Mercury sphygmomanometer reading identification method and automatic calibration device

Abstract

According to the mercury sphygmomanometer reading identification method and the automatic calibration device, image morphological processing is adopted, the mercury column convex liquid level can be accurately identified, and accurate positioning of liquid level scribed lines is achieved; the method has high universality, a network training data set is replaced in a mode of combining image morphological processing and a YOLOv5 target detection algorithm, and the method is applicable to industrial instruments or glass measuring tools with scales arranged in the vertical direction. The industrial camera is driven through the linear guide rail module, the shooting visual angle and the mercury column liquid level scribed line are kept in the same horizontal line, the method is also suitable for glass liquid measuring tools such as thermometers and mercury sphygmomanometers, reading errors caused by the fact that the visual angle is not flush with the liquid level are avoided, full-automatic verification of the sphygmomanometers can be achieved, and the detection efficiency is improved. Manual intervention is not needed in the process, an experiment report can be output according to verification regulations after verification is completed, the labor cost is reduced, and the verification efficiency is improved.

Description

Technical Field

[0001] The invention belongs to the field of mercury sphygmomanometer reading recognition and automatic calibration, and in particular relates to a mercury sphygmomanometer reading recognition method and an automatic calibration device. Background Art

[0002] With the continuous improvement of people's living standards and economic levels, the medical and health industry has also developed vigorously, especially the growing demand for medical devices. As a daily health care product, blood pressure monitors are increasingly entering people's lives and have become a standing item in every family.

[0003] Mercury sphygmomanometers are the gold standard for blood pressure measurement and are the most common in the medical field. However, most of the calibration of mercury sphygmomanometers is still done manually, using a blood pressure simulator to assist. A blood pressure simulator is a medical simulation device that simulates a patient's blood pressure and pulse for medical training and clinical simulation. During calibration, the experimenter outputs a fixed pressure value through the blood pressure simulator to make the mercury column rise to a certain height. After the pressure stabilizes, the sphygmomanometer reading is manually read and then recorded manually. This process is susceptible to human error, such as misoperation, misreading, or negligence, which may lead to inaccurate and unreliable records. Manual calibration usually takes a lot of time and effort, which can lead to inefficiency and waste of resources when multiple sphygmomanometers are calibrated at the same time.

[0004] Therefore, it is of great significance to turn the sphygmomanometer calibration to digitalization, automation and machine learning technologies to reduce the impact of human factors on the record and improve the accuracy and reliability of the data. Automated and digital recording systems can better handle large amounts of data, reduce human subjectivity and improve calibration efficiency.

[0005] At present, related research on mercury sphygmomanometer reading recognition methods mostly uses traditional image morphological processing or recognition algorithms based on machine vision. The position of the mercury column scale line is extracted mainly through edge detection and Hough transform algorithms, and the convolutional neural network is used to recognize the scale value. The sphygmomanometer reading is calculated from the correspondence between the mercury column position, scale value and scale line.

[0006] These recognition methods have certain limitations: 1. It is impossible to accurately locate the convex liquid surface of the mercury column, and the recognition reading that is not flush with the liquid surface will cause reading errors; 2. The camera height needs to be manually adjusted to within the reading range, and the camera cannot automatically follow the mercury column liquid surface mark; 3. Traditional convolutional neural networks can only classify and recognize digital scale values, and still need to be positioned by character cutting, screening area, etc., which is easy to cause positioning errors; 4. The recognition method has great limitations and is not universal for sphygmomanometers of different shapes and specifications. Summary of the invention

[0007] In order to solve the above problems, the main purpose of the present invention is to provide a mercury sphygmomanometer reading recognition method and automatic calibration device with high accuracy and strong universality.

[0008] Based on the existing method for identifying mercury sphygmomanometer readings, the present invention further improves the image processing algorithm and adds a target detection algorithm based on deep learning. By combining image morphological processing and YOLOv5 target detection algorithm, accurate recognition of mercury sphygmomanometer readings is achieved.

