A Substation Switching Board Image Recognition Method and System

Abstract

The present invention provides a method for image recognition of a substation switching board. Based on the original switching board image, a rectangular switching recognition area is set. For the switching monitoring set image, HSV color recognition is performed. Based on the endpoint direct calculation method and the extended side line intersection method, endpoints of the binary image of the switching board are searched in parallel to correct the original switching board image. The position of the switching dividing line is found, and by HSV color recognition, arrows in the "open" part image and the "closed" part image are respectively detected to generate arrow binary images to determine the switching state. The color extraction method with adaptive brightness in this application can extract image color information in environments with different brightness levels. The algorithm calculation time is short, emphasizing real-time performance, and is more suitable for real-time monitoring of substations.

Description

Technical Field

[0001] The present invention relates to the technology of power systems, and particularly to a technology for identifying the opening and closing states of switchboards in substations. Background Art

[0002] Real-time monitoring and identification of opening and closing state indicators through surveillance video images is an auxiliary method. This identification method runs parallel to methods such as automated signal monitoring and manual monitoring, which can improve the real-time performance, visualization, and reliability of the entire monitoring system.

[0003] Existing technologies can be divided into traditional image algorithms and deep learning methods. Among them, traditional algorithms are difficult to adapt to complex environments, lack universality, and have low recognition accuracy in different environments. Deep learning methods have high recognition accuracy, but at the same time require a large number of training samples for training the model. In addition, detection objects outdoors often change due to environmental changes, which requires the design of targeted algorithms. In the substation environment, many recognition objects such as voltage, current, oil volume, opening and closing states or values are relatively stable, lacking abnormal samples and rare samples, resulting in insufficient training samples, which ultimately affects the recognition accuracy.

[0004] For example, the prior art CN 108280454 A is a sensitive image recognition method based on the combination of HSV and LBP features, and the prior art CN 108133219 A is a sensitive image recognition based on the combination of HSV, SURF, and LBP features. Although both of these prior arts use HSV for image recognition, they combine hsv recognition with algorithms such as LBP features and neural networks, which can distinguish the human skin part from the background environment, but the technical algorithm complexity is high and the calculation time is longer. Because the operation safety of substations is very important, faster algorithms are more needed. In addition, existing algorithms require the training process of neural networks and need to be trained with a large amount of picture information. In the specific application of the substation environment, the pointing or display of various instruments and indicators is relatively single, and the difference between pictures is very small. Therefore, there is no rich source of training pictures, and the effect of using neural networks in this case is not good. Summary of the Invention

[0005] The object of the present invention is to address the problems of the prior art. This patent proposes an image recognition method and system for switchboards in substations, and a color extraction method with adaptive brightness, which can extract image color information in environments with different brightness levels, has a short algorithm calculation time, emphasizes real-time performance, and is more suitable for real-time monitoring of substations.

[0006] The technical solution of this application is as follows:

[0007] An image recognition method for switchboards in substations, comprising the following steps:

[0008] Step 1: Use a fixed-position ball camera to shoot the original splitter board image in the substation, set a rectangular splitter identification area based on the original splitter board image, and obtain a splitter monitoring setting image. The pixel area of ​​the splitter board in the splitter identification area in the splitter monitoring setting image is greater than the pixel area threshold value;

[0009] Step 2: Use HSV color recognition to identify the splitting and closing monitoring setting image, obtain the splitting and closing plate binary image, and find the splitting and closing plate position;

[0010] Step 3, based on the endpoint direct calculation method and the extended edge line intersection method, the endpoints of the split-and-combined plate binary graph are found in parallel to obtain the endpoint information of the split-and-combined plate binary graph;

[0011] Step 4, extracting and correcting the original split-and-combined board image;

[0012] Step 5, find the position of the separation and combination dividing line, and cut the corrected original separation and combination board image into a "separation" part image and a "combination" part image;

[0013] Step 6, using HSV color recognition to detect arrows in the "split" part picture and the "combined" part picture respectively to generate an arrow binary map, where the arrow binary map includes the "split" part arrow binary map and the "combined" part arrow binary map;

[0014] Step 7: Calculate the number of white pixels in the binary image of the arrows in the "opening" part and the binary image of the arrows in the "closing" part to determine the opening and closing states.

[0015] Preferably, step 2 specifically includes the following steps:

[0016] According to the color of the split and close board, the HSV coefficient is set, and the split and close monitoring setting image is subjected to HSV color recognition, and the split and close board binary images are obtained respectively. The split and close board binary images include the binary images of the "split" part and the binary images of the "closed" part. Gaussian blur and opening operations are performed on the split and close board binary images to eliminate noise points, and the denoised binary images are obtained;

[0017] The HSV color recognition of the splitting and closing monitoring setting image specifically includes the following steps: setting the HSV coefficient according to the color of the splitting and closing plate, setting the HSV coefficient of the "closed" part in the splitting and closing monitoring setting image to red, setting the HSV coefficient of the "split" part to green, and extracting red and green pixel points to obtain a binary image of the "split" part and a binary image of the "closed" part.

[0018] The HSV coefficient setting specifically includes the following steps: analyzing the pixel grayscale values ​​of the split and combined monitoring setting image that are higher than the abnormal value threshold value as abnormal values, after removing the abnormal values, analyzing the average grayscale of the pixels of the split and combined monitoring setting image, if the average grayscale of the pixels is lower than the average grayscale threshold value, the grayscale of the split and combined monitoring setting image is increased to the range of 0 to 255, and after processing the brightness information, HSV color recognition is performed:

[0019] The Gaussian blur selects a 7*7 Gaussian kernel; for the opening operation, the convolution kernel is constructed using the np.ones() function, and the convolution kernel size is 3*3;

[0020] The convolution kernel for the opening operation is constructed using the np.ones() function, and the convolution kernel size is 3*3.

[0021] Preferably, the endpoint direct calculation method specifically includes the following steps:

[0022] Respectively find the endpoints of the separating plate in the binary image of the "separating" part and the binary image of the "combining" part; adopt the method of traversing pixels to calculate the distances from the pixel points of the white part in the binary image of the "separating" part and the binary image of the "combining" part to the image endpoints, sort the magnitudes of the distances, and respectively obtain the four points with the shortest distances from the white pixel points in the binary image of the "separating" part and the binary image of the "combining" part to the image endpoints as the image endpoints;

[0023] The method of extending the side lines to obtain intersection points specifically includes the following steps. Use the canny edge detection to identify the edge information of the binary image of the separating plate, use the statistical probability Hough line transform function cv2.HoughLinesP() to find the line segment information of the four sides of the binary image of the separating plate, identify the four sides of the white part of the binary image of the separating plate, extend the line segments where the four sides are located, and take the four intersection points of the extended lines. The four intersection points are the four endpoints of the binary image of the separating plate;

[0024] The four points with the shortest distances from the white pixel points in the binary image of the "separating" part to the four endpoints of the image are respectively denoted as A1, A2, A3, and A4 in the counterclockwise direction from the upper left. The four points with the shortest distances from the white pixel points in the binary image of the "combining" part to the four endpoints of the image are respectively denoted as B1, B2, B3, and B4 in the counterclockwise direction from the upper left. Then in the binary image of the separating plate, A3 and A4 respectively correspond to the coincidence of B1 and B2. Denote these two coincident points as C5 and C6. A1, A2, B3, and B4 are the four endpoints of the separating plate. C5 and C6 are the two endpoints of the coincidence line between the "separating" part and the "combining" part in the binary image of the separating plate. A1, A2, B3, B4, C5, and C6 are the endpoint information of the binary image of the separating plate.

