A method of dual pointer detection and read calculation

Abstract

The application provides a double-pointer detection and reading calculation method, and solves the technical problems that the existing double-pointer meter automatic reading scheme relies on color marking, is weak in color interference resistance, has poor cross-table phenotype adaptation capability, and does not utilize bottom map prior feature technology. It comprises collecting a dial image of a double-pointer, calibrating and predicting geometric key points of the pointer in the dial image; extracting pointer line segment pixels based on the geometric key points, and converting to obtain color features; extracting bottom map prior information and calculating reference color features, and performing weighted matching with the color features to determine the pointer attribute, combining the scale information in the bottom map prior information, and adopting a line segment-ray intersection point proportion method to calculate the pointer reading. The application can be widely applied to the technical field of image classification detection.

Description

Technical Field

[0001] This application belongs to the field of image classification and detection technology, and more specifically, it relates to a method for dual-pointer detection and reading calculation. Background Technology

[0002] Pressure gauges, oil temperature gauges, and other dual-pointer instruments are key components for monitoring the operational status of power grid equipment, and the accuracy of their readings directly affects the safety of the power system. To replace inefficient and error-prone manual inspections, computer vision-based automatic reading technology has become the mainstream.

[0003] Current mainstream methods for automatic reading of dual-pointer meters generally adopt a detection logic that binds key points of the pointer to color categories. That is, during the model training phase, key points of the pointer are simultaneously labeled with their corresponding color categories. During inference, the function of different pointers is distinguished by predicting the pointer color. Moreover, most of these are customized solutions for each meter, which do not make full use of the features of the meter's base map. This approach has three problems in industrial deployment: First, it is highly dependent on full color labeling. Industrial meters have diverse colors, making it difficult to collect all colors, resulting in high system adaptation costs and low efficiency. Second, it is prone to misjudgment in scenarios with faded or distorted colors and cannot self-check. Distorted colors are forcibly classified as known colors, leading to misjudgment of pointer functions. Third, the prediction results are disconnected from the prior color information of the meter's base map, making it impossible to use known information to correct errors, resulting in low reliability of the prediction results.

[0004] In summary, existing methods for automatic reading of dual-pointer meters suffer from limitations in customization, data collection bottlenecks, weak resistance to color interference, and disconnect from prior knowledge, making it difficult to meet the high reliability and generalization requirements of industrial applications. Summary of the Invention

[0005] The purpose of this application is to provide a method for dual-pointer detection and reading calculation, so as to solve the technical problems of existing dual-pointer meter automatic reading methods that rely on color marking, have weak resistance to color interference, poor cross-phenotype adaptation ability, and do not utilize the prior features of the base map.

[0006] To achieve the above objectives, this application provides a method for dual-pointer detection and reading calculation, comprising the following steps: Acquire a dial image with two hands, and calibrate and predict the geometric key points of the hands in the dial image; The pointer line segment pixels are extracted based on geometric key points and then converted to obtain color features. Extract prior information from the base map and calculate the baseline color features. Perform weighted matching with the color features to determine the pointer attributes. Combine the scale information in the prior information of the base map and use the line segment-ray intersection ratio method to calculate the pointer reading.

[0007] Preferably, the step of calibrating and predicting the geometric key points of the pointers in the dial image includes: calibrating four key points in a fixed order, including the left pointer endpoint, the right pointer endpoint, the right pointer start point, and the left pointer start point; The YOLO model is trained by focusing on four key points to obtain a key point prediction model, which is then used to predict the geometric key points of the dial image.

[0008] Preferably, the process of extracting pointer line segment pixels based on geometric key points includes: constructing a rectangular ROI for the left and right pointers based on the geometric key points, fitting pointer line segments using the start and end points of the rectangular ROI, sampling along the fitted pointer line segments with a set step size, and obtaining pointer line segment pixels.

[0009] Preferably, the formula for obtaining the fitted pointer line segment is: ; In the formula, To fit the x-coordinate pixel value of any sampling point on the pointer line segment, To fit the ordinate pixel value of any sampling point on the pointer line segment, This is the line segment sampling ratio coefficient, with a value range of [0,1], used for uniform sampling along the pointer line segment. , ) represents the pixel coordinates of the starting point of the pointer line segment. , ) represents the pixel coordinates of the endpoint of the pointer line segment.

