A remote intelligent inspection method and system for substations

Abstract

The present invention relates to the field of image processing technology, and the present invention relates to a remote intelligent patrol method and system for a substation, wherein the method comprises: obtaining edge pixel points of each frame image in a video for protecting substation equipment and its environment; for any frame image, obtaining the image of the previous frame of the frame image, and calculating the edge position stability of the frame image by the gray value difference and gradient amplitude difference between the image of the previous frame and all edge pixel points with the same coordinates in the frame image; for any frame image, calculating its high-frequency component gain; calculating the low-frequency component gain of the frame image; homomorphically filtering the video according to the high-frequency component gain and low-frequency component gain of each frame image in the video; identifying defects in the video, and alarming when there is a defect in at least one frame image. The present invention dynamically adjusts the high-frequency component gain and low-frequency component gain of the video frame image to improve the filtering effect of each frame image.

Description

Technical Field

[0001] The present invention generally relates to the field of image processing technology, and more specifically, to a remote intelligent inspection method and system for a substation. Background Art

[0002] With the rapid development of society, people's demand for electricity is increasing. The power sector should keep up with the development of the times, apply new intelligent technologies, innovate the operation and maintenance mode of substations, and reasonably apply AI recognition to the intelligent patrol system of substations (patrol, also known as inspection, generally uses drones for patrol monitoring), so as to achieve the stable development of substations, maintain the safe operation of substations, and provide people with sufficient power resources. In the remote intelligent patrol system of substations, image acquisition is an important means to obtain equipment status information. For example, Chinese patent application document CN114063643A discloses a remote monitoring control system and control method for a substation, and its specific scheme is: obtaining substation operation information; analyzing the substation operation information according to an analysis model to obtain estimated hot spots; generating a cruise path according to the estimated hot spots and sending the cruise path to an unmanned cruise device, the unmanned cruise device is used to cruise according to the cruise path, the unmanned cruise device is used to monitor the temperature in the substation in real time through infrared thermal imaging during the cruise, and the unmanned cruise device is also used to photograph a position where the temperature value is higher than the preset value to generate image information when the unmanned cruise device monitors that the temperature value is higher than the preset value; receiving image information and the temperature value corresponding to the image information collected by the unmanned cruise device, and sending the image information to the staff terminal.

[0003] However, due to factors such as lighting changes and weather conditions, the quality of the collected images is often unsatisfactory. Homomorphic filtering, as an effective image preprocessing method, can eliminate the impact of uneven illumination, enhance image details, and improve image quality. By applying homomorphic filtering technology, the substation remote intelligent patrol system can more accurately identify equipment status information and improve the accuracy of fault warning and automatic patrol. At the same time, it can also reduce the impact of adverse factors such as lighting changes on image quality and improve the stability and reliability of the system.

[0004] In the process of filtering the video frame image using homomorphic filtering, and Represents the gain of high-frequency components and low-frequency components respectively. High-frequency components usually correspond to the details and edges of the image, while low-frequency components are related to the illumination or overall brightness of the image. In general, you need to set >1 and <1 to enhance image details and edges and reduce the unevenness of lighting in the image to improve image contrast and details.

[0005] However, in the video frame images of the substation, there may be different light sources and shadow areas in the environment where the substation is located, resulting in different brightness in different video frames, and weather changes will also affect the lighting in the video frame images, which will lead to certain changes in the edge clarity of the collected video frame images. At this time, the fixed gains of high-frequency components and low-frequency components cannot respond to this change, resulting in poor filtering effects on some video frames, affecting the final inspection results. Summary of the invention

[0006] In order to solve the technical problem that when there are different light sources in the environment of the substation, the fixed gains of the high-frequency components and the low-frequency components cannot effectively filter each frame image in the substation video, the present invention provides solutions in the following aspects.

[0007] In a first aspect, a remote inspection method for a substation includes: obtaining a video of a substation device, determining edge pixel points of each frame image in the video; for any frame image, obtaining pixel points with the same coordinates as the pixel points in the frame image in the previous frame image of the frame image as its mapping pixel points, calculating the edge position stability of the frame image by the gray value difference and gradient amplitude difference between each edge pixel point in the frame image and its mapping pixel point, wherein the edge position stability represents the difference between the edge positions of multiple frames of images; for any frame image, calculating its high-frequency component gain γ H : ,in h i The first i The gray value corresponding to the edge pixel point is h i,k The first i The edge pixel k The grayscale values ​​corresponding to the adjacent pixels are M i The first i The number of neighboring pixels of an edge pixel, i is a positive integer, N is the number of edge pixels in the frame image, BW is the edge position stability of the frame image; calculate the low-frequency component gain of the frame image γ L : , ,in μ is the average grayscale value corresponding to all pixels of the frame image. σ is the standard deviation of the grayscale values ​​corresponding to all pixels of the frame image, exp() is an exponential function with the real number e as the base, His the information entropy of the frame image, and cv is the brightness variation coefficient; the video is homomorphically filtered according to the high-frequency component gain and the low-frequency component gain of each frame image in the video; and defects are identified in the video.

