Method, System and Medium for Identifying Transformer Oil Level in Distribution Substation

Abstract

The present invention discloses a method, system and medium for identifying the oil level of a transformer in a distribution substation. The method includes acquiring an image of the transformer oil level and performing image preprocessing; positioning the pointer of the transformer oil gauge to obtain the axis position and pointer position of the oil level gauge; identifying the starting point and ending point of the oil level gauge to obtain the main oil level reading; judging the result, and if it is identified as abnormal, outputting an alarm message. The present invention performs homomorphic filtering and median filtering on the oil level image to eliminate the influence of illumination and noise during image acquisition, improving the adaptability to different illuminations; then performing Hough transform on the axis position of the oil level image to detect the pointer, avoiding searching for global edge points, greatly reducing the calculation amount and the running time of the program, and improving the accuracy of searching; finally, the oil level reading can be obtained according to the angle method. The method of the present invention has a high accuracy, a short program running time, can adapt to various illumination conditions, and has high robustness.

Description

Technical Field

[0001] The present invention belongs to the technical field of transformer oil level detection, and particularly relates to a method, a system and a medium for identifying the oil level of a transformer in a distribution room. Background Art

[0002] Transformer equipment is an important equipment in the power system. Whether it can operate normally plays a crucial role in the safety of the entire power system. Aiming at the problems of low efficiency and poor accuracy in manually collecting the transformer oil level during patrol inspection, situations such as "false detection" and "missed detection" are likely to occur. At the same time, the transformer information provided during maintenance has hysteresis, is not suitable for management, and cannot clearly understand and master the oil level status of the transformer in a timely manner. This paper studies and proposes to identify the oil level status of the transformer through the monitoring video in the distribution room, judge whether the oil level meets the requirements for safe operation, and when the oil level reaches the red warning position, timely discover and give an early warning, remind the on-duty personnel to check and eliminate problems in time, control potential hazards in the earliest stage, reduce the safety hazards caused by too low oil level, and ensure the safe production of the distribution room.

[0003] The existing technologies for identifying the transformer oil level image through video monitoring mainly rely on target color features and geometric features. These methods are often affected by factors such as brightness changes and complex backgrounds, resulting in poor generalization ability. Summary of the Invention

[0004] The main purpose of the present invention is to overcome the shortcomings and deficiencies of the existing technologies, and provide a method, a system and a medium for identifying the oil level of a transformer in a distribution room, which access the oil level monitoring video stream, detect the oil level gauge, identify the oil level mark, give an oil level warning when the oil level reaches the warning position, and realize all-day automatic monitoring and warning.

[0005] In order to achieve the above purpose, the present invention adopts the following technical solutions:

[0006] One aspect of the present invention provides a method for identifying the oil level of a transformer in a distribution room, including the following steps:

[0007] Obtain the transformer oil level image and perform image preprocessing;

[0008] Locate the pointer of the transformer oil gauge to obtain the axis position and the pointer position of the oil level gauge;

[0009] Identify the starting point and the ending point of the oil level gauge to obtain the main oil level reading;

[0010] Judge the result. If it is identified as abnormal, output an alarm message.

[0011] As a preferred technical solution, the obtaining the transformer oil level image and performing image preprocessing is specifically:

[0012] Obtain the transformer oil level image, including obtaining the image of the shooting device and intercepting the image from the video; for intercepting the image from the video, comprehensively consider the scene, requirements, and performance, set the video frame selection interval, and convert the intercepted single-frame image into a JPG format image that can be processed by the oil gauge status recognition model.

[0013] Image preprocessing, including homomorphic filtering, median filtering, binarization, and morphological processing; used to remove some noise interference in the picture and make the data preprocessing operations of the training pictures and the pictures to be predicted consistent.

[0014] As a preferred technical solution, the homomorphic filtering is specifically:

[0015] Filter the oil level gauge image through a Gaussian-type homomorphic filter as follows:

[0016] ;

[0017] ;

[0018] where is the distance of point from the center of the frequency rectangle; is the cut-off frequency, and in the present invention, is taken as the median of ; the parameter is used to control the sharpening of the homomorphic filter, and in the present invention, is set; the parameters and control the change range of , and in the present invention, is set, , .

