Converter station pointer type instrument reading identification method and system

Abstract

The invention discloses a converter station pointer type instrument reading identification method and system. The method comprises the following steps: acquiring an instrument panel image and processing the instrument panel image; identifying the position of the instrument panel and the type of the instrument, and obtaining an included angle between a head scale and a tail scale on the instrument panel, a measuring range of the instrument panel and a scale initial value; extracting a scale image and a pointer image; acquiring a first scale, a tail scale, a center coordinate of the first scale and a center coordinate of the tail scale; obtaining a pointer and a center coordinate of the pointer; calculating the center coordinate of the instrument panel; calculating the rotation angle of the pointer, calculating the rotation ratio of the pointer in the scale, and obtaining the reading corresponding to the pointer in the instrument panel image according to the rotation ratio of the pointer in the scale, the measuring range of the instrument panel and the initial value of the scale. According to the invention, the problems of low reading efficiency and poor accuracy of the pointer type instrument of the converter station are solved.

Description

Technical Field

[0001] The invention belongs to the technical field of instrument identification, and in particular relates to a method and system for identifying readings of pointer-type instruments in a converter station. Background Art

[0002] Industrial instruments are devices used to measure, display, record and control various process parameters in industrial production. Industrial instrument reading recognition technology is an important part of the field of industrial automation, so this technology is one of the important research topics of many experts and scholars. In recent years, with the continuous development of computer vision, deep learning and other technologies, people have begun to use deep convolutional neural networks (CNNs) to automatically extract features from images, and more pointer-type industrial instrument reading recognition methods based on deep learning have been proposed. These methods have improved the accuracy and robustness of industrial instrument reading recognition methods, and have been applied to many industrial scenarios, promoting the development of industrial instrument reading recognition technology.

[0003] The converter station is the core component of the HVDC transmission system. The main function of the converter station is to complete the mutual conversion between AC and DC to ensure the safe and stable operation of the power system. Therefore, the converter station plays a key role in the HVDC transmission system. With the continuous development of the power industry, the number of converter stations has also increased. In order to read and record the readings of various pointer instruments used to monitor and control the voltage, current, power and other parameters of the power system in the converter station, the reading method of the converter station instrument is usually to manually read and record the instrument readings at the location of each instrument, or to monitor and shoot each instrument panel and use manual dial reading method.

[0004] The pointer meter reading method and existing pointer meter reading recognition technology in the power industry converter station have the following disadvantages:

[0005] 1. High cost. There are a large number of converter stations. In order to ensure that operators can read the meter readings accurately, a large number of professionally trained operators are needed to read and record the meter readings, which consumes a lot of manpower costs.

[0006] 2. Low efficiency. Manual reading of converter station instrument readings requires operators to read each instrument reading one by one and record the readings manually. When converter station instrument readings need to be read frequently, the limitations of manual reading of instrument readings are more obvious. In addition, manual reading of instrument readings is easily affected by the subjective judgment of operators, resulting in errors in instrument readings, thereby affecting the efficiency and accuracy of generating converter station instrument readings.

[0007] 3. Poor real-time performance. Manual reading of converter station instrument readings cannot achieve 24-hour uninterrupted monitoring and reading, nor can it provide real-time feedback on changes in instrument readings, resulting in the inability to detect and handle abnormal conditions in the converter station in a timely manner, affecting the production efficiency of the converter station and posing certain safety hazards.

[0008] 4. Correcting the instrument will increase errors and delays. Instrument correction in the process of instrument reading recognition usually uses key points marked by computer vision technology to achieve instrument correction. The errors in the correction process will accumulate in the subsequent instrument reading recognition process, affecting the accuracy of the instrument reading. At the same time, instrument correction takes time, thereby reducing the efficiency of instrument reading recognition.

[0009] 5. Data information management is difficult. After manually reading a large number of instrument readings, it is usually necessary to manually record and organize the data, and data entry errors are prone to occur. Therefore, it is difficult to achieve real-time and accurate information management of these production data. In modern industrial production, a large amount of data needs to be processed and analyzed in real time to support production optimization and decision-making. Manual reading of instrument readings is obviously difficult to meet this demand. Summary of the invention

[0010] The object of the present invention is to provide a method and system for identifying the readings of pointer-type instruments in a converter station, so as to solve the problems of low efficiency and poor accuracy of the readings of pointer-type instruments in a converter station.