[0010] Preferably, the process of calculating the pointer reading using the line segment-ray intersection ratio method includes: analyzing the pixel coordinates corresponding to the scale values ​​in the prior information of the base map, sorting them, and then forming line segments from adjacent scale values ​​in sequence; Generate a pointer ray with the pointer's starting point as the origin and the ending point as the direction. Calculate the intersection points of the pointer ray and the line segments. Locate the scale line segment with a unique valid intersection point and calculate the pointer reading using proportional interpolation.

[0011] Preferably, the formula for obtaining the pointer reading is: ; In the formula, This is the final physical reading of the position the pointer is pointing to. The scale value corresponding to the starting point of the valid scale segment. d represents the scale value corresponding to the endpoint of the valid scale segment. p d is the pixel distance from the intersection of the pointer ray and the tick mark segment to the starting point of the tick mark segment. L The total pixel length of the effective scale line segment itself.

[0012] Preferably, the process of converting to obtain color features includes: normalizing the RGB values ​​of the pointer line segment pixels to obtain normalized RGB values, performing extreme value calculations to obtain the maximum, minimum, and color difference of the RGB values, calculating hue, saturation, and brightness based on the maximum, minimum, and color difference, and taking the median of hue, saturation, and brightness to obtain color features.

[0013] Preferably, the process of calculating the reference color features includes: calculating the hue, saturation, and brightness of the pointer pixels within the pointer area based on the pointer area marked in the prior information of the base map, and taking the median of each to obtain the reference color features.

[0014] Preferably, the process of determining pointer attributes includes: calculating the sum of the weighted absolute values ​​of the differences between the color features and the reference color features as a similarity index; matching the dial image with the base image pointer based on the similarity index; and determining the pointer attributes of the dial image.

[0015] Preferably, the formula for the similarity index is: ; In the formula, As a similarity metric, The median hue of the pixel in the pointer line segment to be measured. The median hue of the base pointer of the base map. The median saturation value of the pixel segment of the pointer line to be tested. The median saturation of the base pointer of the base map. The median value of the pixel brightness of the pointer line segment to be measured. The median value of the base pointer in the background map. These are the weighting coefficients for hue, saturation, and brightness, respectively.

[0016] The beneficial effects of this application are as follows: This application provides a method for dual-pointer detection and reading calculation. First, by calibrating and predicting the geometric key points of the pointer in the dial image, pure geometric key point positioning is achieved, eliminating the dependence on pointer color in traditional techniques, reducing the data collection cost brought by full color annotation, and improving the adaptability across phenotypic scales. Next, pointer line segment pixels are extracted based on geometric key points and converted into color features, enhancing the stability of features in color distortion scenarios and improving anti-interference capabilities. Subsequently, weighted matching is performed using prior information from the base map to determine pointer attributes. This process does not require training a color classification model, fully utilizing prior information from the base map to avoid color misjudgment. Finally, combined with the scale information of the base map, the reading is calculated using the line segment-ray intersection ratio method. High-precision ratio interpolation calculation is used to overcome the customization limitations of one meter per device, achieving high-precision, high-robust, and high-generalization automatic reading of dual-pointer meters. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of this application, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 A flowchart illustrating a method for dual-pointer detection and reading calculation provided in an embodiment of this application; Figure 2 A schematic diagram of the overall architecture of a dual-pointer detection and reading calculation method provided in an embodiment of this application; Figure 3 A clockwise calibration diagram of geometric key points provided in an embodiment of this application; Figure 4 This is a calibration diagram of a rectangular frame A provided in an embodiment of this application; Figure 5 A calibration diagram of a rectangular frame B provided in an embodiment of this application; Figure 6 A color matching flowchart provided for one embodiment of this application. Detailed Implementation

[0019] To make the technical problems, technical solutions, and beneficial effects to be solved by this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and are not intended to limit the scope of this application.

[0020] This application constructs a technical solution for clockwise fixed-order keypoint calibration, accurate keypoint prediction, keypoint-based detection box generation, pointer line segment pixel statistics, background map prior color matching, and reading calculation. This solution uses geometric features as its core, combined with background map prior information, to eliminate dependence on pointer color. Through unified calibration and detection logic, it adapts to various dual-pointer meters, solving the problem of misjudgment in scenarios of color fading and distortion, reducing the limitations of customized solutions like those for individual meters, and fully utilizing background map prior information to correlate prediction results with background map information, thereby improving the reliability of detection results and the model's generalization ability.

