Pointer type instrument reading identification method and system of transformer substation inspection robot

Abstract

The invention provides a pointer type instrument reading identification method and system of a transformer substation inspection robot. The method comprises the following steps of 1, obtaining an inspection image shot by the transformer substation inspection robot; 2, preprocessing the inspection image obtained in the step 1; 3, converting the inspection image obtained in the step 2 and a registered image, wherein a pre-established corresponding relation exists between the registered image and the standard model image; and step 4, mapping line segments identified on the inspection image to the registration image, then mapping the line segments to the standard model image through the registration image, analyzing and identifying a pointer, and then performing instrument reading according to the pointer. According to the method, the pointer table can be efficiently and accurately identified and analyzed in a complex environment, and the inspection precision of the robot is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of pointer instrument readings, and in particular to a pointer instrument reading recognition method and system for a substation inspection robot. Background Art

[0002] As a key device in the monitoring system, industrial instruments are responsible for monitoring parameters such as temperature, flow, pressure, etc. in production, and displaying the system operation status in real time to prevent equipment failure. Industrial instruments are divided into two categories: pointer type and digital type. In some dangerous environments, such as substations, power plants, and oil production plants, pointer type has excellent anti-interference due to its mechanical structure design, so it has significant advantages in these industrial fields.

[0003] However, in remote areas or places with harsh conditions or dangerous environments, it may be difficult for staff to read meters during inspections. In addition, manual meter reading is not applicable to some industries that require 24 / 7 monitoring. In view of this, the current method is to use computer vision technology to identify meter readings. In order to improve the accuracy of meter reading recognition, high-quality meter images must be obtained. A common method of collecting meter images is to take pictures through inspection robots, such as Figure 1 as shown in . Figure 1 The middle left picture shows a scene in which a track-mounted robot is conducting inspections and shooting at a fixed position. It is a robot system that moves along a pre-set track or guide rail. It is effective in high-precision inspection tasks, but track-mounted robots are subject to certain limitations in adaptability and flexibility, and are not suitable for scenarios that require frequent changes of position or performing tasks in multiple locations. Figure 1 The middle right picture shows a wheeled inspection robot, which has high flexibility and mobility and can adapt to different environments and tasks. In practical applications, the above two types of robots are usually used according to the application scenario. However, the position of the wheeled robot camera is at a certain height and the zoom range is limited, which will result in poor instrument images, blurry images, tilted angles and other problems.

[0004] A large number of deep learning-based instrument reading recognition methods have emerged in the prior art. These methods mainly use target detection methods to detect the pointer direction to achieve reading recognition of pointer-type instruments. However, these methods still face some challenges and problems: (1) Sensitivity to noise and interference: Noise and interference in the environment may have a negative impact on the performance of deep learning models. For example, the model may be interfered by camera noise, image blur, or instrument damage, resulting in inaccurate readings. (2) The reading error of instrument panels photographed at oblique angles is large: In the actual factory environment, inspection robots are usually used to capture instrument images. This actual situation introduces some challenges. Among them, general deep learning algorithms are difficult to meet the requirements for instrument reading accuracy, especially when the instrument is far away from the inspection robot or there is an angle problem, which further increases the difficulty of correct recognition. Summary of the invention

[0005] The present invention provides a pointer meter reading recognition method and system for a substation inspection robot, aiming to solve the problems existing in the above research background.

[0006] To achieve the above object, the present invention adopts the following technical solution:

[0007] The present invention provides a pointer meter reading recognition method for a substation inspection robot, which is characterized by comprising the following steps:

[0008] Step 1: Obtain inspection images taken by the substation inspection robot;

[0009] Step 2: preprocessing the inspection image obtained in step 1;

[0010] Step 3, converting the inspection image obtained in step 2 with the registered image, where there is a pre-established correspondence between the registered image and the standard model image;

[0011] Step 4: Map the line segments identified on the inspection image to the registration image, and then map the registration image to the standard model image, analyze and identify the pointer, and then read the instrument according to the pointer.

[0012] Furthermore, in the pointer meter reading recognition method of the substation inspection robot provided by the present invention, it can also have the following characteristics: wherein, the preprocessing of the inspection image in step 2 is: adjusting the image to be consistent with the pixel size of the standard model image.

