the condensation state, resulting in a high false alarm rate; (2) Infrared thermal imaging method: using the difference in temperature distribution on the surface of the equipment to identify the condensation area, but it is sensitive to ambient temperature and humidity, the imaging quality is easily affected by water mist, and the equipment cost is high; (3) Fiber optic sensing method: based on the light intensity attenuation characteristics of the optical fiber after being damp, it requires customized modification of the equipment surface, complex deployment and high maintenance costs.

[0004] The above methods generally have technical problems such as limited coverage, insufficient anti-interference ability, and difficulty in balancing real-time and economy, making it difficult to meet the needs of accurate monitoring under complex working conditions. Summary of the invention

[0005] The purpose of the present invention is to provide a substation condensation real-time monitoring method and system based on image recognition, which has fast response speed, large coverage, strong anti-interference ability, and takes into account both real-time and economic performance.

[0006] In view of the above technical problems, the technical solution adopted by the present invention is: a real-time monitoring method for substation condensation based on image recognition, comprising the following steps:

[0007] Step S1: Number the monitored substation terminal, establish a unique corresponding cloud database, collect images with a pixel of 1280×960 every 0.5 to 2 hours for the currently monitored substation terminal through the image acquisition unit, upload the real-time collected images to the cloud computing unit through the wireless communication unit, and perform online image analysis and processing through the cloud computing unit;

[0008] Step S2: Performing a perspective transformation based on a tangent function on each image received by the cloud computing unit, converting the image after the perspective transformation into a grayscale image, and normalizing the obtained grayscale image pixel values ​​so that the normalized image pixel values ​​are within the interval [0, 1];

[0009] Step S3: performing nonlinear stretching based on a cosine function on the normalized image, and performing denormalization on the pixel values ​​of the image after the nonlinear stretching, so that the pixel values ​​of the denormalized image are within the interval [0, 255];

[0010] Step S4: performing Gaussian filtering denoising on the image after the denormalization processing, and performing adaptive histogram equalization on the image after the Gaussian filtering denoising processing;

[0011] Step S5: performing adaptive threshold segmentation on the image processed by adaptive histogram equalization, and then finding the condensation contour in the image through the closing operation and opening operation of the morphological operation, setting the pixel value of the condensation contour line to 1, and setting the pixel values ​​of the rest of the contour line to 0, so as to generate binary image data of the condensation contour;

[0012] Step S6: Sum all element values ​​of the matrix corresponding to the binary image data, estimate the total number of condensed dew drops based on the summation result, and determine whether the warning value is reached based on the total number of dew drops. When the total number of dew drops is greater than the warning value, manual intervention is required to eliminate the condensation in the substation terminal currently being monitored.

[0013] Preferably, in step S2, is the image before perspective transformation. k The pixel matrix of the page j Column, No. i The pixels of the row, is the image after perspective transformation k The pixel matrix of the page j Column, No. i For pixels in a row, the following perspective transformation relationship exists:

[0014] ,

[0015] Where H is a 3×3 perspective transformation matrix, which is:

[0016] ,

[0017] The position relationship of the four corner points of the image before the perspective transformation is known:

[0018] ,

[0019] Assume that the shooting inclination angle of image acquisition is , then the position relationship of the four corner points of the image after perspective transformation based on the tangent function is:

[0020] ,

[0021] when k When any value among 1, 2 or 3 is taken, the above two positional relationships are substituted into the perspective transformation relationship to obtain a linear equation system. By solving the linear equation system, all parameter values ​​in the perspective transformation matrix H are obtained.

[0022] Preferably, all the The value corresponds to the first pixel in the image matrix after perspective transformation. i Row, No. j Column, No. k The pixel value of the pixel position of the page is equal to the coordinate of the corresponding image before the perspective transformation. The pixel value at the pixel position.

[0023] Preferably, when When any of is not an integer, the corresponding pixel value is calculated by the bilinear interpolation algorithm.