[0021] Please see Figure 1-2 A method for dual-pointer detection and reading calculation, provided in one embodiment of this application, includes: S1: Acquire a dial image of two hands, calibrate and predict the geometric key points of the hands in the dial image.

[0022] The system uses cameras to capture images of the dials of dual-pointer meters in substation high-voltage equipment areas, distribution cabinets, transformers, and other similar locations. The cameras used have a resolution of at least 1920×1080 pixels, support autofocus, 1ms to 100ms dynamic exposure adjustment and white balance calibration, and have a lens focal length suitable for installation distances from 50cm to 3m to ensure clear capture of the pointer endpoints.

[0023] In an optional embodiment, after obtaining the dial image with two hands, it is preprocessed, including sharpness assessment and occlusion detection, and images that are blurry, have more than 30% hand occlusion, or have a severely damaged dial are discarded and stored in an edge computing device.

[0024] Please see Figure 3 This application calibrates the geometric key points of the pointers in a dial image. Unlike existing complex schemes that simultaneously annotate key points and colors, this application employs a purely geometric feature-guided calibration method, completely eliminating dependence on color. After acquiring a clear, unobstructed, high-resolution dial image of a dual-pointer meter, the user annotates four key endpoints in a fixed clockwise order, without needing to consider color. Specifically: Key point 1: The left pointer is at the end furthest from the center of rotation (the end point, the outermost point). Key point 2: The end of the pointer on the right that is furthest from the center of rotation (the endpoint); Key point 3: The right hand is closer to the center of rotation (starting point, near the center of the dial). Key point 4: The left pointer is at the end closest to the center of rotation (the starting point).

[0025] This application uses the pointer space geometry and endpoint type as the sole calibration basis, avoiding color judgment errors, eliminating meter type limitations, and standardizing the calibration process. The system records the pixel coordinates of each key point. The data is stored as a JSON file for later use, transforming the color-pointer matching problem into a single deterministic geometric calibration task, thus reducing the cognitive burden on humans.

[0026] After the calibration of the dial's geometric key points is completed, the YOLO model is used to focus on learning the common geometric endpoint features (key points 1 to 4) of dual-pointer meters, without needing to learn color features and phenotypic differences, thereby improving the accuracy, anti-interference ability and generalization ability of the predicted key points.

[0027] Specifically, in an optional embodiment, this application uses the YOLOv11s-pose estimation model for training. The input image resolution is set to 512×512, with 200 training rounds and approximately 5000 training samples, covering various types of meters in substations, including substation meter images with different lighting and levels of contamination. The training process uses the data augmentation method built into the YOLO framework to expand sample diversity. During the model inference phase, the keypoint confidence threshold is set to 0.1 to ensure the recall rate of target keypoint detection, predicting the geometric keypoints of the pointer in the dial image. Since the model uses pure geometric features as the core of learning throughout and does not rely on color information constraints, the keypoint prediction accuracy is significantly better than traditional recognition models bound to color features under complex conditions such as drastic changes in lighting and image color distortion. At the same time, it has good cross-model adaptability and can be directly compatible with dual-pointer meters of various structures without the need for separate parameter tuning or retraining for specific meters.

[0028] S2: Extract pointer line segment pixels based on geometric key points and convert them to obtain color features.

[0029] This application adopts a step-by-step process of first geometric positioning and then pixel extraction to solve the error in pointer color recognition caused by color distortion and local dirt in the dial image.

[0030] Specifically, firstly, based on the coordinates of four key points predicted by the model, axis-aligned matrices are constructed for the left and right pointers, creating regions of interest (ROIs) to isolate the pointers from the background. For example... Figure 4-5 As shown: Rectangle A (ROI1) constructs a minimum bounding rectangle with keypoints 1 and 4 as diagonal vertices, covering the left pointer path; Rectangle B (ROI2) constructs a minimum bounding rectangle with keypoints 2 and 3 as diagonal vertices, covering the right pointer path.