[0009] In one aspect, the present invention provides a mercury sphygmomanometer automatic calibration device, comprising:

[0010] Standard pressure generating device, as a standard gas source, is used to provide the pressure required for calibration;

[0011] An image acquisition unit, used for acquiring a sphygmomanometer image;

[0012] The linear guide module is equipped with an image acquisition unit, which is used to drive the image acquisition unit to move and keep the acquisition angle of view always aligned with the convex liquid surface of the mercury column during calibration;

[0013] The host computer is used to set up the calibration device, process the calibration data, and print the calibration results.

[0014] On the other hand, the present invention provides a method for identifying readings of a mercury sphygmomanometer, which is applied to the above-mentioned automatic calibration device for a mercury sphygmomanometer, and comprises the following steps:

[0015] Obtain a front image of the mercury sphygmomanometer to be inspected, recorded as an original image;

[0016] Based on the pre-trained Yolov5 object detection network model, the scale information in the original image is identified;

[0017] The first round of image morphological processing algorithm is used to extract the mercury column area in the original image, which is recorded as the ROI image;

[0018] The second round of image morphological processing algorithm is used to extract the mercury column height line in the ROI image;

[0019] The reading of the mercury sphygmomanometer is determined by combining the acquired scale information and the proportional relationship of the mercury column height lines.

[0020] In summary, the present invention has the following beneficial effects:

[0021] 1. The reading recognition method of the present invention has strong versatility. By combining image morphological processing and YOLOv5 target detection algorithm, the network training data set is replaced. It is applicable to industrial instruments or glass measuring tools (such as mercury thermometers, measuring cups, measuring cylinders, etc.) with scales arranged in a vertical direction.

[0022] 2. The reading recognition method of the present invention adopts image morphological processing and image corrosion algorithm to accurately identify the convex liquid surface of the mercury column and realize accurate positioning of the liquid surface mark.

[0023] 3. The present invention drives the image acquisition unit through a linear guide module to keep the shooting angle of view and the mercury column liquid level mark at the same horizontal line. The method is also applicable to glass liquid measuring instruments such as thermometers and mercury sphygmomanometers to avoid reading errors caused by the angle of view not being flush with the liquid level.

[0024] 4. The Yolov5 target detection algorithm is used to realize accurate recognition of scale value characters and corresponding scale lines. Compared with the traditional convolutional neural network to recognize scale values, the method of comparing and identifying areas to screen scale lines is more stable and accurate.

[0025] 5. The reading recognition method of the present invention adopts a combination of image morphological processing and improved YOLOv5 target detection algorithm. Each method is relatively independent and can be optimized separately without affecting the implementation of the overall method. In this way, when better image processing algorithms or deep learning network models appear in different fields, they can be applied to this method to improve the performance of this method.

[0026] 6. The method of the present invention adopts the same principle as manual reading in terms of reading strategy, that is, calculation is performed based on the proportional relationship between the upper and lower scale values ​​of the mercury liquid level mark. This reduces the amount of parameter calculation in the network model and outputs the reading value quickly and accurately.

[0027] 7. The present invention has developed a set of automatic calibration device for mercury sphygmomanometers, which can realize fully automatic calibration of sphygmomanometers without manual intervention. After the calibration is completed, the experimental report can be output according to the calibration procedures, thereby reducing labor costs and improving calibration efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] In order to more clearly illustrate the reading recognition scheme and the automatic verification device in the embodiment of the present invention, the following is a brief introduction to the drawings required for describing the embodiment:

[0029] Figure 1 This is a flow chart of step 2 in a method for identifying mercury sphygmomanometer readings according to an embodiment of the present application.

[0030] Figure 2This is a flow chart of steps 3 and 4 in a method for identifying mercury sphygmomanometer readings according to an embodiment of the present application.

[0031] Figure 3 This is a schematic diagram of scale information identified in a method for identifying readings of a mercury sphygmomanometer according to an embodiment of the present application.

[0032] Figure 4 This is a schematic diagram of the overall structure of an automatic calibration device for a mercury sphygmomanometer according to an embodiment of the present application.