[0013] Furthermore, in the pointer meter reading recognition method of the substation inspection robot provided by the present invention, it can also have the following characteristics: wherein, the standard model image is a front image of each type of pointer meter that is pre-processed and stored, and the information of the standard model image includes the image name representing the pointer meter type, the scale, the axis point of the pointer, the inner radius of the pointer area, and the outer radius of the pointer area.

[0014] Furthermore, in the pointer meter reading recognition method of the substation inspection robot provided by the present invention, it can also have the following characteristics: wherein the standard model image is constructed by the following steps:

[0015] 1) Prepare clear front-facing images of each type of pointer meter;

[0016] 2) Resize all types of frontal images to a uniform pixel size;

[0017] 3) Manually mark the scale of the pointer meter in the image;

[0018] 4) Perform circular fitting according to the coordinate position of the marked scale, the radius of the fitted circle is the outer radius of the pointer area, and the center of the circle is the axis point of the pointer;

[0019] 5) Based on the manually set inner radius and the outer radius of the pointer area obtained by fitting the circle, determine the range of the pointer area.

[0020] Furthermore, in the pointer meter reading recognition method of the substation inspection robot provided by the present invention, it can also have the following characteristics: wherein the registration image in step 3 is constructed by the following steps:

[0021] 1) Batch capture inspection images of various postures as sample images of registration images;

[0022] 2) The feature points between the sample image and the standard model image are screened out through the SIFT feature detector;

[0023] 3) The sample image and the standard model image are matched with feature points by the Brute-Force matcher, and the preferred match is selected by the ratio test. The sample image generates the result image based on the feature points of the preferred match;

[0024] 4) Manually mark the scale of the pointer meter in the result image;

[0025] 5) constructing a transformation matrix between the result image and the standard model image based on the scale of the standard model image and the scale of the result image;

[0026] 6) The resulting image of the transformation matrix between the standard model image and the standard model image is the registered image. The transformation matrix that transforms the standard model image into the registered image is denoted as M1, and the transformation matrix that transforms the registered image into the standard model image is denoted as M2.

[0027] Furthermore, in the pointer meter reading recognition method of the substation inspection robot provided by the present invention, it can also have the following characteristics: wherein step 4 includes the following steps:

[0028] 1) Obtain the registered image corresponding to the current inspection image;

[0029] 2) Using the LSD line segment detection algorithm, all line segments on the inspection image that exceed a preset threshold are identified. These line segments are potential pointer candidates.

[0030] 3) Filter out pointers;

[0031] 4) Calculate the instrument reading by the angle method.

[0032] Furthermore, in the pointer meter reading recognition method of the substation inspection robot provided by the present invention, it can also have the following characteristics: wherein, the specific process of obtaining the registered image is as follows: the corresponding corner points between the inspection image and the corresponding registered image are calculated through the corner point matching algorithm, and the transformation matrix from the inspection image to the registered image is calculated based on these corresponding corner points, and the transformation matrix is ​​denoted as M3.

[0033] Furthermore, in the pointer instrument reading recognition method of the substation inspection robot provided by the present invention, it can also have the following characteristics: wherein, the specific process of screening out the pointer is as follows: using the transformation matrix M3, the identified line segment is mapped to the corresponding registered image; then using the transformation matrix M2 from the registered image to the corresponding standard model image, the identified line segment is mapped to the standard model image; the pointer is selected based on the restrictions of the pointer area and the relationship between the pointer center points.

[0034] Furthermore, in the pointer meter reading recognition method of the substation inspection robot provided by the present invention, it can also have the following characteristics: wherein, the angle method calculates the intersection of the pointer and the scale circle, and obtains the central angle between the intersection and the two adjacent scale points A and B on the scale circle, thereby obtaining the proportion of the scale grid, and the mathematical expression is:

[0035]

[0036] Where r is the instrument reading; θ left The left end point of the scale interval pointed by the instrument pointer; θ right The right end point of the scale interval pointed by the instrument pointer; A, B and the axis point form the center angle of the circle; r right 、r leftare the instrument readings corresponding to points A and B respectively; r right -r left is the instrument graduation value.

[0037] The present invention also provides a pointer meter reading recognition system for a substation inspection robot, which is characterized by comprising: an image preprocessing module, used to automatically adjust the current inspection image to be consistent with the size pixels of the standard model image; a first image matching module, used to transform the standard model image into a registered image; a second image matching module, used to transform the registered image into the standard model image; a third image matching module, used to obtain the registered image based on the inspection image; a line segment recognition module, running the LSD line segment detection algorithm to identify all line segments in the image that exceed a preset threshold; and an instrument reading module, running the angle method to calculate the pointer meter reading.