[0011] The present invention is achieved through the following technical solutions:

[0012] A method for identifying readings of pointer-type meters in a converter station comprises the following steps:

[0013] Acquire the dashboard image and process the dashboard image;

[0014] Identify the location of the instrument panel and the instrument type in the instrument panel image, and obtain the angle between the first scale and the last scale on the instrument panel, the range of the instrument panel, and the initial value of the scale according to the instrument type;

[0015] Extracting a scale image and a pointer image from the image at the location where the recognized instrument panel is located;

[0016] Obtain the first scale and the last scale in the recognized scale image, as well as the center coordinates of the first scale and the center coordinates of the last scale;

[0017] Obtaining the pointer and the center coordinates of the pointer in the recognized pointer image;

[0018] Calculate the center coordinates of the instrument panel according to the center coordinates of the first scale, the center coordinates of the last scale, and the angle between the first scale and the last scale;

[0019] Calculate the rotation angle of the pointer based on the central coordinates of the dashboard and the central coordinates of the pointer, and calculate the rotation ratio of the pointer on the scale according to the rotation angle of the pointer and the angle between the starting scale and the ending scale. Obtain the reading corresponding to the pointer in the dashboard image according to the rotation ratio of the pointer on the scale, the range of the dashboard, and the initial value of the scale.

[0020] In some embodiments, the steps of obtaining the starting scale and the ending scale in the recognized scale image include:

[0021] There are N preset scales, and a certain scale i has one adjacent scale or two adjacent scales;

[0022] Calculate the distances from scale i to each of the N - 1 scales, and sort them from small to large. After sorting, select the first two smallest distances. The first two smallest distances are the distance dim from scale i to scale m and the distance din from scale i to scale n. The coordinates of scale i are (xi, yi), and the coordinates of scale m are (xm, ym);

[0023] When the absolute value of the difference between dim and din is less than or equal to a preset value, it is determined that scale i is not the starting scale or the ending scale;

[0024] When the absolute value of the difference between dim and din is greater than the preset value, and when xi < xm, it is determined that scale i is the starting scale, and when xi > xm, it is determined that scale i is the ending scale.

[0025] In some embodiments, the steps of obtaining the central coordinates of the starting scale and the central coordinates of the ending scale include:

[0026] Use the ellipse fitting method for the starting scale and the ending scale respectively to obtain the central coordinates of the starting scale and the central coordinates of the ending scale.

[0027] In some embodiments, the steps of identifying the position of the dashboard and the type of instrument in the dashboard image include:

[0028] Process the dashboard image based on the trained instrument recognition network model to identify the position of the dashboard and the type of instrument.

[0029] In some embodiments, the steps of extracting the scale image and the pointer image from the image of the position where the recognized dashboard is located include:

[0030] Process the image of the position where the dashboard is located based on the trained dashboard segmentation network model to segment out the dashboard segmentation result image;

[0031] Assign the scales in the dashboard segmentation result image to red and the pointers to green;

[0032] The colored instrument panel segmentation result image is split according to the three channels of the image to obtain the scale image and the pointer image.

[0033] In some embodiments, the step of calculating the rotation angle of the pointer includes:

[0034] The preset pointer rotation angle is ω, the angle between the first scale and the last scale is 2θ, the center of the pointer is point M, the center of the first scale is point A, the center of the last scale is point B, the center of the instrument panel is point C, the center of the straight line from point A to point B is point D, and the pointer rotation angle range is 2π-2θ;

[0035] When ACM =d AM When , then ω=π;

[0036] When AM =d BM When , then ω=(2π-2θ) / 2;

[0037] When AM <d BM When ω=Φ, Φ is the angle between the first scale and the pointer;

[0038] When AM >d BM When ω=2π-2θ-Φ, Φ is the angle between the tail scale and the pointer.

[0039] Among them, d ACM is the distance from point A to point C and then to point M, d AM is the distance from point A to point M, d BM is the distance from point B to point M, and Φ is solved using the acute angle formula of a triangle.

[0040] In some embodiments, the calculation formula of the rotation ratio of the pointer in the scale is:

[0041] Ratio = ω / (2π-20);

[0042] Among them, Ratio is the rotation ratio of the pointer in the scale.