[0019] As a preferred technical solution, the recognition of the axis position of the oil level gauge adopts the method of traversing the pointer pixels, specifically:

[0020] Use the threshold method to distinguish the pointer and other areas in the image, as follows:

[0021] ;

[0022] where is the binarized image after threshold method processing, is the set threshold;

[0023] Use the erosion operation in morphology to process the binarized image after threshold method processing.

[0024] As a preferred technical solution, the recognition of the axis position of the oil level gauge adopts the method of traversing the pointer pixels, specifically:

[0025] Perform contour extraction operation on the image after image morphological processing: Frame the pointer interrupted into two segments by the axis center with two minimum bounding rectangles, and calculate the centroid coordinates of these two rectangles respectively. , ;

[0026] Connect the centroid coordinates of the two rectangles and calculate the slope of the line. And the intercept , and the line equation is ;

[0027] Traverse the pixels on this line, make a judgment according to RGB, and obtain the first non - black point on the line. And the last non - black point , and finally find the mid - point to obtain the coordinates of the axis center , as shown in the following formula:

[0028] ;

[0029] .

[0030] As a preferred technical solution, the identification of the pointer position of the oil level gauge adopts an improved Hough transform, specifically:

[0031] According to the previous work of detecting the axis center of the pointer, judge the approximate position where the pointer is located;

[0032] Use the minimum rectangle to frame the ROI where the pointer is located;

[0033] Perform Hough transform within this ROI to detect the pointer line;

[0034] Judge the detection result according to the feature that the pointer is the longest, and only output and display the longest line, and obtain the coordinates of the two endpoints of the line, which are the detected pointer;

[0035] Judge the distance from the two endpoints of the pointer to the axis center, and the end with the larger distance is the coordinate of the pointer head. .

[0036] As a preferred technical solution, the positions of the starting point and the ending point of the oil level gauge are identified by color segmentation and contour area method, specifically as follows:

[0037] Convert the picture from RGB to HSV color space;

[0038] According to the H value of red, make a red mask, and perform a bitwise_and operation with the mask and the original image to obtain the region - of - interest image;

[0039] Perform an erosion operation on the obtained region of interest image to remove interfering points that are not of interest, and then perform a dilation operation to expand the edge of the red region;

[0040] Use the contour method to mark the rectangles at the starting point and the ending point, and obtain the centroid of the rectangle. Then, the coordinates of the two centroids correspond to the position coordinates of the starting point and the ending point , .

[0041] As a preferred technical solution, the angle method is used to identify the main oil level reading, specifically as follows:

[0042] According to the axis center coordinates , the pointer head coordinates , the starting point coordinates and the ending point coordinates , the angle between scale 0 and 10 is obtained by the method of vector included angle , then:

[0043] The angle of the graduated dial is ;

[0044] The included angle between each scale of the oil level gauge ;

[0045] Then, according to the pointer head coordinates and the starting point coordinates and the ending point coordinates , the included angle between the pointer and the 0 scale line is calculated , then the reading of the oil level gauge is as follows:

[0046] .

[0047] Another aspect of the present invention provides a distribution room transformer oil level discrimination system, which is characterized in that it is applied to the above-mentioned distribution room transformer oil level discrimination method, and includes an image acquisition module, a pointer positioning module, an oil level reading acquisition module, and an output module;

[0048] The image acquisition module is used to acquire the transformer oil level image and perform image preprocessing;

[0049] The pointer positioning module is used to position the transformer oil gauge pointer to obtain the axis center position and the pointer position of the oil level gauge;

[0050] The oil level reading acquisition module is used to identify the starting point and the ending point of the oil level gauge to obtain the main oil level reading;

[0051] The output module is used to judge the result. If it is identified as abnormal, an alarm message is output.

[0052] Another aspect of the present invention provides a storage medium storing a program, characterized in that when the program is executed by a processor, the above-mentioned method for identifying the oil level of a transformer in a distribution room is implemented.