[0038] Functions and effects of the present invention:

[0039] The pointer meter reading recognition method of the substation inspection robot provided by the present invention adopts the creation of a standard model, realizes the matching of the standard model with the actual captured image to establish a corresponding relationship, improves the accuracy of pointer meter recognition and analysis, recognizes all line segments on the inspection map that exceed the preset threshold through the LSD line segment detection algorithm, and then maps these line segments to the corresponding registration map through the conversion matrix, and finally maps them to the standard model map, combines the restrictions of the pointer area and the relationship between the pointer center point to accurately screen, select the real pointer, and then realize efficient and accurate recognition and analysis of the pointer meter by reading the scale value pointed to by the pointer. The method of the present invention can efficiently and accurately identify and analyze the pointer meter in a complex environment, and improve the accuracy of robot inspection. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 These are real-life photos of inspection robots collecting instrument images. The left picture is a track-mounted robot, and the right picture is a wheeled inspection robot.

[0041] Figure 2 It is a step diagram of a method for identifying pointer meter readings of a substation inspection robot in an embodiment of the present invention.

[0042] Figure 3 It is a front-view image used to construct a standard model image in an embodiment of the present invention.

[0043] Figure 4 It is a preprocessed image in the embodiment of the present invention.

[0044] Figure 5 It is an image of a drawing rectangle in the embodiment of the present invention.

[0045] Figure 6is an image of a dot-marked scale in an embodiment of the present invention.

[0046] Figure 7 It is an image of a fitted circle in the embodiment of the present invention.

[0047] Figure 8 is a schematic diagram of a pointer area in an embodiment of the present invention.

[0048] Fig. 9 It is a standard model image in the embodiment of the present invention.

[0049] Fig.10 It is the inspection image in the embodiment of the present invention.

[0050] Fig.11 It is an inspection image of the dot-marked scale in the embodiment of the present invention.

[0051] Fig.12 Schematic diagram of the angle of the standard model image in the embodiment of the present invention.

[0052] Fig.13 3 is a schematic diagram of the angle of the image after correction in an embodiment of the present invention.

[0053] Fig.14 Schematic diagram of SIFT corner point matching between an inspection image and a registration image in an embodiment of the present invention.

[0054] Fig.15 This is the case where the identification line segment of the inspection image in the embodiment of the present invention is mapped to the pointer area of ​​the standard image.

[0055] Fig.16 This is the case where a real pointer is identified in the embodiment of the present invention.

[0056] Fig.17 It is a schematic diagram of the principle of calculating readings using the angle method in an embodiment of the present invention.

[0057] In the accompanying drawings, (a) is a pointer table of the first type, and (b) is a pointer table of the second type. DETAILED DESCRIPTION

[0058] In order to make the technical means, creative features, objectives and effects achieved by the present invention easy to understand, the following embodiments are combined with the accompanying drawings to specifically illustrate the technical solutions of the present invention.

[0059] See also Figure 2 The embodiment of the present invention provides a method for identifying pointer meter readings of a substation inspection robot, which includes the following steps:

[0060] Step 1: Obtain inspection images taken by the substation inspection robot.

[0061] The posture images of the inspection points taken by the robot are not necessarily front-facing images. They may show various postures. For inspection images, see Fig.10 .

[0062] Step 2: preprocess the inspection image obtained in step 1: adjust the image to be consistent with the pixel size of the standard model image.

[0063] The standard model image is a front image of each type of pointer meter that is pre-processed and stored. These standard model images and their related information (including the image name representing the pointer meter type, scale, the axis point of the pointer, the inner radius of the pointer area, the outer radius of the pointer area, etc.) are stored.

[0064] The Standard Model image is constructed using the following steps:

[0065] 1) Prepare a clear front view of each type of pointer meter, see Figure 3 .

[0066] In this embodiment, these are stored in the "models / images" folder, and the scale values ​​of each type of pointer model are saved in the "models / parameters" folder, stored in txt text format, ensuring that the file name matches the corresponding image name. When modifying the program, it is necessary to ensure that the model modelIds is consistent with the name of the corresponding standard model image. For example, if the standard model image name is 'dialModel-1003.jpg', modelIds should be set to ['dialModel-1003']. In addition, the inner diameter range of the pointer is determined by manually adjusting the minRadius parameter.