[0025] Preferably, step 4 specifically includes the following steps:

[0026] Based on the endpoint information of the binary image of the split-merge board, the affine transformation method is used to correct the extracted split-merge board image into a rectangle; the affine transformation method uses the cv2.getPerspectiveTransform() function, and the parameters of the cv2.getPerspectiveTransform() function are the corresponding point coordinates of the original image and the target image, and the inclination angle of the split-merge board image is obtained through the corresponding front and back coordinates. After obtaining the inclination angle of the split-merge board image, the relationship between the point coordinates of the original split-merge board image and the corresponding point coordinates of the split-merge board image after rectangular correction is obtained, the correction of the split-merge board image is completed, and the corrected split-merge board image is obtained.

[0027] Preferably, step 5 specifically includes the following steps:

[0028] Calculate the position of the split-merge boundary line through the two endpoints C5 and C6 of the coincidence line of the binary images of the "split" part and the "merge" part in the corrected split-merge board image; the point coordinates of A1, A2, B3, B4, C5, and C6 are (a, b), (a, c), (d, b), (e, c), (f, b), (f, c) respectively, the width of the corrected split-merge board image is w, and the distance from the split-merge boundary line to the left side of the corrected split-merge board image is h. The position formula of the split-merge boundary line is:

[0029]

[0030] Cut the split-merge board into a "split" part image and a "merge" part image through the split-merge boundary line.

[0031] Step 7 specifically includes the following steps:

[0032] Calculate the number of white pixel points in the binary image of the "merge" part arrow and the binary image of the "split" part arrow respectively by traversing the pixel points. The number of white pixel points in the binary image of the "merge" part arrow is recorded as num1, and the number of white pixel points in the binary image of the "split" part arrow is recorded as num2. If num1>num2, it is judged as "merge"; if num1<num2, it is judged as "split"; if num1 = num2, it is judged as "uncertain".

[0033] A substation split-merge board image recognition system includes a split-merge recognition area setting unit, an HSV color recognition unit, an endpoint information acquisition unit, a correction unit, a split-merge cutting unit, an arrow binarization unit, and a split-merge state judgment unit;

[0034] The split-merge recognition area setting unit uses a fixed-position camera to capture the original split-merge board image in the substation, sets a rectangular split-merge recognition area based on the original split-merge board image, and obtains a split-merge monitoring setting image. The pixel area of the split-merge board in the split-merge monitoring setting image is greater than the pixel area threshold value;

[0035] The HSV color recognition unit uses HSV color recognition on the splitting and closing monitoring setting image to obtain the splitting and closing plate binary image and find out the splitting and closing plate position;

[0036] The endpoint information acquisition unit searches for the endpoints of the split-and-joint plate binary graph in parallel based on the endpoint direct calculation method and the extended edge line intersection method, and acquires the endpoint information of the split-and-joint plate binary graph;

[0037] The correction unit extracts and corrects the original split-and-combined plate image;

[0038] The split-and-combine cutting unit finds the split-and-combine dividing line position and cuts the corrected original split-and-combine plate image into a "split" part image and a "combine" part image;

[0039] The arrow binarization unit detects arrows in the "split" part picture and the "combined" part picture respectively through HSV color recognition to generate arrow binary maps, and the arrow binary maps include the "split" part arrow binary map and the "combined" part arrow binary map;

[0040] The split / close state judgment unit respectively calculates the number of white pixels in the "split" part arrow binary image and the "closed" part arrow binary image to judge the split / close state.

[0041] The working process of the HSV color recognition unit specifically includes the following steps:

[0042] According to the color of the split and close board, the HSV coefficient is set, and the split and close monitoring setting image is subjected to HSV color recognition, and the split and close board binary images are obtained respectively. The split and close board binary images include the binary images of the "split" part and the binary images of the "closed" part. Gaussian blur and opening operations are performed on the split and close board binary images to eliminate noise points, and the denoised binary images are obtained;

[0043] The HSV color recognition of the splitting and closing monitoring setting image specifically includes the following steps: setting the HSV coefficient according to the splitting and closing plate color, setting the HSV coefficient of the "closed" part in the splitting and closing monitoring setting image to red, setting the HSV coefficient of the "split" part to green, and extracting red and green pixel points to obtain a binary image of the "split" part and a binary image of the "closed" part;

[0044] The HSV coefficient setting specifically includes the following steps: analyzing the pixel grayscale values ​​of the split and combined monitoring setting image that are higher than the abnormal value threshold value as abnormal values, after removing the abnormal values, analyzing the average grayscale of the pixels of the split and combined monitoring setting image, if the average grayscale of the pixels is lower than the average grayscale threshold value, the grayscale of the split and combined monitoring setting image is increased to the range of 0 to 255, and after processing the brightness information, HSV color recognition is performed:

[0045] The Gaussian blur uses a 7*7 Gaussian kernel. The opening operation uses the np.ones() function to construct a convolution kernel with a size of 3*3.

[0046] The convolution kernel of the opening operation is constructed using the np.ones() function, and the size of the convolution kernel is 3*3.

[0047] The endpoint direct calculation method specifically includes the following steps:

[0048] Respectively find the endpoints of the separation plate in the binary image of the "separation" part and the binary image of the "combination" part; adopt the method of traversing pixels to calculate the distances from the pixel points of the white part in the binary image of the "separation" part and the binary image of the "combination" part to the endpoints of the image, sort the magnitudes of the distances, and respectively obtain the four points with the shortest distances from the white pixel points in the binary image of the "separation" part and the binary image of the "combination" part to the endpoints of the image as the endpoints of the image;

[0049] The method of extending the side lines to obtain intersection points specifically includes the following steps. Use canny edge detection to identify the edge information of the binary image of the separation plate, use the statistical probability Hough line transform function cv2.HoughLinesP() to find the line segment information of the four sides of the binary image of the separation plate, identify the four sides of the white part of the binary image of the separation plate, extend the line segments where the four sides are located, and take the four intersection points of the extension lines. The four intersection points are the four endpoints of the binary image of the separation plate;

