Liquid drop volume estimation method and system based on machine vision

Abstract

The invention belongs to the technical field of droplet volume detection, and particularly relates to a droplet volume estimation method and system based on machine vision, and the method comprises the steps: collecting a burette background image and a droplet critical point image through a camera; then obtaining a high-quality burette background binary image and a liquid drop critical point binary image through an image processing algorithm, then carrying out difference on the burette background binary image and the liquid drop critical point binary image to obtain an approximate liquid drop binary image, then analyzing boundary data of a side section image to obtain an approximate liquid drop section contour map, and finally obtaining a burette image. Establishing a three-dimensional model of the liquid drop; and finally, measuring and calculating the volume of the liquid drop based on the three-dimensional model. The method not only can replace manual measurement and calculation of the droplet volume and improve the measurement and calculation quality and the measurement and calculation precision, but also realizes the capture of the approximate droplet through binary image difference, and is low in photographing difficulty and high in efficiency because only the critical point image of the droplet which is about to drip from the burette needs to be collected.

Description

Technical Field

[0001] The present invention belongs to the technical field of droplet capacity detection, and in particular relates to a droplet volume estimation method and system based on machine vision. Background Art

[0002] The measurement of droplet capacity is of great significance in production practice and can provide a technical support for refined water-saving irrigation. The traditional method of measuring the droplet capacity of the burette solution is to manually observe the scale changes and record the number of solution droplets falling from the burette, then use the scale difference to obtain the total consumed solution capacity, and finally divide the total consumed solution capacity by the total number of droplets dripping to calculate the capacity of each droplet. However, this manual operation often has the following two problems: First, in order to ensure the accuracy of the droplet capacity reading, the droplet reading must be carried out in a relatively large scale interval, and it must be read repeatedly many times, and then the capacity of each droplet is calculated. This process is time-consuming and labor-intensive, and the efficiency is very low; second, repeatedly reading the scale value on the burette by human eyes will cause a certain degree of visual fatigue, resulting in an increase in the human error rate, which deepens the difficulty and uncertainty of manual reading. If the solution used in chemical experiments is often toxic and volatile, the volatilized toxic gas will cause damage to the experimenter's body. Therefore, it is difficult for manual reading of droplet volume to meet the requirements of scientific development in terms of operation, efficiency, accuracy and safety.

[0003] The invention patent with application number 201710795445.7 proposes a real-time tracking and measurement method for droplet volume based on optical contour curve extraction, which uses two high-speed cameras in vertical directions to collect droplet optical graphics and sends them to the image processing terminal to extract the contour lines on both sides of the X and Y directions in real time, and finally calculates the droplet volume by integration. Although this method realizes the real-time calculation of droplet capacity through image intelligent algorithm, due to the common use of CCD industrial cameras to obtain images, such as high-speed cameras with a maximum frame rate of 30 frames per second when the photo resolution is 640×480, its acquisition speed is difficult to capture a clear and complete image of the droplet dripping instantly, and it is necessary to take pictures repeatedly to obtain a clear and complete image of the droplet, which places high demands on the photographing equipment, is difficult to take pictures, has low efficiency, and has large calculation errors. Summary of the invention

[0004] The purpose of the present invention is to address the above-mentioned problems existing in the prior art and to provide a method and system for estimating the volume of a droplet based on machine vision that does not require multiple photographs and has high efficiency.

[0005] To achieve the above objectives, the technical solution of the present invention is as follows:

[0006] In a first aspect, the present invention provides a method for estimating a droplet volume based on machine vision, comprising:

[0007] S1, collecting the burette background image and the droplet critical point image through the camera;

[0008] S2, using an image processing algorithm to process the burette background image and the droplet critical point image to obtain a burette background binary image and a droplet critical point binary image;

[0009] S3, performing difference processing on the burette background binary image and the droplet critical point binary image to obtain an approximate droplet binary image;

[0010] S4, fitting one side of the approximate droplet binary image to obtain the contour of one side of the approximate droplet binary image, and then obtaining the contour of the other side of the approximate droplet binary image by symmetrical drawing, and finally obtaining the approximate droplet cross-sectional contour image;

[0011] S5, filling the approximate liquid droplet cross-section contour map to obtain an approximate liquid droplet side cross-section filling map, and drawing a three-dimensional model of the approximate liquid droplet based on the approximate liquid droplet side cross-section filling map;

[0012] S6. The three-dimensional model approximating the droplet is regarded as a solid of revolution, and the actual volume of the droplet is calculated by the volume calculation formula of the solid of revolution and the camera parameters.