[0043] In some embodiments, the calculation formula for the reading corresponding to the pointer in the instrument panel image is:

[0044]

[0045] In some embodiments, the instrument information is also stored; the instrument information includes a dashboard image, a dashboard reading and a dashboard image information, and the dashboard image is associated with the dashboard reading and the dashboard image information; the dashboard image information includes a converter station name, an instrument number, and an image acquisition time, and the converter station name, the instrument number, and the image acquisition time are associated; the acquired dashboard image is named in sequence according to the flow station name, the instrument type, the instrument number, and the image acquisition time.

[0046] A converter station pointer meter reading recognition system, comprising:

[0047] An instrument panel image acquisition module is used to acquire an instrument panel image of a pointer instrument in a converter station and process the instrument panel image;

[0048] The instrument recognition module is used to identify the position of the instrument panel in the instrument panel image and the instrument type, and obtain the angle between the first scale and the last scale on the instrument panel, the range of the instrument panel and the initial value of the scale;

[0049] An instrument panel recognition module, used to recognize a scale image and a pointer image in an image of a location where the instrument panel is located;

[0050] A scale recognition module is used to recognize the first scale and the last scale in the scale image, as well as the center coordinates of the first scale and the center coordinates of the last scale;

[0051] A pointer recognition module, used for recognizing the pointer and the center coordinates of the pointer in the pointer image;

[0052] The instrument panel reading acquisition module is used to obtain the reading corresponding to the pointer in the instrument panel image based on the center coordinates of the first scale, the center coordinates of the last scale, the center coordinates of the pointer, the angle between the first scale and the last scale, the range of the instrument panel and the initial value of the scale.

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

[0054] 1. The present invention obtains the instrument panel reading by using the center coordinates of the first scale, the center coordinates of the last scale, the angle between the first scale and the last scale, the range of the instrument panel and the initial value of the scale without correcting the tilt of the instrument. The instrument panel reading recognition process is simple, and the instrument reading efficiency and accuracy are high.

[0055] 2. The present invention stores instrument information, which is beneficial to maintaining the safety and normal operation of the converter station, expands the application scope of the converter station instrument data, accelerates the informatization construction of the converter station, improves the decision-making efficiency and management level of the converter station, and contributes to the intelligent development process of the converter station. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings in the embodiments will be briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without creative work.

[0057] Figure 1 The figure is a flow chart of a method for identifying pointer meter readings in a converter station according to an embodiment of the present invention.

[0058] Figure 2 The block diagram of the pointer meter reading recognition system for a converter station in an embodiment of the present invention is shown in FIG.

[0059] Figure 3 It is the central coordinate system of the instrument panel in the embodiment of the present invention.

[0060] Figure 4 It is the rotation angle coordinate system of the pointer in the embodiment of the present invention.

[0061] Figure 5 is the triangle coordinate system in the embodiment of the present invention DETAILED DESCRIPTION

[0062] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments.

[0063] Example 1

[0064] Reference Figure 1 , a method for identifying the reading of a pointer meter in a converter station, comprising the following steps:

[0065] S1, obtaining a dashboard image and processing the dashboard image;

[0066] The specific steps include:

[0067] S11. Obtain instrument panel images through dedicated video surveillance of instruments in the converter station, and the resolution of each instrument panel image is the same. Instruments in the instrument panel images collected by dedicated video surveillance have skew angles. Due to the complex and changeable business scenarios of the converter station, the number of instrument panel images containing each type of instrument is no less than 300.

[0068] S12. The instrument panel image is enhanced by histogram equalization to reduce the influence of factors such as uneven external lighting on the instrument panel image. The histogram of the instrument panel image is transformed into a uniformly distributed form by histogram equalization to improve the contrast of the instrument panel image, thereby achieving the effect of instrument panel image enhancement, which is beneficial to improving the accuracy and efficiency of later instrument readings.

[0069] S13. Use data annotation tools such as labelme to annotate the enhanced dashboard image to annotate the location of the dashboard and the type of instrument in the dashboard image data set.