[0053] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0054] The present invention performs necessary image preprocessing operations such as homomorphic filtering and median filtering on the oil level image to eliminate the influence of illumination and noise during image acquisition, and improve the adaptability of the algorithm to different illuminations; then, it searches for the pointer contour of the oil level image after binarization and morphological operations, and performs the Hough transform within this area according to the axis center position to detect the pointer, avoiding searching for global edge points, greatly reducing the calculation amount and the running time of the program, and improving the accuracy of the search; the oil level image is converted to the HSV color space. Since the starting point and the ending point of the oil level are red, the starting point and the ending point are segmented according to the H value of red, so as to obtain the coordinate positions of the starting point and the ending point of the oil level; finally, the reading of the oil level can be obtained according to the angle method. This algorithm has a high accuracy, a short program running time, and can adapt to various illumination conditions, and has high robustness. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] Figure 1 is a flowchart of the method for identifying the oil level of a transformer in a distribution room according to an embodiment of the present invention;

[0056] Figure 2 is a schematic diagram of the process for determining the axis center of the oil level gauge according to an embodiment of the present invention;

[0057] Figure 3 is a schematic diagram of the principle of the Hough transform according to an embodiment of the present invention;

[0058] Figure 4 is a schematic diagram of using the improved Hough transform to identify the position of the pointer according to an embodiment of the present invention

[0059] Figure 5 is a schematic diagram of the structure of the system for identifying the oil level of a transformer in a distribution room according to an embodiment of the present invention;

[0060] Figure 6 is a schematic diagram of the structure of the storage medium according to an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0061] To enable those skilled in the art to better understand the solution of this application, the following will clearly and completely describe the technical solution in the embodiments of this application with reference to the accompanying drawings in the embodiments of this application. Obviously, the described embodiments are only a part of the embodiments of this application, rather than all the embodiments. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative efforts belong to the scope of protection of this application.

[0062] Embodiment:

[0063] As Figure 1 shown, this embodiment provides a method for distinguishing the oil level of a transformer in a distribution substation, including the following steps:

[0064] S1. Obtain the image of the transformer oil level.

[0065] Preferably, obtaining the image of the transformer oil level includes obtaining the image of the shooting device and intercepting the image from the video; when intercepting the image from the video, the video frame selection interval is set considering the three aspects of scene, requirement, and performance, and the intercepted single-frame picture is converted into an image in JPG format that can be processed by the oil gauge state recognition model;

[0066] S2. Image preprocessing: including image filtering processing, image enhancement processing, and performing image preprocessing work on the picture after format conversion, removing some noise interference in the picture and ensuring that the data preprocessing operations of the training picture and the picture to be predicted are consistent;

[0067] Preferably, the oil level gauge image collected by the camera is easily affected by light, and a large amount of noise will be carried during shooting and transmission. In order to accurately read the oil level gauge subsequently, it is necessary to preprocess the image, including operations such as homomorphic filtering, median filtering, binarization, and morphological processing.

[0068] S2.1. Homomorphic filtering processing:

[0069] In view of the fact that the collected oil level gauge image is easily affected by light, a homomorphic filter is used to enhance the oil level gauge image under weak light. The homomorphic filter is a frequency-domain filtering algorithm based on the principle of the image illumination reflection imaging model. This filter can change the image gray range and enhance the contrast of the image. The present invention uses a Gaussian-type homomorphic filter to perform filtering processing on the oil level gauge image, enhancing the brightness of the weak-light oil level gauge image and improving the contrast of the image. Among them, the high-pass filter has the following calculation formula:

[0070] ;

[0071] ;

[0072] Among them, is the point distance from the center of the frequency rectangle ; is the cut-off frequency, and in the present invention, is the median value; the parameter is used to control the sharpening of the homomorphic filter. In the present invention, is set; the parameters and control the change range. In the present invention, is set, .

[0073] Using the processing result of the Gaussian-type homomorphic filter can brighten the relatively dark oil level gauge image during the camera acquisition process, and at the same time enhance the contrast between the pointer area and the non-pointer area of the oil level gauge image.

[0074] S2.2. Binarization and image morphological processing:

[0075] Since most of the pointers in the oil level gauge image are black, while most of the other areas of the oil level gauge are gray or white, the threshold method can be used to separate the pointer from the other areas. The binarized image after processing with the threshold method is expressed as:

[0076] ;

[0077] Among them, is the binarized image after processing with the threshold method, is the set threshold;

[0078] In the binarized image, it is found that there are still some small interference points in the image. The present invention effectively removes them by using the erosion operation in morphology.