[0013] In S4, fitting one side of the approximate droplet binary image is specifically performed as follows:

[0014] S41, use the slope single variable point analysis method to find the approximate droplet binary value Figure 1 The segmentation boundary point on the side is determined, and the segmentation area is determined based on the segmentation boundary point;

[0015] S42, determining the best mathematical expressions for these segmented regions by polynomial interpolation, least squares fitting, and multivariate linear regression in turn;

[0016] S43. Based on the best mathematical expressions of all segmented regions, draw the approximate droplet binary Figure 1 Side silhouette.

[0017] In S5, the three-dimensional model approximating the droplet is regarded as a segmented rotating body, and the actual volume of each segment of the rotating body is calculated by the rotating body volume calculation formula and the camera parameters, and then the actual volumes of all rotating bodies are summed up, and the sum result is the actual volume of the droplet.

[0018] The formula for calculating the volume of the rotating body includes:

[0019]

[0020] In the above formula, V and v represent the actual capacity and pixel capacity of the rotating body respectively; u and f represent the object distance and focal length of the camera respectively; Q(x) represents the optimal mathematical expression of the segmented area corresponding to the rotating body located at n≤x≤m, and n and m both represent segment boundary points.

[0021] In S2, the image processing algorithm refers to grayscale, denoising, grayscale enhancement, image segmentation, opening operation to remove impurities, image expansion, image erosion, and image filling in sequence. The denoising refers to an image denoising method based on wavelet transform, the grayscale enhancement refers to an image brightening algorithm based on grayscale transform enhancement, and the image segmentation refers to an image segmentation algorithm based on maximum inter-class variance.

[0022] In a second aspect, the present invention provides a droplet volume estimation system based on machine vision, comprising:

[0023] An image acquisition module, used for acquiring a burette background image and a droplet critical point image through a camera;

[0024] An image processing module is used to process the burette background image and the droplet critical point image using an image processing algorithm to obtain a burette background binary image and a droplet critical point binary image;

[0025] A differential processing module is used to perform differential processing on the burette background binary image and the droplet critical point binary image to obtain an approximate droplet binary image;

[0026] A cross-sectional profile fitting module is used to fit one side of the approximate droplet binary image to obtain the contour of one side of the approximate droplet binary image, and then obtain the contour of the other side of the approximate droplet binary image by symmetrical drawing, and finally obtain the approximate droplet cross-sectional profile image;

[0027] A three-dimensional model drawing module is used to fill the approximate droplet cross-section contour map to obtain the approximate droplet side cross-section filling map, and draw a three-dimensional model of the approximate droplet based on the approximate droplet side cross-section filling map;

[0028] The volume calculation module is used to regard the three-dimensional model of the approximate droplet as a rotating body, and calculate the actual volume of the droplet through the rotating body volume calculation formula and camera parameters.

[0029] The cross-sectional profile fitting module fits one side of the approximate droplet binary image according to the following steps:

[0030] S41, use the slope single variable point analysis method to find the approximate droplet binary value Figure 1 The segmentation boundary point on the side is determined, and the segmentation area is determined based on the segmentation boundary point;

[0031] S42, determining the best mathematical expressions for these segmented regions by polynomial interpolation, least squares fitting, and multivariate linear regression in turn;

[0032] S43. Based on the best mathematical expressions of all segmented regions, draw the approximate droplet binary Figure 1 Side silhouette.

[0033] The volume calculation module calculates the actual volume of the droplet according to the following steps:

[0034] The three-dimensional model approximating the droplet is regarded as a segmented rotating body. The actual volume of each segment of the rotating body is calculated by the rotating body volume calculation formula and the camera parameters. Then the actual volumes of all rotating bodies are summed up, and the sum result is the actual volume of the droplet.

[0035] The formula for calculating the volume of the rotating body includes:

[0036]

[0037] In the above formula, V and v represent the actual capacity and pixel capacity of the rotating body respectively; u and f represent the object distance and focal length of the camera respectively; Q(x) represents the optimal mathematical expression of the segmented area corresponding to the rotating body located at n≤x≤m, and n and m both represent segment boundary points.