[0008] The beneficial effects of the present invention are:

[0009] The present invention dynamically adjusts the high-frequency component gain and the low-frequency component gain when filtering a video frame image according to the illumination uniformity and edge prominence in the video frame image, so as to improve the image filtering effect and further improve the accuracy of the patrol result.

[0010] Preferably, the formula for calculating the edge position stability BW of any frame image is:

[0011] ;

[0012] in, The first j The gray value corresponding to the edge pixel point is is the previous frame of the current frame. j The gray value corresponding to the edge pixel point is The first j The gradient amplitude corresponding to the edge pixel point is is the previous frame of the current frame. j The gradient amplitude corresponding to the edge pixel point is j Is a positive integer.

[0013] Preferably, the video of the substation equipment is processed by a Sobel operator to obtain edge pixel points of each frame image in the video.

[0014] The present invention processes the substation equipment and its environment video through the Sobel operator, which can not only effectively extract edge information, but also improve the efficiency and accuracy of the monitoring system, providing more reliable protection for equipment safety.

[0015] Preferably, the homomorphic filtering of the video according to the high-frequency component gain and the low-frequency component gain of each frame image in the video includes: designing a homomorphic filter according to the high-frequency component gain and the low-frequency component gain of each frame image; obtaining a spatial domain image by inverse Fourier transforming the frequency domain image obtained after filtering each frame image with the homomorphic filter, so as to realize homomorphic filtering of the image.

[0016] The present invention uses a homomorphic filter to perform homomorphic filtering on each frame image according to the high-frequency component gain and the low-frequency component gain of each frame image, thereby effectively removing noise in each frame image and improving the quality of each frame image.

[0017] Preferably, a remote intelligent inspection method for a substation further includes: for any pixel point, calculating the horizontal gradient G of the pixel point by using a Sobel operator x and the vertical gradient G y ; Calculate the gradient amplitude of the pixel: , where GM is the gradient amplitude of the pixel.

[0018] Preferably, defect recognition of a video includes: extracting features from each frame image in the video after homomorphic filtering processing by a SIFT algorithm; and inputting the features into a deep learning model to determine whether defects occur in each frame image.

[0019] The present invention can effectively extract features in images through the SIFT algorithm; and the present invention inputs the extracted features into a deep learning model so that the model can identify possible defects in each frame of the image, thereby reducing the dependence of defect identification on manual work and improving detection efficiency.

[0020] Preferably, a remote intelligent inspection method for a substation further includes: obtaining its information entropy includes: for any frame image, calculating the grayscale histogram of the frame image, and obtaining the frequency of occurrence of each grayscale level; for any grayscale level, dividing the frequency of occurrence of the grayscale level by the number of all pixels in the frame image to obtain the probability distribution of the grayscale level; calculating the information entropy of the frame image H : ,in, p a The first a The probability distribution of gray levels, A is the number of gray levels of the frame image, a Is a positive integer.

[0021] In a second aspect, a remote intelligent patrol system for a substation includes a processor and a memory, wherein the memory stores a computer program, and the processor executes the computer program to implement a remote intelligent patrol method for a substation according to any one of the above-mentioned invention contents.

[0022] The beneficial effects of the present invention are:

[0023] The present invention significantly improves the image filtering effect by dynamically adjusting the high-frequency and low-frequency component gains of the video frame image, thereby improving the accuracy of the inspection results. Based on this, the present invention avoids the technical problem that the fixed high-frequency and low-frequency component gains cannot effectively filter each frame image in the substation video. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] By reading the following detailed description with reference to the accompanying drawings, the above and other objects, features and advantages of the exemplary embodiments of the present invention will become readily understood. In the accompanying drawings, several embodiments of the present invention are shown in an exemplary and non-limiting manner, and the same or corresponding reference numerals represent the same or corresponding parts, wherein:

[0025] Figure 1 is a flowchart schematically showing a method of remote intelligent inspection of a substation according to an embodiment of the present invention;

[0026] Figure 2 is a schematic diagram of a pixel point and its adjacent pixel points in any frame image of a video according to an embodiment of the present invention;

[0027] Figure 3 The figure schematically shows a structural block diagram of a remote intelligent inspection system for a substation according to an embodiment of the present invention. DETAILED DESCRIPTION

[0028] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present invention.