[0070] S2, identifying the location of the instrument panel and the instrument type in the instrument panel image, and obtaining the angle between the first scale and the last scale on the instrument panel, the range of the instrument panel, and the initial scale value according to the instrument type;

[0071] The specific steps include:

[0072] S21. Train the instrument recognition network model using the labeled instrument panel image dataset. The structure of the instrument recognition network model is mainly divided into three parts, namely, feature extraction part, feature fusion part and recognition result decision part. Among them, the feature extraction part is used to extract the features of the instrument panel image, and the feature extraction part adopts a series of convolutional network layers and deconvolutional network layers, and simultaneously adopts residual network layers and bottleneck structures to reduce the network size; the recognition result decision part is mainly used to obtain the location frame of the instrument panel and the instrument category.

[0073] S22. Based on the trained instrument recognition network model, the instrument panel image is processed to identify the location of the instrument panel and the instrument type. At the same time, the instrument panel area image is intercepted according to the location of the instrument panel, and the angle between the first scale and the last scale on the instrument panel, the range of the instrument panel and the initial value of the scale are obtained according to the instrument type.

[0074] S23. Mark the scale area and pointer area of ​​the instrument panel using a labeling tool on the data set of the instrument panel area image.

[0075] S3, extracting a scale image and a pointer image from the image at the position where the recognized instrument panel is located;

[0076] The specific steps include:

[0077] S31. Input the labeled dashboard area image into the dashboard segmentation network model for training. The dashboard segmentation network model structure is mainly divided into two parts, namely the encoding part (encoder) and the decoding part (decoder). Among them, the encoding part is used to extract the features of the dashboard area image and obtain deep semantic information; the decoding part is used to restore spatial information.

[0078] S32, processing the image at the location of the dashboard based on the trained dashboard segmentation network model to segment the dashboard segmentation result image;

[0079] S33, assigning red to the scale and green to the pointer in the instrument panel segmentation result diagram;

[0080] S34, performing morphological processing such as corrosion and expansion on the instrument panel segmentation result image after coloring, so as to remove noise in the instrument panel segmentation result image;

[0081] Among them, the corrosion processing is used to eliminate the noise in the instrument panel segmentation result image and smoothly segment the scale boundary and pointer boundary; the expansion processing is used to increase the scale size and pointer size after the corrosion processing, so that the segmented scale boundary and pointer boundary are smooth and close to the actual size.

[0082] S35, splitting the instrument panel segmentation result image after denoising according to the three channels of the image to obtain a scale image and a pointer image.

[0083] Among them, there are two parts in the split red channel image, namely the scale image, one is the foreground, namely the split scale and both are close to white, and the other is the black background; there are two parts in the split green channel image, namely the pointer image: one is the foreground, namely the split pointer and both are close to white, and the other is the black background.

[0084] S4, obtaining the first scale and the last scale in the recognized scale image, as well as the center coordinates of the first scale and the center coordinates of the last scale;

[0085] The specific steps include:

[0086] S41. Get the first and last scales

[0087] Using the ellipse fitting method, each scale is completely surrounded by an ellipse, and the center coordinates of each scale are obtained from the fitted ellipse.

[0088] There are N scales in the preset scale image. Since the inclination angle of the instrument panel image collected by the instrument-specific video monitoring does not exceed 90°, a certain scale i in the scale image has one or two adjacent scales.

[0089] Calculate the distance from scale i to each of the N-1 scales, and sort them from small to large. After sorting, select the first two smallest distances, which are the distance dim from scale i to scale m and the distance din from scale i to scale n. Preset the coordinates of scale i (xi, yi) and scale m (xm, ym).

[0090] When the absolute value of the difference between dim and din is less than or equal to a preset value, it indicates that both scale m and scale n are adjacent scales, and it is determined that scale i is neither the first scale nor the last scale.

[0091] When the absolute value of the difference between dim and din is greater than the preset value, it indicates that scale m is an adjacent scale and scale n is a non - adjacent scale. And when xi < xm, it is determined that scale i is the first scale, and when xi > xm, it is determined that scale i is the last scale.

[0092] For example, if the center coordinates of any two scales are i(xi, yi) and j(xj, yj) respectively, the calculation formula for the distance dij from the center coordinate of scale i to the center coordinate of scale j is:

[0093]

[0094] Set the preset value to 5, then there is:

[0095] When |dim - din| ≤ 5, both scale m and scale n are adjacent scales;

[0096] When |dim - din| > 5, scale m is an adjacent scale and scale n is a non - adjacent scale.