[0050] The four points with the shortest distances from the white pixel points in the binary image of the "separation" part to the four endpoints of the image are respectively denoted as A1, A2, A3, and A4 from the upper left to the lower left in a counterclockwise direction. The four points with the shortest distances from the white pixel points in the binary image of the "combination" part to the four endpoints of the image are respectively denoted as B1, B2, B3, and B4 from the upper left to the lower left in a counterclockwise direction. Then, in the binary image of the separation plate, A3 and A4 respectively correspond to the coincidence of B1 and B2. Denote these two coincident points as C5 and C6. A1, A2, B3, and B4 are the four endpoints of the separation plate, C5 and C6 are the two endpoints of the coincidence line between the "separation" part and the "combination" part in the binary image of the separation plate, and A1, A2, B3, B4, C5, and C6 are the endpoint information of the binary image of the separation plate;

[0051] The working process of the correction unit specifically includes the following steps:

[0052] Based on the endpoint information of the binary image of the separation plate, use the affine transformation method to correct the extracted separation plate image into a rectangle; the affine transformation method uses the cv2.getPerspectiveTransform() function. The parameters of the cv2.getPerspectiveTransform() function are the corresponding point coordinates of the original image and the target image. After obtaining the inclination angle of the separation plate image, find the relationship between the coordinates of the original separation plate image points and the corresponding point coordinates of the rectangle-corrected separation plate picture, complete the correction of the separation plate image, and obtain the corrected separation plate picture;

[0053] The working process of the separation and combination cutting unit specifically includes the following steps:

[0054] Calculate the position of the separation and combination boundary line through the two endpoints C5 and C6 of the coincidence line of the binary images of the "separation" part and the "combination" part in the corrected separation and combination plate image; the coordinates of points A1, A2, B3, B4, C5, and C6 are (a, b), (a, c), (d, b), (e, c), (f, b), (f, c) respectively, the width of the corrected separation and combination plate image is w, and the distance from the separation and combination boundary line to the left side of the corrected separation and combination plate image is h. The position formula of the separation and combination boundary line is:

[0055]

[0056] Cut the separation and combination plate into the "separation" part image and the "combination" part image through the separation and combination boundary line;

[0057] The working process of the separation and combination state judgment unit specifically includes the following steps:

[0058] Calculate the number of white pixel points in the arrow binary image of the "separation" part and the arrow binary image of the "combination" part respectively by traversing the pixel points. Among them, the number of white pixel points in the arrow binary image of the "combination" part is denoted as num1, and the number of white pixel points in the arrow binary image of the "separation" part is denoted as num2. If num1 > num2, it is judged as "combination"; if num1 < num2, it is judged as "separation"; if num1 = num2, it is judged as "uncertain".

[0059] Compared with the prior art, the present invention has the following beneficial effects:

[0060] This paper proposes a method for identifying the separation and combination plate image of a substation, a color extraction method with adaptive brightness, which can extract image color information in environments with different brightness levels;

[0061] This application proposes a method for finding the endpoints of the separation and combination plate based on pixel traversal, which improves the accuracy of endpoint detection of the separation and combination plate; at the same time, it proposes a method for taking the intersection points of the extended side lines based on the Hough line transformation, which can find the original endpoints of the separation and combination plate in the case of corner defects of the separation and combination plate to correct the separation and combination plate image.

[0062] This application proposes an image tilt correction method based on the statistical probability Hough line transformation function, which uses canny edge detection and Hough line transformation to obtain the average tilt angle of the longitudinal line segments in the separation and combination plate image, and corrects the separation and combination plate image based on this. Obtaining the tilt angle of the tilt line consistent with the main direction of the separation and combination plate can correct the separation and combination plate image more accurately.

[0063] This application proposes a method for identifying the separation and combination state of the separation and combination plate based on color information, which can identify the pointing state of the arrow of the separation and combination plate at the pixel level. Description of the Drawings

[0064] Figure 1Flow chart of an image recognition method for the switch - board in a sub - station in this application

[0065] Figure 2 Schematic diagram of the original switch - board image in the sub - station in this embodiment

[0066] Figure 3 Binary image of the "open" part of the switch - board in the sub - station in this embodiment

[0067] Figure 4 Binary image of the "closed" part of the switch - board in the sub - station in this embodiment

[0068] Figure 5 Endpoint position map of the switch - board in the sub - station in this embodiment

[0069] Figure 6 Corrected switch - board image in this embodiment

[0070] Figure 7 Schematic diagram of the "closed" part after cutting the switch - board in this embodiment

[0071] Figure 8 Schematic diagram of the "open" part after cutting the switch - board in this embodiment

[0072] Figure 9 Binary image of the arrow in the "closed" part in this embodiment

[0073] Figure 10 Binary image of the arrow in the "open" part in this embodiment

[0074] Figure 11 Schematic diagram of the image endpoints in this embodiment Detailed implementation manners

[0075] Next, in combination with the drawings in the present invention, the technical solutions of the present invention will be clearly and completely described. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments in the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present invention.

[0076] The core idea of the present invention is to design a suitable algorithm for image pre - processing enhancement according to the characteristics of fundus images, then perform retinal vessel segmentation based on the information - migration fundus image segmentation network, and finally use a neural network combined with ordered classification for intelligent analysis and prediction to achieve the purpose of automatic diagnosis and film reading.

[0077] An image recognition method for the switch - board in a sub - station includes the following steps:

[0078] Step 1: Use a fixed - position camera to capture the original switch - board image in the sub - station(Figure 2 ) Based on the original split-merge board image, set a rectangular split-merge recognition area, and obtain a split-merge monitoring set image, where the pixel area of the split-merge board in the split-merge recognition area is greater than 50%;

[0079] The function of the pixel area of the split-merge board in the split-merge recognition area being greater than 50% is:

[0080] When recognizing the split-merge board, it is necessary to make the split-merge board occupy the main part of the field of view. If the image area occupied by the split-merge board is less than 50%, the image will obtain more environmental information around the split-merge board, including various straight edges and color areas, which will interfere with the recognition of the split-merge board; if the image area occupied by the split-merge board is too large, the image may not contain the entire split-merge board, so some information of the split-merge board may be missed, affecting the recognition of the split-merge board.

[0081] Step 2: Perform HSV color recognition on the split-merge monitoring set image to obtain a binary image of the split-merge board and find the position of the split-merge board;

[0082] Step 3: Parallelly search for the endpoints of the binary image of the split-merge board based on the endpoint direct calculation method and the extended side line intersection method to obtain the endpoint information of the binary image of the split-merge board;

[0083] Step 4: Extract and correct the original split-merge board image:

[0084] Step 5: Find the position of the split-merge boundary line and cut the corrected original split-merge board image into "split" part pictures and "merge" part pictures;

[0085] Step 6: Detect the arrows in the "split" part pictures and "merge" part pictures respectively through HSV color recognition. Since the arrows are yellow, identify the yellow parts in the "split" part and "merge" part images to generate arrow binary images, which include the "split" part arrow binary image and the "merge" part arrow binary image. Among them, the gray value of the yellow part in the arrow binary image is 255, that is, white; the gray values of the remaining parts are all 0, that is, black.