[0031] Subsequently, based on the aforementioned rectangular region, the pixel information of the pointer line segment is extracted. The specific steps are as follows: First, fit the pointer line segments on the left (key points 1 and 4) and the right (key points 2 and 3) respectively. The parametric equations of the line segments are as follows: ; In the formula, To fit the x-coordinate pixel value of any sampling point on the pointer line segment, To fit the ordinate pixel value of any sampling point on the pointer line segment, This is the line segment sampling ratio coefficient, with a value range of [0,1], used for uniform sampling along the pointer line segment. , ) represents the pixel coordinates of the starting point of the pointer line segment. , ) represents the pixel coordinates of the endpoint of the pointer line segment.

[0032] Secondly, sample along the fitted line segment with a step size of 1 pixel to obtain the set of pixels. , For the first Coordinates of each sampling point This represents the total number of sampling points.

[0033] After obtaining the pointer line segment pixels, the RGB values ​​of each sampling point of the pointer line segment pixels are read and converted into HSV values. The core processes include normalization, extreme value calculation, and hue calculation. saturation and brightness The calculation is as follows: First, the RGB values ​​of each pixel are normalized using the following formula:

[0034] In the formula, The normalized original value of the red channel. This is the original value of the red channel of the pixel. The normalized green channel value. This is the original value of the green channel of the pixel. The normalized blue channel value. This is the original value of the blue channel for each pixel.

[0035] Next, extreme value calculation is performed, using the following formula: ; In the formula, for , , The maximum value among the three for , , The minimum of the three, This is due to color difference.

[0036] Next, calculate the hue. saturation and brightness Among them, hue The calculation formula is: ; Saturation The calculation formula is: ; Brightness The calculation formula is: ; in, , .

[0037] Finally, to filter out errors caused by localized dirt and noise interference, the median of the HSV features of all sampling points of the pointer line segment is calculated as the final color feature of the pointer. The color characteristics of the left pointer are obtained. Color characteristics of the right pointer This completes the extraction of pointer color features from geometric key points.

[0038] S3: Extract prior information from the base map and calculate the baseline color features. Perform weighted matching with the color features to determine the pointer attributes. Combine the scale information in the prior information of the base map and use the line segment-ray intersection ratio method to calculate the pointer reading.

[0039] The system retrieves prior information from the base image of the corresponding dual-pointer meter, including scale parameters, pointer area annotations, and pointer color combinations. It calculates the baseline color features and performs a weighted matching with the color features of the aforementioned dial image (new image) to complete color determination. This eliminates reliance on color classification models and improves matching robustness in complex lighting scenarios. The specific process is as follows: Figure 6 As shown: First, the base map configuration file for the corresponding dual-pointer meter is retrieved. Prior information is read from the base map configuration file to obtain the two pre-labeled pointer regions and their corresponding pointer color attribute labels (first color, second color). The median HSV feature of the pointer pixels within the two pointer regions is calculated to obtain the baseline color features of the base map pointer. Among these, the baseline features of pointer 1... The base features of pointer 2 This baseline feature will serve as an interference-free reference threshold for matching the pointer color of the dial image.

[0040] Then, based on the color characteristics of the left pointer in the aforementioned dial image... Color characteristics of the right pointer Then, the sum of the absolute values ​​of the HSV weighted differences is used as the similarity index. The formula is: ; In the formula, The median hue of the pixel in the pointer line segment to be measured. The median hue of the base pointer of the base map. The median saturation value of the pixel segment of the pointer line to be tested. The median saturation of the base pointer of the base map. The median value of the pixel brightness of the pointer line segment to be measured. The median brightness of the base map's reference pointer is set to a weighting factor of 1. Calculate four sets of matching differences between the left and right pointers of the dial image and two baseline features of the background image. The matching of the dial image and the base map pointer is completed by following the principle of minimizing the total difference.

[0041] After matching is complete, the color attribute label corresponding to the base image pointer is assigned to the successfully matched pointer in the dial image, and a similarity threshold is set. If the difference between any matching pairs is greater than the similarity threshold If a matching conflict occurs, a manual verification alarm will be triggered, indicating that the field pointer is severely damaged or obstructed, or that the bottom label does not match the actual meter reading.