[0024] Preferably, in step S2, the calculation formula for normalization processing is:

[0025] ,

[0026] In the formula, is the grayscale image before normalization. i Row, No. j The pixel value of the column pixel; After normalization, the grayscale image i Row, No. j The pixel value of the column pixel.

[0027] Preferably, in step S3, the nonlinear stretching formula based on the cosine function is:

[0028] ,

[0029] In the formula, s is the tensile strength coefficient, and its value range is [0, 0.5]; P is the boundary pixel value, and its value range is [80, 120]; is the stretch range control coefficient, and its value range is [0, 2]; is the pixel value of the image after nonlinear stretching. In particular, if The value of is greater than 1, The value of is 1, if If the value is less than 0, The value of is 0.

[0030] Preferably, in step S3, the calculation formula for the denormalization process is:

[0031] , where After the denormalization process, the grayscale image i Row, No. j The pixel value of the column pixel.

[0032] Preferably, in step S6, the calculation formula for the total number of dew drops is: , where N is the total number of dewdrops; B is the matrix corresponding to the binary image data; is the coincidence coefficient, and its value range is [0.4, 0.6]; L is the average perimeter of a single dew drop contour, and its value range is [35, 60]. The value range of the total number of dew drops warning value is [2000, 2400].

[0033] The present invention also proposes a real-time monitoring system for condensation in a substation based on image recognition based on the above method, including an image acquisition unit, a wireless communication unit, and a cloud computing unit. Data is exchanged between the image acquisition unit and the wireless communication unit; data is exchanged between the wireless communication unit and the cloud computing unit; the image acquisition unit is used to acquire condensation images; the wireless communication unit is used to transmit condensation images and interact with image acquisition instructions; the cloud computing unit is used to identify and calculate condensation images, and to indicate whether manual intervention for cleaning is required based on the results.

[0034] Preferably, the image acquisition unit includes a camera and a controller; the camera and the controller are electrically connected; data is exchanged between the controller and the wireless communication unit; the controller is used to control the camera to acquire condensation images and transmit image data to the wireless communication unit; the plane where all condensation is condensed is the condensation surface, and the angle between the plane where the mirror surface of the camera lens is located and the condensation surface is the shooting inclination angle , The value range is 60° to 75°.

[0035] Compared with the prior art, the present invention has the following advantages: (1) by using perspective transformation based on tangent function, the condensation distribution image on the condensation surface can be quickly obtained, and the response speed is fast; (2) by using nonlinear stretching based on cosine function for the normalized image, the contrast between the pixels of the condensation boundary contour and the pixels of the rest of the image can be effectively improved, which helps the morphological operation to more accurately find the condensation contour in the image; (3) by generating binary image data of the condensation contour, the total number of condensed dew drops can be quickly estimated, taking into account both real-time performance and economy; (4) the camera is arranged at an angle, which, on the one hand, has a wider shooting field of view and a large coverage area, and on the other hand, reduces the risk of condensation on the camera lens and has a strong anti-interference ability. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Figure 1 The present invention is a flow chart of the method for real-time monitoring of condensation in a substation.

[0037] Figure 2 It is a module diagram of the substation condensation real-time monitoring system of the present invention.

[0038] Figure 3 This is a schematic diagram of the layout of the image acquisition unit of the present invention.

[0039] Figure 4 This is a reverse color image of the condensation contour recognition result according to an embodiment of the present invention.

[0040] In the figure: 1-image acquisition unit; 2-wireless communication unit; 3-cloud computing unit; 4-condensation surface; 5-condensation dew; 101-camera; 102-controller. DETAILED DESCRIPTION

[0041] The technical solution of the present invention will be further described below in conjunction with the accompanying drawings and through specific implementation methods, wherein the accompanying drawings are only used for exemplary descriptions and represent only schematic diagrams rather than actual drawings, and should not be understood as limiting the present invention; in order to better illustrate the embodiments of the present invention, some parts of the accompanying drawings may be omitted, enlarged or reduced, and do not represent the size of the actual product; for those skilled in the art, it is understandable that some well-known structures and their descriptions in the accompanying drawings may be omitted.