[0086] Step 7: Calculate the number of white pixel points in the "split" part arrow binary image and the "merge" part arrow binary image respectively to judge the split-merge state.

[0087] Step 2 is specifically:

[0088] Set the HSV coefficients according to the color of the split-merge board, perform HSV color recognition on the split-merge monitoring set image, and obtain the binary image of the split-merge board respectively. The binary image of the split-merge board includes the "split" part binary image (green part binary image, Figure 3 ) and the "merge" part binary image (red part binary image, Figure 4), Gaussian blur and opening operations are performed on the binary image of the split and combined plate to eliminate noise points, and a binary image after denoising is obtained; wherein in the process of eliminating noise points, different masks are set based on the pixels of the binary image of the split and combined plate, and the size of the mask is selected to ensure that all noise points are eliminated;

[0089] The HSV color recognition of the splitting and closing monitoring setting image specifically includes the following steps: setting the HSV coefficient according to the color of the splitting and closing plate, setting the HSV coefficient of the "closed" part in the splitting and closing monitoring setting image to red, setting the HSV coefficient of the "split" part to green, and extracting red and green pixel points to obtain a binary image of the "split" part (a binary image of the green part) and a binary image of the "closed" part (a binary image of the red part).

[0090] The HSV coefficient setting specifically includes the following steps: analyzing the pixel grayscale value of the split-and-close monitoring setting image, which is higher than 250 as an abnormal value. After removing the abnormal value, analyzing the average grayscale of the pixel points of the split-and-close monitoring setting image. If the average grayscale of the pixel points is lower than 130, it means that the image is dark. The grayscale of the split-and-close monitoring setting image is increased to the range of 0 to 255. After processing the brightness information, HSV color recognition is performed. Since the colors in the split-and-close board image are close to red, green, and yellow, the selected HSV coefficient after debugging is:

[0091] 1. Red: hmin:156,0

[0092] smin:43

[0093] vmin:46

[0094] hmax:180,10

[0095] smax:255

[0096] vmax:255

[0097] 2. Green: hmin: 35

[0098] smin:43

[0099] vmin:46

[0100] hmax:99

[0101] smax:255

[0102] vmax:255

[0103] 3. Yellow: hmin: 19

[0104] smin:130

[0105] vmin:52

[0106] hmax: 50

[0107] smax: 255

[0108] vmax: 255

[0109] The Gaussian blur selects a 7*7 Gaussian kernel; for the opening operation, the convolution kernel is constructed using the np.ones() function, and the size of the convolution kernel is 3*3;

[0110] This algorithm uses Gaussian smoothing filtering to smooth the noise. Since the Gaussian kernel parameter is odd and the pixel size of the image used in this algorithm is 1920*1080, after testing, the 7*7 Gaussian kernel has a good smoothing effect on interference factors such as noise and burrs in this type of image. After Gaussian smoothing, the edges of the information in the image are more regular and smooth, which is conducive to edge recognition.

[0111] In addition, the opening operation eliminates the adhered parts and isolated redundant pixel blocks in the image. Among them, after testing, the convolution kernel is constructed using the np.ones() function, and the size of the convolution kernel is 3*3.

[0112] Step 3 specifically includes the following steps:

[0113] Find the four points closest to the image endpoints in the white part of the binary image of the "fen" part (green binary image) and the binary image of the "he" part (red binary image) respectively. The four closest points are used as the endpoints of the fenhe board; by using the method of traversing pixels, calculate the distance from the pixel points in the white part to the image endpoints. The distance is the Euclidean distance, and sort the distance sizes to obtain the four points with the shortest distance from the white pixel points in the red binary image and the green binary image to the image endpoints respectively;

[0114] The image endpoints refer to the four vertices of the rectangular fenhe recognition area, that is, the red points in the attached figure Figure 11 in the attached figure;

[0115] As Figure 11 shown, the four vertices of the rectangular fenhe recognition area (red points in the attached figure) are the image endpoints. Find the white pixel points closest to the image endpoints respectively (blue points in the attached figure). The four white pixel points closest to the image endpoints (the four vertices of the rectangular fenhe recognition area) are the four endpoints of this fenhe board. When calculating the distance, the traversal method for white pixel points is: from top to bottom, from left to right.

[0116] The above first method for finding the endpoints of the split-merge board is based on the direct endpoint calculation method. The following is the second method for finding the endpoints, which is the method of extending the edge lines to find the intersection points. The two algorithms for finding the endpoints run in parallel and output results simultaneously, enhancing the robustness of the system. When there are defects at the corners of the split-merge board, the method of extending the edge lines to find the intersection points is more applicable. When initially detecting a split-merge board, both the direct endpoint calculation method and the method of extending the edge lines to find the intersection points are used to find the endpoints of the split-merge board. When there is a large difference (the distance is greater than 2% of the image length) between the endpoint coordinate results obtained by the direct endpoint calculation method and the method of extending the edge lines to find the intersection points, it is considered that there are defects at the corners of the split-merge board. At this time, the result of the method of extending the edge lines to find the intersection points is adopted.

[0117] The method of extending the edge lines to find the intersection points specifically includes the following steps. Due to the wear problem at the corner parts of the split-merge board, it is difficult to directly find the endpoints when identifying the endpoints of the split-merge board. Therefore, the following method of extending the edge lines to find the intersection points is used: Use the canny edge detection to identify the edge information of the binary image of the split-merge board, where threshold1 and threshold2 are set to 10 and 128 respectively. Use the statistical probability Hough line transform function cv2.HoughLinesP() to find the line segment information of the four sides of the binary image of the split-merge board. The parameter settings of the statistical probability Hough line transform function cv2.HoughLinesP() are: rho = 1.0; theta = numpy.pi / 180; threshod = 100; minLineLength = 50; maxLineGap = 20. After identifying the four sides of the white part of the binary image of the split-merge board, extend the line segments where the four sides are located, and take the four intersection points of the extended lines. The four intersection points are the four endpoints of the binary image of the split-merge board.

[0118] As Figure 5 shown, in the binary image of the red part, the four points with the shortest distances from the white pixel points to the four endpoints of the image are denoted as A1, A2, A3, and A4 in counterclockwise order from the upper left. In the binary image of the green part, the four points with the shortest distances from the white pixel points to the four endpoints of the image are denoted as B1, B2, B3, and B4 in counterclockwise order from the upper left. Then, in the binary image of the split-merge board, A3 and A4 respectively correspond to the coincidence of B1 and B2. Denote these two coincident points as C5 and C6. Then, A1, A2, B3, and B4 are the four endpoints of the split-merge board, C5 and C6 are the two endpoints of the coincident line between the red part and the green part in the binary image of the split-merge board, and A1, A2, B3, B4, C5, and C6 are the endpoint information of the binary image of the split-merge board.