[0029] The specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0030] Figure 1 The figure is a flowchart schematically showing the steps of a remote intelligent inspection method for a substation according to an embodiment of the present invention.

[0031] like Figure 1 As shown, a remote intelligent inspection method for a substation includes steps S1 to S6.

[0032] Step S1: obtaining a video of a substation device, and determining edge pixel points of each frame image in the video.

[0033] In one embodiment, the video of the protection substation equipment and its environment is processed by a Sobel operator to obtain edge pixel points of each frame image in the video.

[0034] Step S2: For any frame image, in the previous frame image of the frame image, obtain the pixel point with the same coordinates as the pixel point in the frame image as its mapping pixel point, and calculate the edge position stability of the frame image through the grayscale value difference and gradient amplitude difference between each edge pixel point in the frame image and its mapping pixel point.

[0035] The edge position stability represents the difference between edge positions of multiple frames of images. The formula for calculating the edge position stability BW of any frame of image is:

[0036] .

[0037] The first j The gray value corresponding to the edge pixel point is is the image in the previous frame of the image and the image in the previous frame of the image. j The grayscale value corresponding to the pixel with the same coordinates as the edge pixel points. The first j The gradient amplitude corresponding to the edge pixel point is is the image in the previous frame of the image and the image in the previous frame of the image. j The gradient amplitude corresponding to the pixel points with the same coordinates as the edge pixel points, j Is a positive integer.

[0038] It should be noted that for any pixel in any frame of the video, the gradient amplitude of the pixel reflects the intensity of the brightness change around the pixel. The larger the gradient amplitude of the pixel, the more drastic the change. Therefore, for any frame of the video, the absolute value of the square difference between the grayscale value difference and the gradient amplitude difference of the edge pixel in the frame and the pixel at the same position in the previous frame of the video is calculated. The larger the value is, the more stable the edge position of the frame image is. The larger the value, the more stable the edge position of the frame image is. BW The bigger.

[0039] When abnormal events occur in the video (such as object movement, shape change, color mutation, etc.), these changes are often reflected on the edges of the image. For example, a stationary object suddenly moves, which causes the position of its edge to change; a change in the shape of an object can cause the edge shape to distort or break. In other words, by monitoring the changes in the edges, abnormal events in the video can be discovered in time and the corresponding alarm or processing mechanism can be triggered.

[0040] In one embodiment, calculating the gradient magnitude of any pixel point of any frame image includes: for any pixel point, calculating the horizontal gradient G of the pixel point by using the Sobel operator. x and the vertical gradient G y ; Calculate the gradient amplitude of the pixel: , where GM is the gradient amplitude of the pixel.

[0041] Step S3: For any frame image, calculate its high-frequency component gain γH .

[0042] Among them, the formula for calculating the high-frequency component gain of the image is: , h i The first i The gray value corresponding to the edge pixel point is h i,k The first i The edge pixel k The grayscale values ​​corresponding to the adjacent pixels are M i The first i The number of neighboring pixels of an edge pixel, i is a positive integer, N is the number of edge pixels in the frame image, BW is the edge position stability of the frame image.

[0043] It should be noted that for any pixel point, in a 3×3 window centered on the pixel point, all pixels other than the pixel point are the neighboring pixels of the pixel point. Furthermore, for any frame image, is the contrast of the i-th edge pixel of the frame image; is the edge conspicuity of the frame image. The larger it is, the greater the edge visibility of the frame image is. The more obvious the edge of the frame image is, the more credible the visibility of the edge of the frame image is, and the larger the high-frequency component gain parameter of the frame image is.

[0044] Figure 2 It is a schematic diagram of pixel points and adjacent pixel points in any frame image of a video according to an embodiment of the present invention.

[0045] like Figure 2 As shown, pixel point D1 and pixel point D2 are pixel points in the frame image.

[0046] The 3×3 window centered on pixel D1 covers 9 pixels including pixel D1 in the frame image, and the 8 pixels around pixel D1 are its neighboring pixels. That is, pixel d1, pixel d2, pixel d3, pixel d4, pixel d5, pixel d6, pixel d7 and pixel d8 are neighboring pixels of pixel D1.