[0042] If rectangle A matches the first color attribute label (e.g., black), then the endpoint of the color pointer is keypoint 1, and the starting point is keypoint 4; if rectangle B matches the second color (e.g., red), then the endpoint of the color pointer is keypoint 2, and the starting point is keypoint 3; if there are cases where the color assignments are reversed (e.g., A is red, B is black), then the start and end points are automatically adjusted according to the actual physical location to ensure that each color pointer is consistent with its actual physical location. This generates a structured pointer information table, as shown in Table 1: Table 1: Structured Pointer Information Table

[0043] The above information, along with the pre-stored dial center coordinates, The data is then transmitted to the reading calculation module, which uses the scale segment-ray intersection ratio method to accurately calculate the physical reading. The specific process and formula are as follows: Pre-parse the pixel coordinates corresponding to all tick values ​​on the dial in the base map configuration file, and record the tick values. (like ), corresponding pixel coordinates are Sort all scale values ​​in ascending order to obtain an ordered scale set. Its corresponding pixel coordinate set is .

[0044] Next, the pixel coordinates corresponding to the sorted adjacent tick values ​​are sequentially arranged into line segments, the th... line segment From the ordered scale set The and the first It consists of several scale points, and the calculation formula is as follows: ; Then, using the coordinates of the pointer's starting point... The origin of the ray and the coordinates of the pointer's endpoint are given. Generate an infinitely extending pointer ray along the ray's direction. Its parametric equation is as follows: ; Calculate the rays sequentially With each scale segment The intersection of the ray and a certain line segment There exists a unique valid intersection point. (The intersection point is within the line segment and) If the intersection point calculation for subsequent line segments is stopped, the line segment is locked. A valid scale segment is selected; if no valid intersection is found among all segments, the reading is considered invalid (triggering manual review).

[0045] Finally, define the effective scale segments. The starting scale value is The endpoint scale value is First, calculate the intersection point. to line segment starting point pixel distance and line segments Its own pixel length As shown in the following formula: ; ; In the formula, This represents the x-coordinate pixel value of the intersection point of the pointer ray and the effective scale line segment. The x-coordinate pixel value corresponding to the starting point of the effective scale line segment. The vertical pixel value of the intersection point of the pointer ray and the effective scale line segment. The ordinate pixel value corresponding to the starting point of the effective scale line segment. The x-coordinate pixel value corresponding to the endpoint of the valid scale line segment. This represents the ordinate pixel value corresponding to the endpoint of the valid scale line segment.

[0046] Then, the actual physical reading of the pointer is calculated using proportional interpolation. The calculation formula is shown below: ; In the formula, This refers to the actual physical reading of the pointer. The starting point of the effective scale segment is the scale value. This is the end-point scale value of the valid scale segment.

[0047] Example 1: The test was conducted using a 5-megapixel industrial camera and processed by an NVIDIA Jetson Xavier edge computing device. 800 images covering three types of meters, different lighting conditions, and varying degrees of color distortion were selected. The test results are shown in Table 2.

[0048] Table 2: Comprehensive Test Results of Three Different Meters

[0049] The results show that the accuracy of key point prediction in this application is 9.4% higher than that of the traditional solution, and the accuracy of color matching is 10.3% higher. The accuracy of adaptation for the three meter types is over 99%, while the traditional solution requires separate model training, with an accuracy of 89.2% (pressure gauge), 90.5% (ammeter), and 87.8% (oil temperature gauge), and requires redeployment when switching meter types. Even in complex scenarios such as strong light and fading, the color matching accuracy of this solution is still greater than 95.0%, while the traditional RGB difference solution drops to 82.3%.

[0050] The average inference time for the rectangle generation, pixel extraction, and color matching modules is less than 15ms. In a test running continuously for a week in a real-world scenario, processing 3680 images, the effective reading accuracy reached 98.7%, with a single image processing time of less than 200ms. Even in 1200 images with complex lighting and color distortion, the effective reading accuracy was still 97.5%, compared to only 86.3% for traditional solutions. This application does not require adjusting parameters or retraining the model for different meters, nor does it require collecting all pointer color samples. When switching phenotypic models, high-precision detection can be achieved simply by reusing the existing model and retrieving the corresponding base image, fully verifying the superiority and practicality of the innovative design.

[0051] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0052] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application, and should all be included within the protection scope of this application.