[0097] Note that in this article, the coordinates used in the calculation are based on the upper - left corner of the scale image and the pointer image as the origin, the horizontal right direction as the x - axis direction, and the vertical downward direction as the y - axis direction.

[0098] S42. Respectively use the ellipse fitting method for the first scale and the last scale to obtain the center coordinates of the first scale and the center coordinates of the last scale.

[0099] S5. Obtain the pointer in the recognized pointer image and the center coordinate of the pointer;

[0100] Adopt the ellipse fitting method to completely enclose the pointer in the pointer image with an ellipse, and obtain the pointer center coordinate from the fitted ellipse.

[0101] S6. Calculate the center coordinate of the instrument panel according to the center coordinate of the first scale, the center coordinate of the last scale, and the included angle between the first scale and the last scale;

[0102] The specific steps include:

[0103] Refer to Figure 3, the center coordinates of the preset first scale are point A (x1, y1), the center coordinates of the tail scale are point B (x2, y2), the center coordinates of the instrument panel are point C (x3, y3), the midpoint of the straight line from point A to point B is point D (x4, y4), the angle between the first scale and the tail scale is 2θ, the length of the straight line from point A to point B is AB, the length of the straight line from point B to point C is BC, the length of the straight line from point C to point D is CD, the length of the straight line from point A to point C is AC, and the length of the straight line from point B to point C is BC, then AB=BC=AC, CD⊥AB, the angle between point A and point D is θ, and the angle between point B and point D is θ.

[0104] AB is calculated according to formula (1), and AD = AB*1 / 2;

[0105] CD is calculated based on AD and the angle θ between point A and point D. The calculation formula of CD is:

[0106] CD=AD / tanθ (2);

[0107] According to the Pythagorean theorem and CD calculation, the coordinates of point C are obtained. The calculation formula of point C (x3, y3) is:

[0108]

[0109] S7, calculating the rotation angle of the pointer according to the center coordinates of the instrument panel and the center coordinates of the pointer;

[0110] The specific steps include:

[0111] Reference Figure 4 , the preset pointer rotation position is point M (x5, y5), the pointer rotates to the middle position of the instrument panel's range is point M' (x6, y6), the pointer rotation angle is ω, the angle between the pointer rotation position and the first scale is α, and the angle α is greater than π, the angle between the pointer rotation position and the last scale is β, the pointer rotation angle range is α+β=2π-2θ, the distance from point A to point C and then to point M is d ACM , the distance from point A to point M is d AM , the distance from point B to point M is d BM .

[0112] When ACM =d AM When , it means that point A, point C and point M are on the same straight line, then ω=π.

[0113] When AM =d BM , it means that the pointer moves to the middle position of the instrument panel's range, then ω = (2π-2θ) / 2.

[0114] Reference Figure 5 , the calculation formula for the acute angle of a triangle is:

[0115]

[0116] When AM <d BM When , it means α<β, then ω=α. The value of α can be solved according to formula (1) and formula (4).

[0117] When AM >d BM When α>β, then ω=2π-2θ-β. The value of β can be solved according to formula (1) and formula (4).

[0118] S8, calculating the rotation ratio of the pointer in the scale according to the rotation angle of the pointer and the angle between the first scale and the last scale;

[0119] The formula for calculating the rotation ratio of the pointer in the scale is:

[0120] Ratio = ω / (2π-2θ) (5);

[0121] Among them, Ratio is the rotation ratio of the pointer in the scale.

[0122] S9. Obtain the reading corresponding to the pointer in the instrument panel image according to the rotation ratio of the pointer in the scale and the range and initial value of the scale of the instrument panel.

[0123] The calculation formula for the reading corresponding to the pointer in the instrument panel image is:

[0124]

[0125] Without correcting the tilt of the instrument, the instrument panel reading is obtained by using the center coordinates of the first scale, the center coordinates of the last scale, the angle between the first scale and the last scale, the range of the instrument panel and the initial value of the scale. The instrument panel reading recognition process is simple, and the instrument reading efficiency and accuracy are high.

[0126] In some embodiments, the converter station pointer meter reading identification method further includes storing meter information.