[0038] The image processing algorithm refers to grayscale, denoising, grayscale enhancement, image segmentation, opening operation to remove impurities, image expansion, image erosion, and image filling in sequence; wherein the denoising refers to an image denoising method based on wavelet transform, the grayscale enhancement refers to an image brightening algorithm based on grayscale transform enhancement, and the image segmentation refers to an image segmentation algorithm based on maximum inter-class variance.

[0039] Compared with the prior art, the present invention has the following beneficial effects:

[0040] 1. The present invention first uses a camera to dynamically collect the burette background image and the droplet critical point image, then obtains a high-quality burette background binary image and a droplet critical point binary image through an image processing algorithm, obtains an approximate droplet binary image by differentiating the two binary images, then analyzes the boundary data of the side section image, establishes a three-dimensional model of the droplet, and finally realizes the measurement of the droplet capacity based on the three-dimensional model. This method can not only replace the manual measurement of the droplet capacity and improve the measurement quality and accuracy, but also realizes the approximate droplet capture through the binary image difference. It only needs to collect the critical point image of the droplet about to drip from the burette. The acquisition difficulty is low, and an ordinary CCD industrial camera can meet the photography needs.

[0041] 2. Aiming at the interference of the captured image information, the present invention processes the image collected by the camera through the image processing algorithm, thereby realizing the rapid positioning of the critical droplet and the nozzle of the burette under complex background conditions; the image processing algorithm refers to graying, denoising, gray enhancement, image segmentation, opening operation removal, image expansion, image erosion, and image filling in sequence; since the photos taken by the camera are RGB true color images, it is difficult to perform graphic analysis, so it is necessary to convert the RGB color image into a gray image through graying; in order to remove noise, the gray image is denoised by an image denoising method based on wavelet transform, and the image visibility is improved by an image brightening algorithm based on gray transformation enhancement; the image segmentation is realized by the maximum inter-class variance method; in order to eliminate the light spots that appear due to the reflected light when the CCD camera shoots the droplet, the opening operation is used for removal of impurities; finally, the irregular and hollow droplet edge caused by shooting is removed by morphological processing techniques such as image expansion and image erosion and image filling technology, and the droplet edge quality is improved to obtain a high-quality binary image, thereby improving the subsequent fitting accuracy and reliability. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 The present invention is a flowchart of the method.

[0043] Figure 2 This is the burette background image captured by the camera.

[0044] Figure 3 This is the droplet critical point image captured by the camera.

[0045] Figure 4 This is a binary image of the burette background obtained through image processing algorithm.

[0046] Figure 5 It is a binary image of the droplet critical point obtained by image processing algorithm.

[0047] Figure 6 It is the approximate droplet binary image obtained by difference processing.

[0048] Figure 7 This is the fitted approximate droplet cross-sectional profile.

[0049] Figure 8 Comparison diagram of the approximate droplet cross-section profile and the droplet critical point image.

[0050] Fig. 9 Comparison diagram of the approximate droplet cross-section profile and the burette background image.

[0051] Fig.10 Comparison diagram of the approximate droplet cross-sectional contour image and the approximate droplet binary image.

[0052] Fig.11 This is a three-dimensional model diagram of an approximate droplet.

[0053] Fig.12 The structure diagram of the system of the present invention is shown in FIG. DETAILED DESCRIPTION

[0054] The present invention is further described in detail below in conjunction with specific implementation methods and drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0055] Embodiment 1:

[0056] See also Figure 1 , a droplet volume estimation method based on machine vision is carried out in the following steps:

[0057] S1. Capture the burette background image and the droplet critical point image by using a camera. The captured burette background image and the droplet critical point image are respectively as follows: Figure 2 , Figure 3 As shown;

[0058] S2. Process the burette background image and the droplet critical point image using an image processing algorithm to obtain a burette background binary image and a droplet critical point binary image; obtain the burette background binary image and the droplet critical point binary image as shown in the figure. Figure 4 , Figure 5 As shown;

[0059] Specifically, the image processing algorithm includes: the first step, grayscale conversion; the second step, denoising; the third step, grayscale enhancement; the fourth step, image segmentation; the fifth step, opening operation to remove impurities; the sixth step, image expansion; the seventh step, image erosion; the eighth step, image filling; wherein the denoising refers to an image denoising method based on wavelet transform, the grayscale enhancement refers to an image brightening algorithm based on grayscale transform enhancement, and the image segmentation refers to an image segmentation algorithm based on maximum inter-class variance;