[0119] Step 4 is specifically:

[0120] After finding the endpoints of the binary image of the split-merge board, use perspective transformation to extract the split-merge board image from the original split-merge board image, and correct the tilted split-merge board image into a rectangle to obtain the corrected split-merge board picture ( Figure 6)。

[0121] The endpoint information required for correction is obtained from the binary image of the split-merge plate in Step 3.

[0122] In Step 3, the endpoint direct calculation method or the extended side line intersection method is used to obtain the endpoint information of the binary image of the split-merge plate. In Step 4, based on the endpoint information of the binary image of the split-merge plate, the affine transformation method is used to correct the extracted split-merge plate image into a rectangle. The affine transformation method uses the cv2.getPerspectiveTransform() function. The parameters of the cv2.getPerspectiveTransform() function are the corresponding point coordinates of the original image and the target image. After obtaining the tilt angle of the split-merge plate image, the relationship between the point coordinates of the original split-merge plate image and the corresponding point coordinates of the split-merge plate image after rectangular correction is obtained to complete the correction of the split-merge plate image and obtain the corrected split-merge plate image.

[0123] Step 5 is specifically as follows:

[0124] Calculate the position of the split-merge boundary line through the two endpoints C5 and C6 of the coincidence line of the binary image (red part binary image) of the "split" part and the binary image (red part binary image) of the "merge" part in the corrected split-merge plate image; the point coordinates of A1, A2, B3, B4, C5, and C6 are (a, b), (a, c), (d, b), (e, c), (f, b), and (f, c) respectively. The width of the corrected split-merge plate image is w, and the distance from the split-merge boundary line to the left side of the corrected split-merge plate image is h. The position formula of the split-merge boundary line is:

[0125]

[0126] Cut the split-merge plate into the "split" part image ( Figure 8 ) and the "merge" part image ( Figure 7 ) through the split-merge boundary line.

[0127] In the split-merge plate, the background colors of the split and merge parts are red and green, and the arrow color is yellow. After the original split-merge plate image is cut through Step 5, for the two split images, the "split" part image and the "merge" part image ( Figure 7 , 8 ), HSV color recognition is performed respectively to identify yellow. Then, the two result images of HSV color recognition are converted into the "split" part arrow binary image and the "merge" part arrow binary image ( Figure 9 , 10 ). Among them, the gray value of the yellow part in the arrow binary image is 255, that is, white; the gray values of the remaining parts are all 0, that is, black.

[0128] Step 7 is specifically as follows: It specifically includes the following steps:

[0129] By traversing the pixel points, calculate the number of white pixel points in the "separation" part arrow binary image and the "combination" part arrow binary image respectively. Denote the number of white pixel points in the "combination" part arrow binary image as num1, and the number of white pixel points in the "separation" part arrow binary image as num2. If num1 > num2, it is judged as combination; if num1 < num2, it is judged as separation; if num1 = num2, it is judged as uncertain.

[0130] Traversal process: The pixel points are traversed from top to bottom and from left to right, and compare the pixel sizes occupied by the indicated arrows in the "separation" part arrow binary image ( Figure 10 ) and the "combination" part arrow binary image ( Figure 9 ).

[0131] A substation separation and combination board image recognition system includes a separation and combination recognition area setting unit, an HSV color recognition unit, an endpoint information acquisition unit, a correction unit, a separation and combination cutting unit, an arrow binarization unit, and a separation and combination state judgment unit;

[0132] The separation and combination recognition area setting unit uses a fixed-position dome camera to capture the original separation and combination board image in the substation, sets a rectangular separation and combination recognition area based on the original separation and combination board image, and obtains a separation and combination monitoring setting image. The pixel area of the separation and combination board in the separation and combination monitoring setting image is greater than the pixel area threshold;

[0133] The HSV color recognition unit performs HSV color recognition on the separation and combination monitoring setting image to obtain a separation and combination board binary image and find the position of the separation and combination board;

[0134] The endpoint information acquisition unit parallelly searches for the endpoints of the separation and combination board binary image based on the endpoint direct calculation method and the extended side line intersection method to obtain the endpoint information of the separation and combination board binary image;

[0135] The correction unit extracts and corrects the original separation and combination board image;

[0136] The separation and combination cutting unit finds the position of the separation and combination dividing line and cuts the corrected original separation and combination board image into a "separation" part picture and a "combination" part picture;

[0137] The arrow binarization unit respectively detects the arrows in the "separation" part picture and the "combination" part picture through HSV color recognition to generate arrow binary images, and the arrow binary images include a "separation" part arrow binary image and a "combination" part arrow binary image;

[0138] The separation and combination state judgment unit respectively calculates the number of white pixel points in the "separation" part arrow binary image and the "combination" part arrow binary image to judge the separation and combination state.

[0139] The working process of the HSV color recognition unit specifically includes the following steps:

[0140] According to the color of the split and close board, the HSV coefficient is set, and the split and close monitoring setting image is subjected to HSV color recognition, and the split and close board binary images are obtained respectively. The split and close board binary images include the binary images of the "split" part and the binary images of the "closed" part. Gaussian blur and opening operations are performed on the split and close board binary images to eliminate noise points, and the denoised binary images are obtained;

[0141] The HSV color recognition of the splitting and closing monitoring setting image specifically includes the following steps: setting the HSV coefficient according to the splitting and closing plate color, setting the HSV coefficient of the "closed" part in the splitting and closing monitoring setting image to red, setting the HSV coefficient of the "split" part to green, and extracting red and green pixel points to obtain a binary image of the "split" part and a binary image of the "closed" part;

[0142] The HSV coefficient setting specifically includes the following steps: analyzing the pixel grayscale values ​​of the split and combined monitoring setting image that are higher than the abnormal value threshold value as abnormal values, after removing the abnormal values, analyzing the average grayscale of the pixels of the split and combined monitoring setting image, if the average grayscale of the pixels is lower than the average grayscale threshold value, the grayscale of the split and combined monitoring setting image is increased to the range of 0 to 255, and after processing the brightness information, HSV color recognition is performed:

[0143] The Gaussian blur uses a 7*7 Gaussian kernel. The opening operation uses the np.ones() function to construct a convolution kernel with a size of 3*3.

[0144] The convolution kernel for the open operation is constructed using the np.ones() function, and the convolution kernel size is 3*3.