[0067] 2) All types of frontal images are resized to a uniform pixel size.

[0068] 3) In this embodiment, the standard model image is a square image with a side length of 500 pixels, see Figure 4 Processing the current inspection image, the captured image as a sample, and the captured image as a standard model into pixels of uniform size can ensure the consistency and comparability of the images, as well as the stability of the program in different scenarios, and improve the recognition accuracy of the model.

[0069] 4) Manually mark the scale of the pointer meter in the image.

[0070] Draw a blue rectangle at the scale position in the image, such as Figure 5As shown. The center of the rectangle is used as the center of the scale. After the program is set to draw the rectangle, the operator presses the number key "1" and the rectangle will instantly turn into a striking red dot mark. This setting allows the operator to mark and identify more intuitively and conveniently. Figure 6 As shown in FIG. 1 , the dot mark is the corresponding scale mark.

[0071] 4) Perform circular fitting according to the coordinate positions of the marked scales.

[0072] Use the circle fitting program to fit a circle using the coordinate position of the marked scale. The radius of the fitted circle is the outer radius of the pointer area, and the center of the circle is the axis point of the pointer. Figure 7 After completing the scale calibration and circle fitting, the system will automatically save the scale point coordinates, the coordinates of the fitted circle center and the circle radius and other related parameters in the bin file in the models / parameters folder. The name is consistent with the txt file name for subsequent use and calling.

[0073] 5) Determine the scope of the pointer area.

[0074] Determine the range of the pointer area based on the manually set inner radius and the outer radius and pointer pivot point of the pointer area obtained by fitting the circle, see Figure 8 . Clarifying the pointer area can provide an accurate pointer limit range for subsequent detection operations, thereby effectively improving the accuracy of pointer recognition and making the reading more accurate.

[0075] Step three, converting the inspection image obtained in step two with the registered image, there is a pre-established correspondence between the registered image and the standard model image.

[0076] The registration image is constructed by the following steps:

[0077] 1) For each type of pointer table, repeatedly capture inspection images of various postures as sample images of the registration image.

[0078] 2) Filter out the feature points between the sample image and the standard model image.

[0079] Fig. 9 is the standard model image, Fig.10 It is the inspection image taken at the inspection point. The key points are detected and descriptors are calculated on the standard model image and the inspection image through the SIFT feature detector. According to the given mask, the key points that meet the conditions in the image are screened out.

[0080] 3) The sample image is matched with the standard model image through the Brute-Force matcher, and the preferred match is screened out through the ratio test. The sample image generates the result image based on the feature points of the preferred match.

[0081] Through the above image matching process, cv2.findHomography function is used to calculate the transformation matrix from the standard model image to the registered image and from the registered image to the standard model image using the RANSAC algorithm. The matrix describes how to convert the standard model image into the pointer image of the inspection point posture and how to convert the image of the inspection point posture into the standard model image. Use the drawMatches function to draw the matching key points on the sample image and the standard model image to generate the result image.

[0082] 4) Manually mark the scale of the pointer meter in the result image.

[0083] This step is similar to the steps of manually marking the scale of the pointer table in constructing a standard model image. Similarly, first draw a blue rectangle, which is displayed in real time on the result image. The size and position of the rectangle can be adjusted in real time so that the user can select the scale point on the registration image. Then, the operator presses the number key 1, and the blue rectangle turns into a red dot. Fig.11 .

[0084] The purpose of this step is to establish a correspondence between the standard model image and the registered image. After completing the scale point marking, the scale points can be used to calculate the transformation matrix between the two.

[0085] 5) Based on the scale of the standard model image and the scale of the result image, a transformation matrix between the result image and the standard model image is constructed. The result image with the transformation matrix between the standard model image and the result image is the registration image. The transformation matrix from the standard model image to the registration image is denoted as M1. The transformation matrix from the registration image to the standard model image is denoted as M2.