[0033] Figure 5 This is an overall control block diagram of an automatic calibration device for a mercury sphygmomanometer according to an embodiment of the present application. DETAILED DESCRIPTION

[0034] In order to be able to more clearly understand the above-mentioned objects, features and advantages of the present invention, the present invention is described in detail below in conjunction with the accompanying drawings and specific embodiments. It should be noted that, in the absence of conflict, the embodiments of the present application and the features in the embodiments can be combined with each other. In the following description, a lot of specific details are set forth to facilitate a full understanding of the present invention, and the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those generally understood by those skilled in the art of the present invention. The terms used in the specification of the present invention herein are only for the purpose of describing specific embodiments and are not intended to limit the present invention.

[0035] This application is based on the pre-trained Yolov5 target detection network model to identify the scale information in the original image. The first round of image morphological processing algorithm extracts the mercury column ROI area in the original image. The second round of image morphological processing algorithm extracts the mercury column height scale in the mercury column image. Combined with the acquired scale information and the proportional relationship of the mercury column height scale, accurate recognition of the mercury sphygmomanometer reading is achieved.

[0036] The standard pressure generating device, image acquisition unit, linear guide module and host computer interface are designed to realize the automatic calibration of mercury sphygmomanometer. The image corrosion algorithm is used to accurately locate the convex surface of the mercury column; during the calibration process, the linear guide module drives the image acquisition unit to follow the scale line of the mercury column; the Yolov5 target detection algorithm is used to accurately identify the scale value in the captured image and accurately locate the corresponding scale line; the combination of image morphological processing and YOLOv5 target detection algorithm is used to improve the effectiveness and universality of measuring tools with different shapes and specifications.

[0037] A set of automatic calibration device for mercury sphygmomanometers has been developed to realize fully automatic calibration of sphygmomanometers. The calibration process does not require human intervention. After the calibration is completed, the experimental report can be output according to the calibration procedures, reducing labor costs and improving calibration efficiency.

[0038] The technical solution of the present invention is further described in detail below through specific embodiments and in conjunction with the accompanying drawings.

[0039] like Figure 1-Figure 3 As shown, the embodiment of the present application provides a method for identifying the reading of a mercury sphygmomanometer, comprising the following steps:

[0040] Step 1: Obtain a front image of the mercury sphygmomanometer to be inspected, recorded as the original image;

[0041] Step 2: Based on the pre-trained Yolov5 object detection network model, identify the scale information in the original image;

[0042] Step 3: Use the first round of image morphological processing algorithm to extract the mercury column area in the original image, which is recorded as the ROI image;

[0043] Step 4: Use the second round of image morphological processing algorithm to extract the mercury column height line in the ROI image;

[0044] Step 5: Determine the reading of the mercury sphygmomanometer by combining the obtained scale information and the proportional relationship of the mercury column height scale.

[0045] Furthermore, the pre-training of the Yolov5 target detection network model in step 2 is performed as follows:

[0046] 500 images of mercury sphygmomanometers in different positions and lighting conditions were taken as training data sets.

[0047] The Labelimg annotation tool is used to annotate the captured images. The scale values ​​in millimeters of mercury are "0", "20", ... "300", a total of 30 scale values, as identification categories. The annotation box area is the size of the scale value digital rectangular area.

[0048] The Yolov5 network model was trained using the PyCharm development tool. The data set was a collection of mercury sphygmomanometer images with labeled information. The batch size was set to 8 and the epoch was set to 200.

[0049] The trained target detection model is called, and the detection source is set to real-time detection by the image acquisition unit to identify the scale information in the original image.

[0050] like Figure 1As shown, the Yolov5 target detection network model includes an input layer (Input), a backbone network (Backbone), a feature pyramid (Neck) and a prediction output layer (Prediction).

[0051] The input layer uses the Mosaic data enhancement method, and for different data sets, there will be an anchor box with an initial set length and width, which is then compared with the real box, the difference between the two is calculated, and then the network parameters are updated in reverse and iterated; the adaptive image scaling method is also used to make the output layer compatible with image inputs of different sizes.

[0052] The backbone network refers to the network used to extract image features. Its main function is to convert the original input image into a multi-layer feature map for subsequent target detection tasks. In Yolov5, CSPDarknet53 or ResNet backbone networks are used. Both networks are relatively lightweight and can minimize the amount of calculation and memory usage while ensuring high detection accuracy. The main structures in Backbone are Conv module, C3 module, and SPPF module.