[0079] S3. Oil gauge pointer positioning: The method of traversing the pointer pixels is used to identify the axis position of the oil level gauge, and then the improved Hough transform is used to identify the position of the pointer.

[0080] S3.1. Determination of the oil level gauge axis:

[0081] As Figure 2As shown, for the identification of the axis of the oil level gauge, the present invention uses the method of traversing the pointer pixels to identify the axis of the oil level gauge. It can be found that there are still significant differences in the color between the pointer axis and the black pointer. In the binary image, it can be found that the pointer is interrupted into two segments by the axis. According to this feature, first perform contour extraction on the image after morphological operations. Two minimum bounding rectangles can be used to frame the two interrupted segments of the pointer, and then calculate the centroids of these two rectangles respectively, and obtain the coordinates of the two mass points. , . Connect these two points to obtain a straight line on the pointer, and the slope of this straight line can be obtained based on these two points. and the intercept , thus obtaining the straight line equation: ;

[0082] Traverse the pixels on this straight line, and judge according to RGB to find the first non-black point and the last non-black point on the straight line, and record them. Finally, perform the operation of finding the midpoint of the coordinates of these two points to obtain the coordinates of the axis. .

[0083] ;

[0084] .

[0085] S3.2. Pointer position identification:

[0086] (1) Principle of Hough transform:

[0087] The Hough transform works based on the duality of points and lines. As Figure 3 shown, a straight line in the rectangular coordinate system is transformed into the straight line equation in the parameter space . Considering that the slope may be infinite, it is finally transformed into the curve equation in the polar coordinate space . Let Traverse the image with a step size of 1 according to . The schematic diagram is as Figure 3 shown.

[0088] (2) Improvement of Hough transform:

[0089] For the identification of the pointer position, a large number of scholars have adopted the global Hough transform. This method first transforms the coordinates into the polar coordinate system. A straight line in the polar coordinate system can be expressed as: ;

[0090] Among them, represents the distance from the origin to the straight line; Represents the inclination angle of the perpendicular line to the straight line.

[0091] Then, each point on the image edge is brought into the polar coordinate system conversion formula for calculation, and the result is recorded in the accumulator. . Because points on the same straight line have the same and , so for each pair of calculated, then . Finally, taking the maximum value in the accumulator can obtain the straight line detection result. This method requires calculating all edge points, with a large amount of calculation, and it is easy to detect many non-existent straight lines, which has a greater impact on the subsequent determination of the pointer position and the accuracy of the fuel level gauge reading.

[0092] As Figure 4 shown, the present invention adopts an improved Hough transform. First, according to the prior work of detecting the pointer axis center, the approximate position of the pointer is judged, and the ROI where the pointer is located is framed by the smallest rectangular box; then, the Hough transform is performed within this small range to detect the pointer straight line, which can greatly reduce the amount of calculation. Secondly, according to the characteristic that the pointer is the longest, the detection result is judged, and only the longest straight line is output and displayed, that is, the detected pointer. After detecting the pointer, the coordinates of the two endpoints of the straight line can be obtained, and then according to the distance from the head of the pointer to the axis center should be greater than the distance from the tail of the pointer to the axis center condition ( ), the coordinate of the head of the pointer is . Compared with the traditional Hough transform, this method reduces the search range of the transform by restricting the parameter space, greatly reducing the storage space and the amount of calculation, thereby improving the calculation speed.

[0093] S4. Main fuel level reading: Perform color segmentation based on the fact that the starting and ending points of the fuel level gauge are red to obtain the starting and ending coordinates, and then input them into the fuel gauge status recognition model to determine whether the fuel gauge is detected.

[0094] S4.1. Use the angle method to read the fuel level gauge. Determining the starting and ending positions of the fuel level gauge is quite important. Through observation, it is found that there are mostly red rectangular frames at the starting and ending points on the instrument scale ring. Therefore, the starting and ending positions are obtained by the methods of color segmentation and contour area. The specific steps are as follows:

[0095] (1) Convert the picture from RGB to the HSV color space; the reason for choosing to convert the picture to the HSV color space is that the HSV color space is easier to represent a color, and the hue represented by the H value can basically determine a certain color.