[0145] The endpoint direct calculation method specifically includes the following steps:

[0146] Find the endpoints of the split and close plates in the binary image of the "split" part and the binary image of the "close" part respectively; use the method of traversing pixels to calculate the distances from the pixels of the white part of the binary image of the "split" part and the binary image of the "close" part to the image endpoints, sort the distances, and obtain the four points with the shortest distances from the white pixels of the binary image of the "split" part and the binary image of the "close" part to the image endpoints as the image endpoints;

[0147] The method of extending the edge line and taking the intersection point specifically includes the following steps: using canny edge detection to identify the edge information of the split plate binary image, using the statistical probability Hough line transformation function cv2.HoughLinesP() to find the line segment information of the four edges of the split plate binary image, identifying the four edges of the white part of the split plate binary image, extending the line segments where the four edges are located, and taking the four intersection points of the extended lines, which are the four endpoints of the split plate binary image;

[0148] In the "separation" part of the binary image, the four points with the shortest distances from the white pixel points to the four endpoints of the image are denoted as A1, A2, A3, and A4 in counterclockwise order from the upper left. In the "combination" part of the binary image, the four points with the shortest distances from the white pixel points to the four endpoints of the image are denoted as B1, B2, B3, and B4 in counterclockwise order from the upper left. Then, in the binary image of the separation and combination board, A3 and A4 respectively correspond to and coincide with B1 and B2. These two coincident points are denoted as C5 and C6. A1, A2, B3, and B4 are the four endpoints of the separation and combination board, C5 and C6 are the two endpoints of the coincidence line between the "separation" part and the "combination" part in the binary image of the separation and combination board, and A1, A2, B3, B4, C5, and C6 are the endpoint information of the binary image of the separation and combination board;

[0149] The working process of the correction unit specifically includes the following steps:

[0150] Based on the endpoint information of the binary image of the separation and combination board, use the affine transformation method to correct the extracted separation and combination board image into a rectangle; the affine transformation method uses the cv2.getPerspectiveTransform() function, and the parameters of the cv2.getPerspectiveTransform() function are the corresponding point coordinates of the original image and the target image. After obtaining the inclination angle of the separation and combination board image, find the relationship between the point coordinates of the original separation and combination board image and the corresponding point coordinates of the rectangle-corrected separation and combination board picture to complete the correction of the separation and combination board image and obtain the corrected separation and combination board picture;

[0151] The working process of the separation and combination cutting unit specifically includes the following steps:

[0152] Calculate the position of the separation and combination boundary line through the two endpoints C5 and C6 of the coincidence line between the "separation" part binary image and the "combination" part binary image in the corrected separation and combination board picture; the point coordinates of A1, A2, B3, B4, C5, and C6 are (a, b), (a, c), (d, b), (e, c), (f, b), and (f, c) respectively. The width of the corrected separation and combination board picture is w, and the distance from the separation and combination boundary line to the left side of the corrected separation and combination board picture is h. The position formula of the separation and combination boundary line is:

[0153]

[0154] Cut the separation and combination board into the "separation" part picture and the "combination" part picture through the separation and combination boundary line;

[0155] The working process of the separation and combination state judgment unit specifically includes the following steps:

[0156] By traversing the pixel points, calculate the number of white pixel points in the binary image of the "combine" part arrow and the binary image of the "separate" part arrow respectively. Denote the number of white pixel points in the binary image of the "combine" part arrow as num1, and the number of white pixel points in the binary image of the "separate" part arrow as num2. If num1 > num2, it is judged as combine; if num1 < num2, it is judged as separate; if num1 = num2, it is judged as uncertain.

[0157] In the specification provided herein, a large number of specific details are set forth. It will be understood, however, that embodiments of the invention may be practiced without these specific details. In some instances, well-known methods, structures and techniques have not been shown in detail in order not to obscure an understanding of this description.

[0158] Similarly, it should be understood that in order to streamline this disclosure and assist in understanding one or more of the various inventive aspects, in the foregoing description of the exemplary embodiments of the invention, various features of the invention are sometimes grouped together in a single embodiment, figure, or description thereof. However, the disclosed method should not be construed as reflecting an intention that the claimed invention requires more features than are expressly recited in each claim. Rather, as the claims reflect, inventive aspects lie in less than all of the features of the single foregoing disclosed embodiment. Thus, the claims following the detailed description are hereby expressly incorporated into this detailed description, with each claim standing on its own as a separate embodiment of the invention.

[0159] Those skilled in the art should understand that the modules or units or groups of the devices in the examples disclosed herein may be arranged in the devices as described in this embodiment, or alternatively may be located in one or more devices different from the devices in this example. The modules in the foregoing examples may be combined into one module or further divided into multiple sub-modules.

[0160] Those skilled in the art can understand that the modules in the devices of the embodiments can be adaptively changed and arranged in one or more devices different from this embodiment. The modules or units or groups in the embodiments can be combined into one module or unit or group, and further can be divided into multiple sub-modules or sub-units or sub-groups. Except that at least some of such features and / or processes or units are mutually exclusive, any combination can be used to combine all the features disclosed in this specification (including the accompanying claims, abstract and drawings) and all the processes or units of any method or device so disclosed. Unless otherwise expressly stated, each feature disclosed in this specification (including the accompanying claims, abstract and drawings) can be replaced by an alternative feature that provides the same, equivalent or similar purpose.

[0161] In addition, those skilled in the art will appreciate that although some of the embodiments described herein include certain features included in other embodiments but not others, the combination of features of different embodiments is meant to be within the scope of the present invention and forms different embodiments. For example, in the following claims, any one of the claimed embodiments can be used in any combination.

[0162] In addition, some of the embodiments are described herein as a method or a combination of method elements that can be implemented by a processor of a computer system or by other devices performing the functions. Therefore, a processor having the necessary instructions for implementing the method or method elements forms a means for implementing the method or method elements. In addition, the elements described herein of the apparatus embodiments are examples of the following apparatus: the apparatus is used to implement the functions performed by the elements for the purpose of implementing the present invention.

[0163] The various techniques described herein can be implemented in conjunction with hardware or software, or a combination thereof. Thus, the methods and apparatuses of the present invention, or certain aspects or portions of the methods and apparatuses of the present invention, may take the form of program code (i.e., instructions) embedded in a tangible medium, such as a floppy disk, a CD-ROM, a hard disk drive, or any other machine-readable storage medium, wherein when the program is loaded into and executed by a machine such as a computer, the machine becomes an apparatus for practicing the present invention.

[0164] In the case where the program code is executed on a programmable computer, the computing device generally includes a processor, a processor-readable storage medium (including volatile and non-volatile memory and / or storage elements), at least one input device, and at least one output device. Among them, the memory is configured to store the program code; the processor is configured to execute the method of the present invention according to the instructions in the program code stored in the memory.

[0165] By way of example and not limitation, computer-readable media include computer storage media and communication media. Computer-readable media include computer storage media and communication media. Computer storage media stores information such as computer-readable instructions, data structures, program modules, or other data. Communication media typically embodies computer-readable instructions, data structures, program modules, or other data in a modulated data signal such as a carrier wave or other transmission mechanism, and includes any information delivery media. Combinations of any of the above are also included within the scope of computer-readable media.