[0053] The feature pyramid is used to combine feature maps of different levels to generate feature maps with multi-scale information to improve the accuracy of target detection. In Yolov5, the FPN (Feature Pyramid Network) feature pyramid structure is used to fuse feature maps of different levels through upsampling and downsampling operations to generate a multi-scale feature pyramid. The top-down part mainly achieves the fusion of features at different levels by upsampling and fusing with coarser-grained feature maps, while the bottom-up part uses a convolutional layer to fuse feature maps from different levels. Finally, the prediction output layer returns the classification, confidence, and location area of ​​the scale value.

[0054] Sending the original image of the mercury sphygmomanometer into the pre-trained Yolov5 object detection network model can return the specific numerical values ​​of all scale values ​​in the image, the position area covered by the characters of all scale values ​​in the image, and the scale line positions corresponding to all scale values ​​in the image.

[0055] like Figure 3 As shown, the scale information in the recognized original image specifically includes:

[0056] (1) The specific numerical values ​​of all scale values ​​present in the image, which are used as the classification output by the object detection network.

[0057] (2) The position area covered by all the scale value characters in the image is used as the rectangular box position corresponding to the target network output classification, and the upper left vertex P1 (x1, y1) and lower right corner coordinates P2 (x2, y2) of the rectangular box are returned.

[0058] (3) The scale line positions corresponding to all scale values ​​in the image. The relative positions of the scale line positions of the mercury sphygmomanometer and the position area covered by the scale value characters are fixed. The scale line positions can be obtained according to the position area of ​​the scale value characters.

[0059] like Figure 2 As shown, the step 3 specifically includes:

[0060] Step 3.1: Use median filtering to remove noise from the original image, which is recorded as the filtered image.

[0061] Step 3.2: Convert the color space of the filtered image to the HSV space and record it as the HSV image.

[0062] Step 3.3: The HSV image is used to extract the mercury column ROI area by adjusting the upper and lower thresholds of the H, S, and V variables.

[0063] Further, the HSV image is:

[0064] The color space consists of three parts: hue, saturation, and value. By setting the upper and lower limits of the HSV variable threshold, the part of the image with hue, saturation, and value within a certain range can be extracted, that is, the mercury column ROI area.

[0065] like Figure 2 As shown, the step 4 specifically includes:

[0066] Step 4.1: Use contour extraction and fitting algorithm to extract the mercury column strip area in the ROI image, and fit it with a rectangular area, which is recorded as the fitting image.

[0067] Using the contour detection method based on Suzuki boundary following algorithm, assuming f ij Represents the pixel value at position (i, j). The top row, bottom row, leftmost column, and rightmost column of the picture constitute its frame. NBD indicates that each new boundary found is assigned a unique number. The NBD of the frame is 1, and the remaining boundaries are numbered in sequence. The parent information of any boundary is saved in LNBD or the last NBD. The specific steps include the following:

[0068] Step 4.1.1: Start scanning the image from left to right until the target pixel is found, determine whether it is an outer contour or an inner contour, increase the value of NBD, and set the value of (i2, j2);

[0069] Step 4.1.2: (i2, j2) is used as the new starting point, and the pixels in the neighborhood of (i, j) are scanned clockwise to find a non-zero pixel and represent it as (i1, j1);

[0070] Step 4.1.3: Set (i2, j2) = (i1, j1) and (i3, j3) = (i, j);

[0071] Step 4.1.4: Starting from the next element of the pixel (i2, j2), traverse the neighborhood of (i3, j3) again in a counterclockwise direction, find the first non-zero pixel, and set it to (i4, j4);

[0072] Step 4.1.5: Determine whether the current pixel (i3, j3) is an outer contour or an inner contour and change the value of NBD;

[0073] Step 4.1.6: Repeat the previous operation, i.e., (i4, j4) = (i, j) and (i3, j3) = (i1, j1). Each time a new row is scanned, LNBD is reset to 1, and the entire image is scanned until the lower right corner of the image is reached.