[0086] About the correction angle: When the three o'clock direction is zero degrees in the instrument of the standard model image, Fig.12 As shown in the figure, point A can represent both -π degrees and π degrees, but most instrument scales are from point B to point C, which makes the angle at point A not continuous, resulting in inaccurate readings. When it is corrected to 0° at 12 o'clock, as shown in the figure below: Fig.13As shown, there is only one angle at point A, and the discontinuity is at point D, but most instrument scales are from point B to point C and do not pass through point D, so there is no effect on the pointer reading. Through such corrections, the angle representation can be made more continuous and natural throughout the scale range, and most center-mark and pointer instruments are designed with 12 o'clock as the starting point. This correction helps to keep consistent with the pointer scale on the instrument, making the angle representation more in line with our intuitive understanding of the instrument when simulating the instrument.

[0087] Step 4: Map the line segments identified on the inspection image to the registration image, and then map the registration image to the standard model image, analyze and identify the pointer, and then read the instrument according to the pointer.

[0088] Step 4 specifically includes the following steps:

[0089] 5) Obtain the registered image corresponding to the current inspection image.

[0090] See also Fig.14 , the corresponding corner points between the inspection image and the corresponding registered image are calculated through the SIFT corner point matching algorithm, and the transformation matrix from the inspection image to the registered image is calculated based on these corresponding corner points. The transformation matrix is ​​recorded as M3.

[0091] 6) See Fig.15 ,The LSD line segment detection algorithm is used to identify all line segments on the inspection image that exceed the preset threshold and these line segments are potential pointer candidates.

[0092] 7) Filter out the pointers.

[0093] The identified line segments are mapped to the corresponding registered image using the conversion matrix M3 from the inspection image to the registered image, and then mapped to the standard model image using the conversion matrix M2 from the registered image to the corresponding standard model image.

[0094] The real pointer is selected based on the pointer area restrictions and the relationship between the pointer center points, such as Fig.16 shown.

[0095] 8) Calculate the meter reading by the angle method.

[0096] The angle method is to calculate the intersection of the pointer and the scale circle, and find the central angle between the intersection and the two adjacent scale points A and B on the scale circle, so as to obtain the proportion of the scale grid. The specific expression is:

[0097]

[0098] Where r is the instrument reading; θ left The left end point of the scale interval pointed by the instrument pointer; θright The right end point of the scale interval pointed by the instrument pointer; A, B and the axis point form the center angle of the circle; r right 、r left are the instrument readings corresponding to points A and B respectively; r right -r left is the instrument graduation value.

[0099] Finally, the scale value corresponding to the pointer is read. This process of the present invention ensures that the pointer can still be accurately identified in a complex environment, thereby improving the accuracy and efficiency of the overall inspection.

[0100] The embodiment of the present invention also provides a pointer meter reading recognition system for a substation inspection robot, in which a corresponding operating program is provided for executing the pointer meter reading recognition method for the substation inspection robot. The system includes an image preprocessing module, a first image matching module, a second image matching module, a third image matching module, a line segment recognition module, and a meter reading module.

[0101] The functions of each module in the pointer meter reading recognition system of the substation inspection robot are as follows:

[0102] The image preprocessing module is used to automatically adjust the current inspection image to be consistent with the pixel size of the standard model image.

[0103] The first image matching module transforms the standard model image into a registered image through the M1 transformation matrix.

[0104] The second image matching module transforms the registered image into a standard model image through the M2 transformation matrix.

[0105] The third image matching module calculates the corresponding corner points between the inspection image and the corresponding registered image through the SIFT corner point matching algorithm to obtain the registered image corresponding to the inspection image and construct the M3 conversion matrix.

[0106] The line segment recognition module runs the LSD line segment detection algorithm to identify all line segments in the image that exceed a preset threshold.

[0107] The instrument reading module operates the angle method to calculate the pointer instrument reading.

[0108] The above embodiments are only exemplary preferred embodiments of the present invention and are not intended to limit the protection scope of the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention are included in the protection scope of the present invention.

Claims

1. A method for identifying pointer meter readings of a substation inspection robot, characterized in that: The following steps are involved: Step 1: Obtain inspection images taken by the substation inspection robot; Step 2: preprocessing the inspection image obtained in step 1; Step 3, converting the inspection image obtained in step 2 with the registered image, wherein there is a pre-established corresponding relationship between the registered image and the standard model image; Step 4: Map the line segments identified on the inspection image to the registration image, and then map the registration image to the standard model image, analyze and identify the pointer, and then read the instrument according to the pointer.

2. The method for identifying pointer meter readings of a substation inspection robot according to claim 1, characterized in that: in, The preprocessing of the inspection image in step 2 is: adjusting the image to be consistent with the pixel size of the standard model image.