[0096] (2) According to the red H value, create a red mask, and perform a bitwise_and operation between the mask and the original image to obtain the region of interest image.

[0097] (3) There are some small interference points in the obtained region of interest image, and erosion operation is required to remove the non - interested interference points, and then dilation operation is performed to expand the edge of the red region.

[0098] (4) Use the contour method to mark the rectangles of the starting point and the ending point, and obtain the centroid of the rectangle. Then the coordinates of the two centroids correspond to the position coordinates of the starting point and the ending point , .

[0099] S4.2, Oil level gauge reading:

[0100] Use the angle method to read the pointer - type oil level gauge. According to the axis coordinates obtained above , the pointer head coordinates , the starting point coordinates and the ending point coordinates . Obtain the angle between scale 0 and 10 through the method of vector angle , then the angle of the graduated dial is , and the angle between each scale of the oil level gauge is . Then, according to the calculated pointer head coordinates and the starting point coordinates and the ending point coordinates , calculate the angle between the pointer and the 0 - scale line , then the reading of the oil level gauge can be calculated according to the following formula:

[0101] .

[0102] S5, Judgment result & Output warning information. According to the oil gauge status of the recognized picture, if it is normal, directly return "normal" and continue to recognize. If it is recognized as "abnormal", output a warning message.

[0103] As Figure 5 shown, in another embodiment of the present application, a distribution room transformer oil level discrimination system is provided. The system includes an image acquisition module, a pointer positioning module, an oil level reading acquisition module, and an output module;

[0104] The image acquisition module is used to acquire the transformer oil level image and perform image pre - processing;

[0105] The pointer positioning module is used to position the transformer oil gauge pointer to obtain the axis position and pointer position of the oil level gauge;

[0106] The oil level reading acquisition module is used to identify the starting point and the ending point of the oil level gauge to obtain the main oil level reading;

[0107] The output module is used to judge the result. If the identification is abnormal, an alarm message is output.

[0108] It should be noted here that the system provided in the above embodiment is only illustrated by the division of the above-mentioned functional modules. In practical applications, the above functions can be allocated to different functional modules according to needs, that is, the internal structure is divided into different functional modules to complete all or part of the functions described above. This system is applied to the method for identifying the oil level of the transformer in the distribution room in the above embodiment.

[0109] Such as Figure 6 As shown, in another embodiment of the present application, a storage medium is further provided, storing a program, and when the program is executed by a processor, the method for identifying the oil level of the transformer in the distribution room in the above embodiment is implemented, specifically:

[0110] Obtain the transformer oil level image and perform image preprocessing;

[0111] Locate the pointer of the transformer oil gauge to obtain the axis position and the pointer position of the oil level gauge;

[0112] Identify the starting point and the ending point of the oil level gauge to obtain the main oil level reading;

[0113] Judge the result. If the identification is abnormal, an alarm message is output.

[0114] It should be understood that each part of the present application can be implemented by hardware, software, firmware, or a combination thereof. In the above embodiment, multiple steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, any one or a combination of the following well-known technologies in the art can be used: discrete logic circuits with logic gate circuits for implementing logic functions on data signals, application-specific integrated circuits with appropriate combinational logic gate circuits, programmable gate arrays (PGA), field programmable gate arrays (FPGA), etc.

[0115] The above embodiment is a preferred embodiment of the present invention, but the embodiments of the present invention are not limited to the above embodiment. Any other changes, modifications, substitutions, combinations, and simplifications made without departing from the spirit and principle of the present invention shall be equivalent substitution methods and are all included in the protection scope of the present invention.