[0166] As used herein, unless otherwise specified, the use of ordinal numbers such as "first", "second", "third", etc. to describe ordinary objects merely indicates different instances of similar objects and is not intended to imply that the objects so described must be in a given order, whether in terms of time, space, ranking, or in any other way.

[0167] Although the invention has been described based on a limited number of embodiments, those skilled in the art in this technical field will understand, based on the above description, that other embodiments can be conceived within the scope of the invention as thus described. In addition, it should be noted that the language used in this specification has been mainly selected for the purpose of readability and teaching, rather than for the purpose of explaining or limiting the subject matter of the invention. Therefore, many modifications and variations will be obvious to those of ordinary skill in the art in this technical field without departing from the scope and spirit of the appended claims. For the scope of the invention, the disclosure of the invention is illustrative rather than restrictive, and the scope of the invention is defined by the appended claims.

Claims

1. A method for image recognition of the switching board in a substation, characterized in that: The following steps are involved: Step 1: Use a fixed-position ball camera to shoot the original splitter board image in the substation, set a rectangular splitter identification area based on the original splitter board image, and obtain a splitter monitoring setting image. The pixel area of ​​the splitter board in the splitter identification area in the splitter monitoring setting image is greater than the pixel area threshold value; Step 2: Use HSV color recognition to identify the splitting and closing monitoring setting image, obtain the splitting and closing plate binary image, and find the splitting and closing plate position; Step 3, based on the endpoint direct calculation method and the extended edge line intersection method, the endpoints of the split-and-combined plate binary graph are found in parallel to obtain the endpoint information of the split-and-combined plate binary graph; Step 4, extracting and correcting the original split-and-combined board image; Step 5, find the position of the separation and combination boundary, and cut the corrected original separation and combination board image into a "separated" part image and a "combined" part image; Step 6, using HSV color recognition to detect arrows in the "split" part picture and the "combined" part picture respectively to generate an arrow binary map, where the arrow binary map includes the "split" part arrow binary map and the "combined" part arrow binary map; Step 7: Calculate the number of white pixels in the binary image of the arrows in the "open" part and the binary image of the arrows in the "close" part to determine the open and close states.

2. A substation switchboard image recognition method according to claim 1, characterized in that: The step 2 specifically includes the following steps: According to the color of the split and close board, the HSV coefficient is set, and the split and close monitoring setting image is subjected to HSV color recognition, and the split and close board binary images are obtained respectively. The split and close board binary images include the "split" part binary image and the "close" part binary image. The split and close board binary images are subjected to Gaussian blur and opening operations to eliminate noise points, and the denoised binary images are obtained; The HSV color recognition of the splitting and closing monitoring setting image specifically includes the following steps: setting the HSV coefficient according to the color of the splitting and closing plate, setting the HSV coefficient of the "closed" part of the splitting and closing monitoring setting image to red, setting the HSV coefficient of the "closed" part to green, and extracting red and green pixel points to obtain the binary image of the "closed" part and the binary image of the "closed" part.

3. A substation switchboard image recognition method according to claim 2, characterized in that: The HSV coefficient setting specifically includes the following steps: analyzing the pixel grayscale values ​​of the split and combined monitoring setting image that are higher than the abnormal value threshold value as abnormal values, after removing the abnormal values, analyzing the average grayscale of the pixels of the split and combined monitoring setting image, if the average grayscale of the pixels is lower than the average grayscale threshold value, the grayscale of the split and combined monitoring setting image is increased to the range of 0 to 255, and after processing the brightness information, HSV color recognition is performed: The Gaussian blur uses a 7*7 Gaussian kernel. The opening operation uses the np.ones() function to construct a convolution kernel with a size of 3*3. The convolution kernel for the open operation is constructed using the np.ones() function, and the convolution kernel size is 3*3.

4. A substation switchboard image recognition method according to claim 2, characterized in that: The endpoint direct calculation method specifically includes the following steps: Find the endpoints of the separating and combining plates in the binary images of the "separating" part and the "combining" part respectively; adopt the method of traversing pixels to calculate the distances from the white pixel points in the binary images of the "separating" part and the "combining" part to the image endpoints, sort the magnitudes of the distances, and obtain the four points with the shortest distances from the white pixel points to the image endpoints in the binary images of the "separating" part and the "combining" part respectively as the image endpoints; The method of extending the side lines to obtain intersection points specifically includes the following steps: use canny edge detection to identify the edge information of the binary image of the separating and combining plates, use the statistical probability Hough line transform function cv2.HoughLinesP() to find the line segment information of the four sides of the binary image of the separating and combining plates, identify the four sides of the white part of the binary image of the separating and combining plates, extend the line segments where the four sides are located, and take the four intersection points of the extended lines. The four intersection points are the four endpoints of the binary image of the separating and combining plates; The four points with the shortest distances from the white pixel points to the four endpoints of the image in the binary image of the "separating" part are denoted as A1, A2, A3, and A4 counterclockwise from the upper left, and the four points with the shortest distances from the white pixel points to the four endpoints of the image in the binary image of the "combining" part are denoted as B1, B2, B3, and B4 counterclockwise from the upper left. Then, in the binary image of the separating and combining plates, A3 and A4 correspond to B1 and B2 respectively and coincide. Denote these two coincident points as C5 and C6. A1, A2, B3, and B4 are the four endpoints of the separating and combining plates, C5 and C6 are the two endpoints of the coincident line between the "separating" part and the "combining" part in the binary image of the separating and combining plates, and A1, A2, B3, B4, C5, and C6 are the endpoint information of the binary image of the separating and combining plates.

5. The method for identifying the separating and combining plate image of a substation according to claim 4, wherein: The specific steps of step 4 are as follows: Based on the endpoint information of the binary image of the separating and combining plates, use the affine transformation method to correct the extracted separating and combining plate image into a rectangle; the affine transformation method uses the cv2.getPerspectiveTransform() function to obtain the relationship between the coordinates of the original separating and combining plate image points and the corresponding points of the rectangle-corrected separating and combining plate image, complete the correction of the separating and combining plate image, and obtain the corrected separating and combining plate image.

6. The method for identifying the separating and combining plate image of a substation according to claim 5, wherein: The specific steps of step 5 are as follows: Calculate the position of the separating and combining boundary line through the two endpoints C5 and C6 of the coincident line between the binary images of the "separating" part and the "combining" part in the corrected separating and combining plate image; the coordinates of points A1, A2, B3, B4, C5, and C6 are (a, b), (a, c), (d, b), (e, c), (f, b), and (f, c) respectively. The width of the corrected separating and combining plate image is w, and the distance from the separating and combining boundary line to the left side of the corrected separating and combining plate image is h. The position formula of the separating and combining boundary line is: Cut the separating and combining plate into a "separating" part image and a "combining" part image through the separating and combining boundary line.