[0042] Figures 1 to 4 It is a preferred embodiment of the present invention.

[0043] like Figure 1 As shown, a real-time monitoring method for substation condensation based on image recognition is given, which includes the following steps:

[0044] Step S1: Number the monitored substation terminal, establish a unique corresponding cloud database, collect a 1280×960 pixel image of the currently monitored substation terminal every 0.5 hours through the image acquisition unit 1, upload the real-time collected image to the cloud computing unit 3 through the wireless communication unit 2, and perform online image analysis and processing through the cloud computing unit 3;

[0045] Step S2: Perform a perspective transformation based on the tangent function on each image received by the cloud computing unit 3. is the image before perspective transformation. k The pixel matrix of the page j Column, No. i The pixels of the row are set The corresponding image after perspective transformation k The pixel matrix of the page j Column, No. i For pixels in a row, the following perspective transformation relationship exists: ,

[0046] Where H is a 3×3 perspective transformation matrix, which is: ,

[0047] The position relationship of the four corner points of the image before the perspective transformation is known: ,

[0048] Assume that the shooting inclination angle of image acquisition is , then the position relationship of the four corner points of the image after perspective transformation based on the tangent function is: ,when k When any value among 1, 2 or 3 is taken, the above two positional relationships are substituted into the perspective transformation relationship to obtain a linear equation system. By solving the linear equation system, all parameter values ​​of the perspective transformation matrix H are obtained; all value; the corresponding image pixel matrix after perspective transformation i Row, No. j Column, No. k The pixel value of the pixel position of the page is equal to the coordinate of the corresponding image before the perspective transformation. The pixel value of the pixel position; when When any of is not an integer, the corresponding pixel value is calculated by the bilinear interpolation algorithm; the image after perspective transformation is converted into a grayscale image, and the obtained grayscale image pixel value is normalized so that the normalized image pixel value is in the interval [0, 1]; the calculation formula for normalization is:

[0049] ,

[0050] In the formula, is the grayscale image before normalization. i Row, No. j The pixel value of the column pixel; After normalization, the grayscale image i Row, No. j The pixel value of the column pixel;

[0051] Step S3: Apply nonlinear stretching based on the cosine function to the normalized image. The nonlinear stretching formula based on the cosine function is:

[0052] ,

[0053] In the formula, s is the tensile strength coefficient, and its value range is [0, 0.5]; P is the boundary pixel value, and its value range is [80, 120]; is the stretch range control coefficient, and its value range is [0, 2]; is the pixel value of the image after nonlinear stretching. In particular, if The value of is greater than 1, The value of is 1, if If the value is less than 0, The value of is 0; the pixel values ​​of the image after nonlinear stretching are denormalized so that the pixel values ​​of the denormalized image are within the interval [0, 255]; the calculation formula for denormalization is:

[0054] , where After the denormalization process, the grayscale image i Row, No. j The pixel value of the column pixel;

[0055] Step S4: performing Gaussian filtering denoising on the image after the denormalization process. In this embodiment, the Gaussian filtering denoising is implemented by the imgaussfilt function of MATLAB, and the standard deviation value of the Gaussian filtering denoising is selected as 2; performing adaptive histogram equalization on the image after the Gaussian filtering denoising process. In this embodiment, the adaptive histogram equalization is implemented by the adaptathisteq function of MATLAB;

[0056] Step S5: Adaptive threshold segmentation is performed on the image after adaptive histogram equalization processing. In this embodiment, the adaptive threshold segmentation is implemented by the imbinarize function of MATLAB. Then, the condensation contour in the image is found out by the closing operation and opening operation of morphological operations. In MATLAB, the opening operation and closing operation can be conveniently implemented by using the imopen function and the imclose function, and the contour is extracted by the bwboundaries function. The pixel value of the contour line of the condensation is set to 1, and the pixel values ​​of the rest of the contour lines are set to 0, so as to generate binary image data of the condensation contour.