[0074] Furthermore, because the shape of the mercury column is a long rectangle, the rectangular fitting outer contour line is selected to realize the mercury column area correction, so as to obtain the vertex information of all rectangular contours in the image. In order to prevent the influence of tiny contours that may appear in other parts of the image, the area, perimeter and other features of the contour are calculated according to the position of each contour vertex. The contour is screened according to its area, perimeter and other features, and the contour corresponding to the maximum value is the required mercury column contour.

[0075] Step 4.2: Perform regional cropping on the fitting image, which is recorded as the cropped image. Regional cropping on the mercury column fitting image is also performed to prevent the influence of other image contours on recognition.

[0076] Step 4.3: Use a 3*3 convolution kernel to erode the cropped image, which is recorded as the eroded image. The 3*3 convolution kernel is used to erode the cropped image, mainly to erase the details outside the image, and the following pixel blocks (also called kernels) are used for opening operations:

[0077]

[0078] Step 4.4: Use the probabilistic Hough line detection algorithm to extract the boundary line of the corrosion image, which is the height line of the mercury column. The specific steps are as follows:

[0079] Step 4.4.1: Randomly extract a feature point in the image, that is, an edge point. If the point has been calibrated as a point on a straight line, continue to randomly extract an edge point from the remaining edge points until all edge points have been extracted.

[0080] Step 4.4.2: Perform Hough transform on the point and perform cumulative sum calculation.

[0081] Step 4.4.3: Select the point with the largest value in the Hough space. If the point is greater than the threshold, proceed to step 4, otherwise return to step 4.4.1.

[0082] Step 4.4.4: Based on the maximum value obtained by Hough transform, start from this point and move along the direction of the straight line to find the two endpoints of the straight line.

[0083] Step 4.4.5: Calculate the length of the straight line. If it is greater than a certain threshold, it is considered a good straight line output and return to step 4.4.1.

[0084] Further, the step 5 specifically includes:

[0085] The measurement value is obtained by extracting the readings of the upper and lower scales closest to the mercury column line, that is, the mercury column line and the scale position are vertically aligned, and then the distance between the upper and lower scales is divided into several equal segments. The height of each segment is the spacing of each scale. The upper and lower scale values ​​are read respectively. According to the upper and lower scale values, the reading can be calculated according to the proportional formula. The calculation formula is:

[0086]

[0087] Among them, x is the height of the mercury column scale in the acquired image, P1 and P2 are the upper and lower scale values, y1 is the position of the long scale line corresponding to the scale value P1 in the image, and y2 is the position of the scale line corresponding to the scale value P2 in the image.

[0088] like Figure 4 As shown, the embodiment of the present application also provides a set of automatic calibration device for mercury sphygmomanometers. Among them, the standard pressure generating device 5 is mainly composed of a micro-pressure pump, an STM32 single-chip microcomputer, a reduction motor and a digital pressure gauge, etc., which is used as a standard gas source, and the calibration pressure can be set through the PC terminal 11 through the RS232 protocol; the solenoid valve 6 is connected to the PC terminal 11 through the relay 7, and can control the on-off of the gas path between the standard pressure generating device 5 and the sphygmomanometer 8 to be tested; the industrial camera 9 (i.e., the image acquisition unit) is fixed by the mold and the linear guide module 10, which can control the up and down movement of the camera, and keep the camera angle of view always aligned with the convex liquid surface of the mercury column during calibration.

[0089] 1) Standard pressure generating device, used to provide the pressure required for calibration, mainly composed of a micro-pressure pump, an STM32 single-chip microcomputer, a reduction motor and a digital pressure gauge.

[0090] As a standard gas source, it provides the pressure required for calibration. During calibration, the calibration pressure is set by the host computer, and the digital pressure gauge collects the reading of the pressure standard gauge in real time as the standard value of the system, compares it with the set pressure value, and controls the frequency of the STM32 microcontroller output pulse according to the difference between the two, thereby controlling the rotation speed of the reduction motor, driving the piston of the micro-pressure pump to advance, and when the set pressure point is reached, the motor speed is 0.