Claims

1. A method for identifying the oil level of a transformer in a distribution substation, characterized in that It includes the following steps: Obtain the transformer oil level image and perform image preprocessing; Locate the pointer of the transformer oil gauge to obtain the axis position and pointer position of the oil level gauge; The recognition of the axis position of the oil level gauge adopts the method of traversing the pointer pixels, specifically: Perform a contour extraction operation on the image after image morphological processing: frame the pointer interrupted into two segments by the axis center with two minimum bounding rectangles, and respectively obtain the centroid coordinates of these two rectangles , ; Connect the centroid coordinates of the two rectangles and calculate the slope of the line and the intercept , and the equation of the line is ; Traverse the pixels on this line, make a judgment based on RGB, and obtain the first non-black point on the line and the last non-black point , and finally find the midpoint to obtain the coordinates of the axis center , as shown in the following formula: ； ； The recognition of the pointer position of the oil level gauge adopts the improved Hough transform, specifically: According to the prior work of detecting the pointer axis, judge the approximate position where the pointer is located; Use the minimum rectangular box to frame the ROI where the pointer is located; Perform the Hough transform within this ROI to detect the pointer line; Judge the detection result according to the feature that the pointer is the longest, and only output and display the longest straight line, and obtain the coordinates of the two endpoints of the straight line, which is the detected pointer; Judge the distances from the two endpoints of the pointer to the axis center. The end with the larger distance is the coordinate of the pointer head ; Identify the starting point and ending point of the oil level gauge to obtain the main oil level reading; Identify the positions of the starting point and ending point of the oil level gauge by means of color segmentation and contour area, specifically as follows: Convert the picture from RGB to HSV color space; According to the H value of red, make a red mask, and perform a bitwise_and operation with the mask and the original image to obtain the region of interest image; Perform an erosion operation on the obtained region of interest image to remove non-interested interference points, and then perform a dilation operation to expand the edge of the red region; The rectangles at the starting point and the ending point are marked using the contour method, and the centroid of the rectangle is obtained. Then, the coordinates of the two centroids correspond to the position coordinates of the starting point and the ending point. , ; Judge the result. If it is identified as abnormal, output an alarm message.

2. The method for identifying the transformer oil level in the distribution substation according to claim 1, wherein The obtaining of the transformer oil level image and the performance of image preprocessing are specifically: Obtain the transformer oil level image, including obtaining the image of the shooting device and intercepting the image from the video; for intercepting the image from the video, set the video frame selection interval, and convert the intercepted single-frame picture into a JPG format image that can be processed by the oil gauge state recognition model; Image preprocessing includes homomorphic filtering, median filtering, binarization and morphological processing; It is used to remove some noise interference in the picture and make the data preprocessing operations of the training picture and the picture to be predicted consistent.

3. The method for identifying the transformer oil level in a distribution substation according to claim 2, characterized in that, The homomorphic filtering is specifically: By means of a Gaussian-type homomorphic filter filter the oil level gauge image as follows: ； ； Among them, is the point distance from the center of the frequency rectangle ; is the cut-off frequency, taking as the median value; the parameter is used to control the sharpening of the homomorphic filter, making ; the parameters and control the change range, let , .

4. The method for distinguishing the transformer oil level in the distribution room according to claim 2, wherein The binarization and image morphological processing are specifically: Use the threshold method to distinguish the pointer in the image from other regions, as shown in the following formula: ； Among them, is the binary image processed by the threshold method, is the set threshold value; Use the erosion operation in morphology to process the binarized image processed by the threshold method.

5. The method for identifying the transformer oil level in a distribution substation according to claim 1, wherein Use the angle method to recognize the main oil level reading, specifically as follows: According to the axis center coordinates , the pointer head coordinates , the starting point coordinates and the ending point coordinates , the angle between scale 0 and 10 is obtained by the method of vector included angle , then: The angle of the graduated dial is ; The included angle between each scale of the oil level gauge ; According to the coordinates of the pointer head again and the starting point coordinates as well as the ending point coordinates , calculate the angle between the pointer and the 0 scale line , then the reading of the fuel level gauge is as follows: 。 6. Distribution room transformer oil level discrimination system, characterized in that Applied to the distribution room transformer oil level discrimination method described in any one of claims 1-5, it includes an image acquisition module, a pointer positioning module, an oil level reading acquisition module, and an output module; The image acquisition module is used to obtain the transformer oil level image and perform image preprocessing; The pointer positioning module is used to locate the pointer of the transformer oil gauge to obtain the axis position and pointer position of the oil level gauge; The oil level reading acquisition module is used to identify the starting point and ending point of the oil level gauge to obtain the main oil level reading; The output module is used to judge the result. If it is identified as abnormal, output an alarm message.

7. A storage medium stores a program, characterized in that: When the program is executed by the processor, it implements the distribution room transformer oil level discrimination method described in any one of claims 1-5.

Images