7. The method for identifying the separating and combining plate image of a substation according to claim 6, wherein: The specific steps of step 7 are as follows: By traversing pixel points, calculate the number of white pixel points in the "split" part arrow binary image and the "merge" part arrow binary image respectively. Denote the number of white pixel points in the "merge" part arrow binary image as num1, and the number of white pixel points in the "split" part arrow binary image as num2. If num1 > num2, it is judged as merge; if num1 < num2, it is judged as split; if num1 = num2, it is judged as uncertain.

8. A substation switching board image recognition system, characterized in that: It includes a split-merge recognition area setting unit, an HSV color recognition unit, an endpoint information acquisition unit, a correction unit, a split-merge cutting unit, an arrow binarization unit, and a split-merge status judgment unit; The split-merge recognition area setting unit uses a fixed-position camera to capture the original split-merge board image in the substation, sets a rectangular split-merge recognition area based on the original split-merge board image, and obtains a split-merge monitoring setting image. The pixel area of the split-merge board in the split-merge recognition area in the split-merge monitoring setting image is greater than the pixel area threshold value. The HSV color recognition unit performs HSV color recognition on the split-merge monitoring setting image to obtain a split-merge board binary image and find the position of the split-merge board. The endpoint information acquisition unit parallelly searches for the endpoints of the split-merge board binary image based on the direct endpoint calculation method and the intersection method of extending the side line to obtain the endpoint information of the split-merge board binary image. The correction unit extracts and corrects the original split-merge board image. The split-merge cutting unit finds the position of the split-merge dividing line and cuts the corrected original split-merge board image into a "split" part picture and a "merge" part picture. The arrow binarization unit respectively detects the arrows in the "split" part picture and the "merge" part picture through HSV color recognition to generate arrow binary images, and the arrow binary images include a "split" part arrow binary image and a "merge" part arrow binary image. The split-merge status judgment unit calculates the number of white pixel points in the "split" part arrow binary image and the "merge" part arrow binary image respectively to judge the split-merge status.

9. The substation split-merge board image recognition system according to claim 8, characterized in that: The working process of the HSV color recognition unit specifically includes the following steps: Set hsv coefficients according to the split-merge board color, perform hsv color recognition on the split-merge monitoring setting image, and respectively obtain split-merge board binary images. The split-merge board binary images include a "split" part binary image and a "merge" part binary image. Perform Gaussian blur and opening operation on the split-merge board binary images to eliminate noise and obtain the denoised binary images. Performing hsv color recognition on the split-merge monitoring setting image specifically includes the following steps: Set hsv coefficients according to the split-merge board color, respectively set the hsv coefficients of the "merge" part in the split-merge monitoring setting image to red, set the hsv coefficients of the "split" part to green, and extract red and green pixel points to obtain a "merge" part binary image and a "split" part binary image. The HSV coefficient setting specifically includes the following steps: analyzing the pixel grayscale values ​​of the split and combined monitoring setting image that are higher than the abnormal value threshold value as abnormal values, after removing the abnormal values, analyzing the average grayscale of the pixels of the split and combined monitoring setting image, if the average grayscale of the pixels is lower than the average grayscale threshold value, the grayscale of the split and combined monitoring setting image is increased to the range of 0 to 255, and after processing the brightness information, HSV color recognition is performed: The Gaussian blur uses a 7*7 Gaussian kernel. The opening operation uses the np.ones() function to construct a convolution kernel with a size of 3*3. The convolution kernel for the open operation is constructed using the np.ones() function, and the convolution kernel size is 3*3.

10. The substation switchboard image recognition system according to claim 8, characterized in that: The endpoint direct calculation method specifically includes the following steps: Find the endpoints of the split and close plates in the binary image of the "split" part and the binary image of the "close" part respectively; use the method of traversing pixels to calculate the distances from the white pixel points in the binary image of the "split" part and the binary image of the "close" part to the image endpoints, sort the distances, and obtain the four points with the shortest distances from the white pixel points in the binary image of the "split" part and the binary image of the "close" part to the image endpoints as the image endpoints; The method of extending the edge line and taking the intersection point specifically includes the following steps: using canny edge detection to identify the edge information of the split plate binary image, using the statistical probability Hough line transformation function cv2.HoughLinesP() to find the line segment information of the four edges of the split plate binary image, identifying the four edges of the white part of the split plate binary image, extending the line segments where the four edges are located, and taking the four intersection points of the extended lines, which are the four endpoints of the split plate binary image; The four points with the shortest distances from the white pixels in the binary image of the "split" part to the four endpoints of the image are recorded as A1, A2, A3, and A4 from the upper left counterclockwise, and the four points with the shortest distances from the white pixels in the binary image of the "combined" part to the four endpoints of the image are recorded as B1, B2, B3, and B4 from the upper left counterclockwise. Then, in the binary image of the split-combined plate, A3 and A4 coincide with B1 and B2 respectively, and these two coincident points are recorded as C5 and C6. A1, A2, B3, and B4 are the four endpoints of the split-combined plate, C5 and C6 are the two endpoints of the coincidence line between the "split" part and the "combined" part in the binary image of the split-combined plate, and A1, A2, B3, B4, C5, and C6 are the endpoint information of the binary image of the split-combined plate; The working process of the correction unit specifically includes the following steps: Based on the endpoint information of the split-and-combined board binary image, the extracted split-and-combined board image is corrected into a rectangle using the affine transformation method; the affine transformation method uses the cv2.getPerspectiveTransform() function to obtain the relationship between the coordinates of the original split-and-combined board image points and the coordinates of the corresponding points of the split-and-combined board image after the rectangle correction, complete the correction of the split-and-combined board image, and obtain the corrected split-and-combined board image; The working process of the splitting and cutting unit specifically includes the following steps: Calculate the position of the separation and combination boundary line through the two end points C5 and C6 of the coincidence line of the binary images of the "separation" part and the "combination" part in the corrected separation and combination plate image; the coordinates of points A1, A2, B3, B4, C5, and C6 are (a, b), (a, c), (d, b), (e, c), (f, b), and (f, c) respectively. The width of the corrected separation and combination plate image is w, and the distance from the separation and combination boundary line to the left side of the corrected separation and combination plate image is h. The position formula of the separation and combination boundary line is: Cut the separation and combination plate into the "separation" part image and the "combination" part image through the separation and combination boundary line; The working process of the separation and combination state judgment unit specifically includes the following steps: Calculate the number of white pixel points in the binary image of the "combination" part arrow and the binary image of the "separation" part arrow respectively by traversing the pixel points. The number of white pixel points in the binary image of the "combination" part arrow is denoted as num1, and the number of white pixel points in the binary image of the "separation" part arrow is denoted as num2. If num1 > num2, it is judged as "combination"; if num1 < num2, it is judged as "separation"; if num1 = num2, it is judged as "uncertain".

Images