[0047] Similarly, the 3×3 window centered on pixel D2 covers 4 pixels including pixel D2 in the frame image, and the 3 pixels around pixel D2 are its neighboring pixels, that is, pixel d9, pixel d10, and pixel d11 are neighboring pixels of pixel D2.

[0048] Step S4: For any frame image, calculate its low-frequency component gain γ L .

[0049] Among them, calculation , , μ is the average grayscale value corresponding to all pixels of the frame image. σ is the standard deviation of the grayscale values ​​corresponding to all pixels of the frame image, exp() is an exponential function with the real number e as the base, H is the information entropy of the frame image, and cv is the brightness variation coefficient.

[0050] Furthermore, obtaining the information entropy of the frame image includes: for any frame image, calculating the grayscale histogram of the frame image and obtaining the frequency of occurrence of each grayscale level; for any grayscale level, dividing the frequency of occurrence of the grayscale level by the number of all pixels in the frame image to obtain the probability distribution of the grayscale level; calculating the information entropy of the frame image H : ,in, p a The first a The probability distribution of gray levels, A is the number of gray levels of the frame image, a Is a positive integer.

[0051] It should be noted that the ratio of the mean value to the variance of the grayscale values ​​of all pixels in the frame image is The larger the value, the smaller the credibility of the illumination uniformity of the frame image. cv The smaller.

[0052] Furthermore, the absolute value of the ratio of the information entropy of the frame image to the brightness variation coefficient is As the correlation between the two, the correlation is used to describe the credibility of the illumination uniformity of the frame image. The larger the value is, the higher the correlation between the two is, the more reliable the illumination uniformity of the frame image is, and the smaller the gain of the low-frequency component of the frame image is.

[0053] Step S5: performing homomorphic filtering on the video according to the high-frequency component gain and the low-frequency component gain of each frame image in the video.

[0054] In one embodiment, performing homomorphic filtering on the video according to the high-frequency component gain and the low-frequency component gain of each frame image in the video includes: designing a homomorphic filter according to the high-frequency component gain and the low-frequency component gain of each frame image; obtaining a spatial domain image by inverse Fourier transforming the frequency domain image obtained after filtering each frame image with the homomorphic filter, so as to achieve homomorphic filtering of the image.

[0055] Among them, homomorphic filtering is used to improve the brightness and contrast of the image. The homomorphic filter is used to perform homomorphic filtering on each frame image to enhance the contrast and details of each frame image, so that the subsequent defect recognition is more accurate.

[0056] Step S6: Identify defects in the video.

[0057] In one embodiment, defect recognition of a video includes: extracting features from each frame image in the video after homomorphic filtering through a SIFT algorithm; and inputting the features into a deep learning model to determine whether defects occur in each frame image.

[0058] Among them, SIFT (Scale Invariant Feature Transform) is an algorithm for detecting and describing local features, which can maintain stability under different scaling and rotation conditions. The extracted features can be used to describe the key points and local areas of the image. The extracted features are input into the deep learning model for analysis. The model can be a trained convolutional neural network (CNN) that can identify specific defect patterns. By comparing the extracted features with the features learned by the model, the model can determine whether there is a defect in the current frame.

[0059] Furthermore, when multiple consecutive frames of images in the video are judged to have defects, an alarm is issued, so that the substation maintenance personnel can perform maintenance and inspection on the substation.

[0060] Figure 3 The figure schematically shows a structural block diagram of a remote intelligent inspection system for a substation according to an embodiment of the present invention.

[0061] The present invention also provides a remote intelligent inspection system for substations. Figure 3 As shown, the system includes a processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, a remote intelligent inspection method for a substation according to the first aspect of the present invention is implemented.

[0062] In the present invention, the aforementioned memory may be any tangible medium containing or storing a program that can be used by or in combination with an instruction execution system, apparatus or device. For example, a computer-readable storage medium may be any appropriate magnetic storage medium or magneto-optical storage medium, such as a resistive random access memory RRAM (Resistive Random Access Memory), a dynamic random access memory DRAM (Dynamic Random Access Memory), a static random access memory SRAM (Static Random-Access Memory), an enhanced dynamic random access memory EDRAM (Enhanced Dynamic Random Access Memory), a high-bandwidth memory HBM (High-Bandwidth Memory), a hybrid memory cube HMC (Hybrid Memory Cube), etc., or any other medium that can be used to store the required information and can be accessed by an application, a module, or both. Any such computer storage medium may be part of a device or accessible or connectable to a device. Any application or module described in the present invention may be implemented using computer-readable / executable instructions that may be stored or otherwise maintained by such a computer-readable medium.