[0127] Among them, the instrument information includes the instrument panel image, instrument panel reading and instrument panel image information, and the instrument panel image is associated with the instrument panel reading and the instrument panel image information; the instrument panel image information includes the converter station name, instrument number, and image acquisition time, and the converter station name, instrument number, and image acquisition time are associated.

[0128] When the instrument panel image is obtained through the dedicated video monitoring of the instrument in the converter station, the number of the instrument in the image, the instrument image acquisition time, the converter station name and other information are recorded at the same time, and the instrument panel image is named in sequence according to the flow station name, instrument type, instrument number, image acquisition time, etc.

[0129] The meter readings obtained through the meter identification network model and the dial segmentation network model are stored for use in data viewing, management and application.

[0130] Storing instrument information is conducive to maintaining the safety and normal operation of the converter station, expanding the application scope of converter station instrument data, accelerating the informatization construction of the converter station, improving the decision-making efficiency and management level of the converter station, and contributing to the intelligent development process of the converter station.

[0131] The present invention also relates to a converter station pointer meter reading recognition system, referring to Figure 2 ,include:

[0132] An instrument panel image acquisition module is used to acquire an instrument panel image of a pointer instrument in a converter station and process the instrument panel image;

[0133] The instrument recognition module is used to identify the position of the instrument panel in the instrument panel image and the instrument type, and obtain the angle between the first scale and the last scale on the instrument panel, the range of the instrument panel and the initial value of the scale;

[0134] An instrument panel recognition module, used to recognize a scale image and a pointer image in an image of a location where the instrument panel is located;

[0135] A scale recognition module is used to recognize the first scale and the last scale in the scale image, as well as the center coordinates of the first scale and the center coordinates of the last scale;

[0136] A pointer recognition module, used for recognizing the pointer and the center coordinates of the pointer in the pointer image;

[0137] The instrument panel reading acquisition module is used to obtain the reading corresponding to the pointer in the instrument panel image based on the center coordinates of the first scale, the center coordinates of the last scale, the center coordinates of the pointer, the angle between the first scale and the last scale, the range of the instrument panel and the initial value of the scale;

[0138] The instrument information storage module is used to store the instrument panel image, instrument panel reading and instrument panel image information, and associate the instrument panel image with the instrument panel reading and instrument panel image information.

[0139] The above description is only a preferred embodiment of the present invention and does not limit the present invention in any form. Any simple modification or equivalent change made to the above embodiment based on the technical essence of the present invention shall fall within the protection scope of the present invention.

Claims

1. A method for identifying the reading of a pointer meter in a converter station, characterized in that: It includes the following steps: Obtain the dashboard image and process the dashboard image; Identify the position of the dashboard and the type of instrument in the dashboard image, and obtain the angle between the starting scale and the ending scale on the dashboard, the range of the dashboard, and the initial value of the scale according to the type of instrument; Extract the scale image and the pointer image from the image of the position where the identified dashboard is located; Obtain the starting scale and the ending scale in the identified scale image, as well as the central coordinates of the starting scale and the central coordinates of the ending scale; Obtain the pointer and the central coordinates of the pointer in the identified pointer image; Calculate the central coordinates of the dashboard according to the central coordinates of the starting scale, the central coordinates of the ending scale, and the angle between the starting scale and the ending scale; Calculate the rotation angle of the pointer according to the central coordinates of the dashboard and the central coordinates of the pointer, calculate the rotation ratio of the pointer in the scale according to the rotation angle of the pointer and the angle between the starting scale and the ending scale, and obtain the reading corresponding to the pointer in the dashboard image according to the rotation ratio of the pointer in the scale, the range of the dashboard, and the initial value of the scale.

2. A method for identifying pointer meter readings in a converter station according to claim 1, characterized in that: The steps of obtaining the starting scale and the ending scale in the identified scale image include: It is preset that there are N scales, and for a certain scale i, there is one adjacent scale or two adjacent scales; Calculate the distances from scale i to each of the N - 1 scales respectively, and sort them from small to large. After sorting, select the first two smallest distances. The first two smallest distances are the distance dim from scale i to scale m and the distance din from scale i to scale n. The coordinates of scale i are (xi, yi), and the coordinates of scale m are (xm, ym); When the absolute value of the difference between dim and din is less than or equal to the preset value, it is determined that scale i is not the starting scale or the ending scale; When the absolute value of the difference between dim and din is greater than the preset value, and when xi < xm, it is determined that scale i is the starting scale, and when xi > xm, it is determined that scale i is the ending scale.