[0057] Step S6: sum all element values ​​of the matrix corresponding to the binary image data, and estimate the total number of condensed dew drops according to the summation result. The calculation formula for the total number of dew drops is: , where N is the total number of dewdrops; B is the matrix corresponding to the binary image data; is the overlap coefficient, and its value range is [0.4, 0.6]. In this embodiment, The value is 0.5; L is the average perimeter of a single dewdrop contour, ranging from [35, 60]. In this embodiment, L The value is 49.5; the value range of the total number of dew drops warning value is [2000, 2400]; according to the total number of dew drops, it is judged whether the warning value is reached. When the total number of dew drops is greater than the warning value, manual intervention is required to eliminate the condensation in the substation terminal currently being monitored; in this embodiment, the warning value is 2400.

[0058] like Figure 2 As shown, a system of a substation condensation real-time monitoring method based on image recognition includes an image acquisition unit 1, a wireless communication unit 2, and a cloud computing unit 3; data is exchanged between the image acquisition unit 1 and the wireless communication unit 2; data is exchanged between the wireless communication unit 2 and the cloud computing unit 3; the image acquisition unit 1 is used to acquire condensation images; the wireless communication unit 2 is used for the transmission of condensation images and the interaction of image acquisition instructions; the cloud computing unit 3 is used for the recognition and calculation of condensation images.

[0059] like Figure 3 As shown, the image acquisition unit 1 includes a camera 101 and a controller 102; the camera 101 and the controller 102 are electrically connected; the controller 102 and the wireless communication unit 2 perform data exchange; the controller 102 is used to control the camera 101 to collect condensation images and transmit the image data to the wireless communication unit 2; all condensation ( Figure 3 Condensation 5) The condensation plane is the condensation plane 4, and the angle between the plane where the mirror surface of the camera 101 lens is located and the condensation plane 4 is the shooting inclination angle , The value range of is 60° to 75°. In this embodiment, The value of is 67.5°.

[0060] like Figure 4 As shown, a reverse color image of the condensation contour recognition result derived by the monitoring method and system according to the above embodiment is given. The reverse color means that the pixel value of the condensation contour line is set to 0, and the remaining pixel values ​​​​except the contour line are set to 1, so as to display an image with black lines and a white background. In this case, the total number of dew drops calculated is 2736, which is greater than the warning value of 2400, and the cloud computing unit 3 needs to alarm and prompt manual cleaning.

Claims

1. A real-time monitoring method for substation condensation based on image recognition, characterized in that: The following steps are involved: Step S1: number the substation terminal to be monitored, establish a unique corresponding cloud database, collect an image with a pixel of 1280×960 for the substation terminal currently being monitored once every 0.5 hour to 2 hours through the image acquisition unit (1), upload the real-time collected image to the cloud computing unit (3) through the wireless communication unit (2), and perform online image analysis and processing through the cloud computing unit (3); Step S2: Perform a perspective transformation based on the tangent function on each image received by the cloud computing unit (3). is the image before perspective transformation. k The pixel matrix of the page j Column, No. i The pixels of the row, is the image after perspective transformation k The pixel matrix of the page j Column, No. i For pixels in a row, the following perspective transformation relationship exists: ， Where H is a 3×3 perspective transformation matrix, which is: ， The position relationship of the four corner points of the image before the perspective transformation is known: ， Assume that the shooting inclination angle of image acquisition is , then the position relationship of the four corner points of the image after perspective transformation based on the tangent function is: ， when k When any value among 1, 2 or 3 is taken, the above two positional relationships are substituted into the perspective transformation relationship to obtain a linear equation system. By solving the linear equation system, all parameter values ​​in the perspective transformation matrix H are obtained; the image after perspective transformation is converted into a grayscale image, and the obtained grayscale image pixel values ​​are normalized so that the normalized image pixel values ​​are in the interval [0,1]; the calculation formula for normalization is: ， In the formula, is the grayscale image before normalization. i Row, No. j The pixel value of the column pixel; After normalization, the grayscale image i Row, No. j The pixel value of the column pixel; Step S3: Apply nonlinear stretching based on the cosine function to the normalized image. The nonlinear stretching formula based on the cosine function is: ， In the formula, s is the tensile strength coefficient, and its value range is [0,0.5]; P is the boundary pixel value, and its value range is [80,120]; is the stretch range control coefficient, and its value range is [0,2]; is the pixel value of the image after nonlinear stretching, if The value of is greater than 1, The value of is 1, if If the value is less than 0, The value of is 0; the pixel values ​​of the image after nonlinear stretching are denormalized so that the pixel values ​​of the denormalized image are within the interval [0,255]; Step S4: performing Gaussian filtering denoising on the image after the denormalization processing, and performing adaptive histogram equalization on the image after the Gaussian filtering denoising processing; Step S5: performing adaptive threshold segmentation on the image processed by adaptive histogram equalization, and then finding the condensation contour in the image through the closing operation and opening operation of the morphological operation, setting the pixel value of the condensation contour line to 1, and setting the pixel values ​​of the rest of the contour line to 0, so as to generate binary image data of the condensation contour; Step S6: Sum all element values ​​of the matrix corresponding to the binary image data, estimate the total number of condensed dew drops based on the summation result, and determine whether the warning value is reached based on the total number of dew drops. When the total number of dew drops is greater than the warning value, manual intervention is required to eliminate the condensation in the substation terminal currently being monitored.