[0091] Furthermore, the standard pressure generating device is connected to the sphygmomanometer to be tested by an airway, with an electromagnetic valve provided therebetween, and the on-off of the airway can be controlled by a host computer control module.

[0092] Furthermore, the standard pressure generating device is connected to the PC via an RS232 interface of the Modbus communication protocol, and the pressure value and pressure curve can be set via the host computer.

[0093] 2) Industrial camera, used to obtain the image of the blood pressure monitor.

[0094] The industrial camera is photographing the sphygmomanometer, obtaining the sphygmomanometer image, and sending it to the image display module for real-time image display, and automatically obtaining the sphygmomanometer reading through the image processing and recognition module. The industrial camera and the linear guide module are fixed by a mold.

[0095] 3) Linear guide module, used to drive the camera to move up and down. During the calibration process, the mercury column height scale line in the camera's field of view is controlled to always be located at the longitudinal center scale line of the image, that is, the camera's field of view is always kept aligned with the convex liquid surface of the mercury column.

[0096] 4) The host computer interface is used to set up the calibration device, process the calibration data and print the calibration results. It is written using the Python-based PyQt5 function library and mainly includes:

[0097] Image display module, image processing and recognition module, pressure display module, control module, data storage processing module. Among them, the industrial camera collects the sphygmomanometer image and displays it in real time through the image display module, the image processing and recognition module processes the collected sphygmomanometer image, and combines the deep learning network model to identify the reading; the data storage processing module processes and stores the recognition results and the collected standard pressure values, and displays the collected standard pressure values ​​in real time through the pressure display module; the control module sends instructions to the STM32 microcontroller to make the standard pressure generating device generate pressure for the calibration experiment, and also controls the on and off of the solenoid valve through the relay to control the on and off of the gas circuit, and can also control the biaxial movement of the guide rail through the guide rail controller; finally, after the calibration experiment is completed, the data storage processing module can automatically generate an experimental report.

[0098] like Figure 3 As shown, 1 is an image taken by an industrial camera, that is, the original image; P1 and P2 are two scale values ​​in "mmHG" on the mercury sphygmomanometer; 2 is the scale area detection frame corresponding to the P1 scale value output by the YOLOv5 deep learning network model; 3 is the scale line corresponding to the scale value P1, and 4 is the scale line corresponding to the scale value P2; the height of the mercury column in the image is the scale line x, the upper and lower scale values ​​are P1 and P2 respectively, the position of the long scale line corresponding to the scale value P1 in the image is y1, and the position of the scale line corresponding to the scale value P2 in the image is y2, and the calculation formula for the reading is:

[0099]

[0100] like Figure 5 As shown, the host computer stores a calibration unit, including: an image processing and recognition module, an image display module, a data storage processing module, a pressure display module and a control module. The industrial camera collects the sphygmomanometer image and displays it in real time through the image display module. The image processing and recognition module processes the collected sphygmomanometer image and recognizes its reading in combination with the deep learning network model; the data storage processing module processes and stores the recognition result and the collected standard pressure value, and displays the collected standard pressure value in real time through the pressure display module; the control module sends instructions to the STM32 single-chip microcomputer to make the standard pressure generating device generate pressure for the calibration experiment, and also controls the on and off of the solenoid valve through the relay to control the on and off of the gas circuit, and can also control the biaxial movement of the guide rail through the guide rail controller; finally, after the calibration experiment is completed, the data storage processing module can automatically generate an experimental report.

[0101] The above description is only a preferred embodiment of the present invention, and does not limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made by using the contents of the present invention specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A mercury sphygmomanometer automatic calibration device, characterized in that: include: Standard pressure generating device, as a standard gas source, is used to provide the pressure required for calibration; An image acquisition unit, used for acquiring a sphygmomanometer image; The linear guide module is equipped with an image acquisition unit, which is used to drive the image acquisition unit to move and keep the acquisition angle of view always aligned with the convex liquid surface of the mercury column during calibration; The host computer is used to set up the calibration device, process the calibration data, and print the calibration results.

2. The automatic calibration device for mercury sphygmomanometer according to claim 1, characterized in that: The standard pressure generating device is connected to the sphygmomanometer to be tested by an air guide tube, with an electromagnetic valve arranged therebetween, and the on-off of the air path is controlled by a host computer.