Claims

1. A method for dual-pointer detection and reading calculation, characterized in that, Includes the following steps: Acquire a dial image with two hands, and calibrate and predict the geometric key points of the hands in the dial image; Based on the geometric key points, pointer line segment pixels are extracted and converted to obtain color features; Extract prior information from the base map and calculate the baseline color features. Perform weighted matching with the color features to determine the pointer attributes. Combine the scale information in the prior information of the base map and use the line segment-ray intersection ratio method to calculate the pointer reading.

2. The method for dual-pointer detection and reading calculation as described in claim 1, characterized in that, The steps of calibrating and predicting the geometric key points of the pointers in the dial image include: calibrating four key points in a fixed order, including the left pointer endpoint, the right pointer endpoint, the right pointer start point, and the left pointer start point; The YOLO model is trained by focusing on four key points to obtain a key point prediction model, which is then used to predict the geometric key points of the dial image.

3. The method for dual-pointer detection and reading calculation as described in claim 2, characterized in that, The process of extracting pointer line segment pixels based on the geometric key points includes: constructing a rectangular ROI for left and right pointers based on the geometric key points; fitting pointer line segments using the start and end points of the rectangular ROI; sampling along the fitted pointer line segments with a set step size to obtain the pointer line segment pixels.

4. The method for dual-pointer detection and reading calculation as described in claim 3, characterized in that, The formula for obtaining the fitted pointer line segment is: ； In the formula, To fit the x-coordinate pixel value of any sampling point on the pointer line segment, To fit the ordinate pixel value of any sampling point on the pointer line segment, This is the line segment sampling ratio coefficient, with a value range of [0,1], used for uniform sampling along the pointer line segment. , ) represents the pixel coordinates of the starting point of the pointer line segment. , ) represents the pixel coordinates of the endpoint of the pointer line segment.

5. The method for dual-pointer detection and reading calculation as described in claim 1, characterized in that, The process of calculating the pointer reading using the line segment-ray intersection ratio method includes: analyzing the pixel coordinates corresponding to the scale values ​​in the prior information of the base map, sorting them, and then forming line segments from adjacent scale values ​​in sequence; A pointer ray is generated with the pointer's starting point as the origin and its ending point as the direction. The intersection of the pointer ray and the line segment is calculated. The scale line segment with a unique valid intersection point is locked, and the pointer reading is calculated using a proportional interpolation method.

6. The method for dual-pointer detection and reading calculation as described in claim 5, characterized in that, The formula for obtaining the pointer reading is: ； In the formula, This is the final physical reading of the position the pointer is pointing to. The scale value corresponding to the starting point of the valid scale segment. d represents the scale value corresponding to the endpoint of the valid scale segment. p d is the pixel distance from the intersection of the pointer ray and the tick mark segment to the starting point of the tick mark segment. L The total pixel length of the effective scale line segment itself.

7. The method for dual-pointer detection and reading calculation as described in claim 1, characterized in that, The process of converting to obtain the color features includes: normalizing the RGB values ​​of the pointer line segment pixels to obtain normalized RGB values, performing extreme value calculations to obtain the maximum, minimum, and color difference of the RGB values, calculating hue, saturation, and brightness based on the maximum, minimum, and color difference, and taking the median of the hue, saturation, and brightness to obtain the color features.

8. The method for dual-pointer detection and reading calculation as described in claim 1, characterized in that, The process of calculating the reference color features includes: calculating the hue, saturation, and brightness of the pointer pixels within the pointer area based on the pointer area marked in the prior information of the base map, and taking the median of each to obtain the reference color features.

9. The method for dual-pointer detection and reading calculation as described in claim 1, characterized in that, The process of determining the pointer attribute includes: calculating the sum of the weighted absolute values ​​of the differences between the color feature and the reference color feature as a similarity index; matching the dial image with the base image pointer based on the similarity index; and determining the pointer attribute of the dial image.

10. The method for dual-pointer detection and reading calculation as described in claim 9, characterized in that, The formula for the similarity index is: ； In the formula, As a similarity metric, The median hue of the pixel in the pointer line segment to be measured. The median hue of the base pointer of the base map. The median saturation value of the pixel segment of the pointer line to be tested. The median saturation of the base pointer of the base map. The median value of the pixel brightness of the pointer line segment to be measured. The median value of the base pointer in the background map. These are the weighting coefficients for hue, saturation, and brightness, respectively.

Images