2. A method for real-time monitoring of condensation in a substation based on image recognition as claimed in claim 1, characterized in that: All the perspective transformation relationships are calculated The value corresponds to the first pixel in the image matrix after perspective transformation. i Row, No. j Column, No. k The pixel value of the pixel position of the page is equal to the coordinate of the corresponding image before the perspective transformation. The pixel value at the pixel position.

3. A method for real-time monitoring of condensation in a substation based on image recognition as claimed in claim 2, characterized in that: when When any of is not an integer, the corresponding pixel value is calculated by the bilinear interpolation algorithm.

4. A method for real-time monitoring of condensation in a substation based on image recognition as claimed in claim 3, characterized in that: In step S3, the calculation formula for the denormalization process is: ， In the formula, After the denormalization process, the grayscale image i Row, No. j The pixel value of the column pixel.

5. A method for real-time monitoring of condensation in a substation based on image recognition as claimed in claim 4, characterized in that: In step S6, the calculation formula for the total number of dew drops is: ， In the formula, N is the total number of dewdrops; B is the matrix corresponding to the binary image data; is the overlap coefficient, and its value range is [0.4,0.6]; L is the average perimeter of a single dew drop contour, and its value range is [35,60]. The value range of the total number of dew drops warning value is [2000,2400].

6. A system based on the substation condensation real-time monitoring method based on image recognition according to claim 5, comprising an image acquisition unit (1), a wireless communication unit (2), and a cloud computing unit (3), characterized in that: Data is exchanged between the image acquisition unit (1) and the wireless communication unit (2); data is exchanged between the wireless communication unit (2) and the cloud computing unit (3); the image acquisition unit (1) is used to acquire condensation images; the wireless communication unit (2) is used to transmit condensation images and exchange image acquisition instructions; the cloud computing unit (3) is used to identify and calculate condensation images, and indicates whether manual intervention is required for cleaning according to the results.

7. The system of the substation condensation real-time monitoring method based on image recognition according to claim 6 is characterized in that: The image acquisition unit (1) comprises a camera (101) and a controller (102); the camera (101) and the controller (102) are electrically connected; the controller (102) and the wireless communication unit (2) exchange data; the controller (102) is used to control the camera (101) to acquire a condensation image and transmit the image data to the wireless communication unit (2); the plane on which all condensation is condensed is a condensation surface (4), and the angle between the plane where the mirror surface of the camera (101) lens is located and the condensation surface (4) is a shooting inclination angle, and the value range of the angle is 60° to 75°.

Images