3. The automatic calibration device for a mercury sphygmomanometer according to claim 1, characterized in that: The image acquisition unit and the linear guide rail module are fixed by a mold, facing the sphygmomanometer instrument, and after acquiring the sphygmomanometer image, it is sent to the image display module for real-time image display, and the sphygmomanometer reading is automatically acquired through the image processing and recognition module.

4. A mercury sphygmomanometer automatic calibration device according to claim 1 or 2, characterized in that: The host computer stores a verification unit, including: Image display module, image processing and recognition module, pressure display module, control module and data storage and processing module; The image display module is used to display the collected sphygmomanometer image in real time; The image processing and recognition module is used to process the collected sphygmomanometer image and recognize its reading in combination with the deep learning network model; The data storage processing module is used to process and store the recognition results and the collected standard pressure values, and to display the collected standard pressure values ​​in real time through the pressure display module; The control module is used to enable the standard pressure generating device to generate pressure, control the on-off of the gas circuit, and control the biaxial movement of the guide rail through the guide rail controller.

5. A method for identifying readings of a mercury sphygmomanometer, applied to an automatic calibration device for a mercury sphygmomanometer as claimed in any one of claims 1 to 4, characterized in that: The following steps are involved: Step 1: Obtain a front image of the mercury sphygmomanometer to be inspected, recorded as the original image; Step 2: Based on the pre-trained Yolov5 object detection network model, identify the scale information in the original image; Step 3: Use the first round of image morphological processing algorithm to extract the mercury column area in the original image, which is recorded as the ROI image; Step 4: Use the second round of image morphological processing algorithm to extract the mercury column height line in the ROI image; Step 5: Determine the reading of the mercury sphygmomanometer by combining the obtained scale information and the proportional relationship of the mercury column height scale.

6. The method for identifying the reading of a mercury sphygmomanometer according to claim 5, characterized in that: In step 2, the Yolov5 target detection network model is pre-trained as follows: Take images of mercury sphygmomanometers at different positions and under different lighting conditions as training data sets; Use the annotation tool to annotate the captured image, use the scale value in millimeters of mercury as the identification category, and the annotation box area is the size of the scale value digital rectangular area; The Yolov5 network model was trained using the PyCharm development tool. The data set was a picture of a mercury sphygmomanometer with labeled information. The trained target detection model is called, and the detection source is set to real-time detection by the image acquisition unit to identify the scale information in the original image.

7. A method for identifying mercury sphygmomanometer readings according to claim 6, characterized in that: The scale information includes: The specific values ​​of all scale values ​​in the image, the position area covered by all scale value characters, and the scale line positions corresponding to all scale values.

8. The method for identifying the reading of a mercury sphygmomanometer according to claim 5, characterized in that: The step 3 specifically includes: Step 3.1: Use median filtering to remove noise from the original image, recorded as the filtered image; Step 3.2: Convert the color space of the filtered image to the HSV space, recorded as the HSV image; Step 3.3: The HSV image is used to extract the mercury column ROI area by adjusting the upper and lower thresholds of the H, S, and V variables.

9. The method for identifying the reading of a mercury sphygmomanometer according to claim 8, characterized in that: The step 4 specifically includes: Step 4.1: Extract the mercury column strip area in the ROI image and fit it with a rectangular area, which is recorded as the fitting image; Step 4.2: Perform regional cropping on the fitted image, recorded as the cropped image; Step 4.3: Use the convolution kernel to corrode the cropped image, recorded as the corroded image; Step 4.4: Use the probabilistic Hough line detection algorithm to extract the boundary line of the corrosion image, which is the height line of the mercury column.

10. The method for identifying the reading of a mercury sphygmomanometer according to claim 5, characterized in that: The step 5 specifically includes: Step 5.1: Determine the scale values ​​above and below the extracted mercury column height scale line according to its height; Step 5.2: Extract the scale line positions corresponding to the upper and lower scale values; Step 5.3: Calculate the reading of the mercury sphygmomanometer based on the height of the mercury column, the upper and lower scale values, and the position of the scale lines.

Images