3. The method for identifying pointer meter readings of a substation inspection robot according to claim 2, characterized in that: in, The standard model image is a front image of each type of pointer meter that is pre-processed and stored, and the information of the standard model image includes an image name representing the pointer meter type, a scale, a center point of the pointer, an inner radius of the pointer area, and an outer radius of the pointer area.

4. The method for identifying pointer meter readings of a substation inspection robot according to claim 3, characterized in that: in, The standard model image is constructed by the following steps: 1) Prepare clear front-facing images of each type of pointer meter; 2) Resize all types of frontal images to a uniform pixel size; 3) Manually mark the scale of the pointer meter in the image; 4) Perform circular fitting according to the coordinate position of the marked scale, the radius of the fitted circle is the outer radius of the pointer area, and the center of the circle is the axis point of the pointer; 5) Based on the manually set inner radius and the outer radius of the pointer area obtained by fitting the circle, determine the range of the pointer area.

5. The pointer meter reading recognition method of the substation inspection robot according to claim 1, Features: The registered image in step 3 is constructed by the following steps: 1) Batch capture inspection images of various postures as sample images of registration images; 2) filtering out feature points between the sample image and the standard model image through a SIFT feature detector; 3) The sample image and the standard model image are matched with feature points by the Brute-Force matcher, and the preferred match is selected by the ratio test. The sample image generates the result image based on the feature points of the preferred match; 4) Manually mark the scale of the pointer meter in the result image; 5) constructing a transformation matrix between the result image and the standard model image based on the scale of the standard model image and the scale of the result image; 6) The resulting image of the transformation matrix between the standard model image and the standard model image is the registered image. The transformation matrix for transforming the standard model image into the registered image is denoted as M1, and the transformation matrix for transforming the registered image into the standard model image is denoted as M2.

6. The pointer meter reading recognition method of the substation inspection robot according to claim 1, Features: Wherein, step 4 includes the following steps: 1) Obtain the registered image corresponding to the current inspection image; 2) Using the LSD line segment detection algorithm, all line segments on the inspection image that exceed a preset threshold are identified. These line segments are potential pointer candidates. 3) Filter out pointers; 4) Calculate the instrument reading by the angle method.

7. The method for identifying pointer meter readings of a substation inspection robot according to claim 6, characterized in that: in, The specific process of obtaining the registered image is as follows: the corresponding corner points between the inspection image and the corresponding registered image are calculated through the corner point matching algorithm, and the transformation matrix from the inspection image to the registered image is calculated based on these corresponding corner points. The transformation matrix is ​​recorded as M3.

8. The method for identifying pointer meter readings of a substation inspection robot according to claim 7, characterized in that: in, The specific process of screening out the pointer is as follows: using the transformation matrix M3, the identified line segments are mapped to the corresponding registered image; then using the transformation matrix M2 from the registered image to the corresponding standard model image, the identified line segments are mapped to the standard model image; the pointer is selected based on the restrictions of the pointer area and the relationship between the pointer center points.

9. The method for identifying pointer meter readings of a substation inspection robot according to claim 6, characterized in that: in, The angle method calculates the intersection of the pointer and the scale circle, and obtains the central angle between the intersection and the two adjacent scale points A and B on the scale circle, thereby obtaining the proportion of the scale grid. The mathematical expression is: Where r is the instrument reading; θ left The left end point of the scale interval pointed by the instrument pointer; θ right The right end point of the scale interval pointed by the instrument pointer; A, B and the axis point form the center angle of the circle; r right 、r left are the instrument readings corresponding to points A and B respectively; r right -r left is the instrument graduation value.

10. A pointer meter reading recognition system for a substation inspection robot, used to execute the pointer meter reading recognition method for a substation inspection robot as claimed in any one of claims 1 to 9, characterized in that it comprises: An image preprocessing module is used to automatically adjust the current inspection image to be consistent with the size pixels of the standard model image; A first image matching module, used to transform the standard model image into a registered image; A second image matching module, used for transforming the registered image into a standard model image; A third image matching module is used to obtain a registration image based on the inspection image; The line segment recognition module runs the LSD line segment detection algorithm to identify all line segments in the image that exceed a preset threshold; The instrument reading module uses the operating angle method to calculate the pointer instrument reading.

Images