3. A method for identifying pointer meter readings in a converter station according to claim 1 or 2, characterized in that: The steps of obtaining the central coordinates of the starting scale and the central coordinates of the ending scale include: Respectively adopt the fitting ellipse method for the starting scale and the ending scale to obtain the central coordinates of the starting scale and the central coordinates of the ending scale.

4. A method for identifying pointer meter readings in a converter station according to claim 1, characterized in that: The steps of identifying the position of the dashboard and the type of instrument in the dashboard image include: Process the dashboard image based on the already trained instrument recognition network model to identify the position of the dashboard and the type of instrument.

5. A method for identifying pointer meter readings in a converter station according to claim 1 or 4, characterized in that: The steps of extracting the scale image and the pointer image from the image of the position where the identified dashboard is located include: Process the image of the position where the dashboard is located based on the already trained dashboard segmentation network model to segment out the dashboard segmentation result image; Assign the scale in the dashboard segmentation result image to red and the pointer to green; Split the colored dashboard segmentation result image according to the three channels of the image to obtain the scale image and the pointer image.

6. A method for identifying pointer meter readings in a converter station according to claim 1, characterized in that: The steps of calculating the rotation angle of the pointer include: It is preset that the rotation angle of the pointer is ω, the angle between the starting scale and the ending scale is 2θ, the center of the pointer is point M, the center of the starting scale is point A, the center of the ending scale is point B, the center of the dashboard is point C, and the center of the straight line from point A to point B is point D. The range of the pointer rotation angle is 2π - 2θ; When ACM =d AM When , then ω=π; When AM =d BM When , then ω=(2π-2θ) / 2; When AM <d BM When ω=Φ, Φ is the angle between the first scale and the pointer; When AM >d BM When ω=2π-2θ-Φ, Φ is the angle between the tail scale and the pointer. Among them, d ACM is the distance from point A to point C and then to point M, d AM is the distance from point A to point M, d BM is the distance from point B to point M, and Φ is solved using the acute angle formula of a triangle.

7. A method for identifying pointer meter readings in a converter station according to claim 6, characterized in that: The calculation formula of the rotation ratio of the pointer in the scale is: Ratio = ω / (2π-20); Among them, Ratio is the rotation ratio of the pointer in the scale.

8. A method for identifying pointer meter readings in a converter station according to claim 7, characterized in that: The calculation formula for the reading corresponding to the pointer in the instrument panel image is:

9. A method for identifying pointer meter readings in a converter station according to claim 1, characterized in that: It also includes storing instrument information; the instrument information includes instrument panel images, instrument panel readings and instrument panel image information, and the instrument panel image is associated with the instrument panel readings and instrument panel image information; the instrument panel image information includes the converter station name, instrument number, and image acquisition time, and the converter station name, instrument number, and image acquisition time are associated; the acquired instrument panel image is named in sequence according to the flow station name, instrument type, instrument number, and image acquisition time.

10. A converter station pointer meter reading recognition system, characterized in that: include: An instrument panel image acquisition module is used to acquire instrument panel images of pointer instruments in the converter station and process the instrument panel images; The instrument recognition module is used to identify the position of the instrument panel in the instrument panel image and the instrument type, and obtain the angle between the first scale and the last scale on the instrument panel, the range of the instrument panel and the initial value of the scale; An instrument panel recognition module, used to recognize a scale image and a pointer image in an image of a location where the instrument panel is located; A scale recognition module is used to recognize the first scale and the last scale in the scale image, as well as the center coordinates of the first scale and the center coordinates of the last scale; A pointer recognition module, used for recognizing the pointer and the center coordinates of the pointer in the pointer image; The instrument panel reading acquisition module is used to obtain the reading corresponding to the pointer in the instrument panel image based on the center coordinates of the first scale, the center coordinates of the last scale, the center coordinates of the pointer, the angle between the first scale and the last scale, the range of the instrument panel and the initial value of the scale.

Images