[0063] In the description of this specification, "plurality" or "several" means at least two, such as two, three or more, etc., unless otherwise clearly defined.

[0064] Although this specification has shown and described a number of embodiments of the present invention, it will be apparent to those skilled in the art that such embodiments are provided by way of example only. Those skilled in the art will conceive of many modifications, changes and alternatives without departing from the ideas and spirit of the present invention. It should be understood that in the practice of the present invention, various alternatives to the embodiments of the present invention described herein may be employed.

Claims

1. A remote intelligent inspection method for a substation, characterized in that: include: Obtain a video of the substation equipment, and determine edge pixel points of each frame image in the video; For any frame image, in the previous frame image of the frame image, a pixel point with the same coordinates as a pixel point in the frame image is obtained as its mapping pixel point, and the edge position stability of the frame image is calculated by the gray value difference and gradient amplitude difference between each edge pixel point in the frame image and its mapping pixel point, and the edge position stability represents the difference between the edge positions of multiple frame images; For any frame image, calculate its high-frequency component gain γ H : ,in h i The first i The gray value corresponding to the edge pixel point is h i,k The first i The edge pixel k The grayscale values ​​corresponding to the adjacent pixels are M i The first i The number of neighboring pixels of an edge pixel, i is a positive integer, N is the number of edge pixels in the frame image, BW is the edge position stability of the frame image; Calculate the low-frequency component gain of the frame image γ L : , ,in μ is the average grayscale value corresponding to all pixels of the frame image. σ is the standard deviation of the grayscale values ​​corresponding to all pixels of the frame image, exp() is an exponential function with the real number e as the base, H is the information entropy of the frame image, and cv is the brightness variation coefficient; The video is homomorphically filtered according to the high-frequency component gain and the low-frequency component gain of each frame image in the video; and defects are identified in the video.

2. A remote intelligent inspection method for substations according to claim 1, characterized in that: The formula for calculating the edge position stability BW of any frame image is: ； in, The first j The gray value corresponding to the edge pixel point is is the previous frame of the current frame. j The gray value corresponding to the edge pixel point is The first j The gradient amplitude corresponding to the edge pixel point is is the previous frame of the current frame. j The gradient amplitude corresponding to the edge pixel point is j Is a positive integer.

3. A remote intelligent inspection method for substations according to claim 1, characterized in that: The video of the substation equipment is processed by a Sobel operator to obtain edge pixel points of each frame image in the video.

4. A remote intelligent inspection method for substations according to claim 1, characterized in that: The performing homomorphic filtering on the video according to the high-frequency component gain and the low-frequency component gain of each frame image in the video comprises: A homomorphic filter is designed according to the high-frequency component gain and the low-frequency component gain of each frame image; The frequency domain image obtained after filtering each frame image with the homomorphic filter is inversely Fourier transformed to obtain the spatial domain image, so as to realize homomorphic filtering of the image.

5. A remote intelligent inspection method for substations according to claim 2, characterized in that: Also includes: For any pixel point, the horizontal gradient G of the pixel point is calculated by the Sobel operator. x and the vertical gradient G y ; Calculate the gradient magnitude of the pixel: , where GM is the gradient amplitude of the pixel.

6. A remote intelligent inspection method for substations according to claim 1, characterized in that: Defect recognition of videos includes: Extracting features from each frame image in the video after homomorphic filtering by using a SIFT algorithm; The features are input into a deep learning model to determine whether defects occur in each frame image.

7. A remote intelligent inspection method for substations according to claim 1, characterized in that: Also includes: Obtaining its information entropy includes: For any frame image, calculate the grayscale histogram of the frame image and obtain the frequency of occurrence of each grayscale level; For any gray level, the frequency of occurrence of the gray level is divided by the number of all pixels in the frame image to obtain the probability distribution of the gray level; Calculate the information entropy of the frame image H : ,in, p a The first a The probability distribution of gray levels, A is the number of gray levels of the frame image, a Is a positive integer.

8. A remote intelligent inspection system for a substation, comprising a processor and a memory, wherein the memory stores a computer program, characterized in that: The processor executes the computer program to implement a remote intelligent inspection method for a substation as described in any one of claims 1 to 7.

Images