[0060] S3. Since the target object is a droplet, the burette background binary image and the droplet critical point binary image are differentially processed to obtain an approximate droplet binary image. The obtained approximate droplet binary image is as follows: Figure 6 As shown;

[0061] Specifically, since spots may remain in the differential processing process, which may have a negative impact on image analysis, it is also necessary to perform image despeckle on the approximate droplet binary image through an opening operation;

[0062] S4. Since the droplet is subject to internal tension and gravity, the left and right boundary contours of the droplet side section should be symmetrical. Therefore, the ideal side contour of the droplet can be fitted, and then the symmetrical drawing can be performed, thereby achieving the purpose of trimming the approximate droplet cross-section image;

[0063] analyze Figure 6 It can be seen that Figure 6 The right boundary of the image is distorted, so the embodiment selects a more ideal left boundary for fitting, obtains the left contour of the approximate droplet binary image, and then obtains the left contour of the approximate droplet binary image by symmetrical drawing, and finally obtains the approximate droplet cross-sectional contour image; specifically, the fitting process is:

[0064] S41, using the slope single variable point analysis method to find the segmentation boundary point on the left side of the approximate droplet binary image, and determining the segmentation area based on the segmentation boundary point;

[0065] S42, determining the best mathematical expressions for these segmented regions through polynomial interpolation, least squares fitting, multivariate linear regression, and residual analysis;

[0066] Polynomial interpolation:

[0067] Given the points x0, x1, …, x n The values ​​of n , suppose there is an m-degree polynomial:

[0068]

[0069] According to the above polynomial, we can get the following equation:

[0070] a0x m +a1x m-1 +…+a m-1 x+a m = y i, i=0, 1, ..., n;

[0071] When m = n and x0, x1, ..., x n When they are different, there is only one polynomial. By solving the above equation, we can get the function expression. If we only want to get the value of the function at the interpolation point, we can use the Lagrange interpolation formula:

[0072]

[0073] Least squares fitting:

[0074] The least squares method is the most commonly used method in mathematical modeling. In essence, it seeks to minimize the error vector in terms of norm (i.e., vector length). is the error vector, find the parameter Minimize the sum of squared errors, that is:

[0075]

[0076] Known empirical formula Using a series of information containing errors ((x i ,y i ), i = 1, ..., n determines the number The least-two-value multiplication principle can be expressed as:

[0077]

[0078] In the above formula, Represents a column vector consisting of data information of variable y, Represents a matrix of data information of variable x (possibly multiple variables), so For the general The vector value function obtained by substituting each set of data into the function f;

[0079] Multiple Linear Regression:

[0080] The multiple linear regression model is:

[0081] y=b0x0+b1x1+…+b p x p +ε;

[0082] In the above formula, ε~N(0,σ 2 ), σ 2 is the variance;

[0083] make The multiple linear regression model can be expressed as:

[0084]

[0085] Now we have y and x0, x1, ..., x p n groups of observations (generally n is required to be much larger than p, when the constant term is considered in the regression model, it is equivalent to x0 taking a constant of 1), and the regression coefficient is required to be calculated Assume and are n×1 and n×(p+1) matrices of the corresponding n groups of observations, respectively, and we can get:

[0086] Then we can get and σ 2 Estimated value:

[0087] For any y and x0, x1, ..., x pThe regression coefficients of n groups of observations can be obtained through the above formula An estimated value of

[0088] Residual analysis:

[0089] The residual ε is defined as the difference between the true value and the fitted estimate. The residual ε is averaged and squared to test the fitting effect.

[0090] S43, based on the optimal mathematical expressions of all segmented regions, draw the left contour of the approximate droplet binary image;

[0091] The final approximate droplet cross-sectional profile is as follows: Figure 7 As shown; the approximate droplet cross-sectional contour image is compared with the droplet critical point image, the burette background image, and the approximate droplet binary image, thereby providing a basis for the fitting effect and reliability; the comparison diagram of the approximate droplet cross-sectional contour image and the droplet critical point image is shown in Figure 8 As shown, from Figure 8 It can be seen that the fitted side cross-sectional profile has a very high degree of overlap with the droplet critical point image. Except for a few flashes, the fitted profile includes the droplet target to be investigated. The comparison diagram of the approximate droplet cross-sectional profile and the burette background image is shown in Figure 1. Fig. 9 As shown, from Fig. 9 It can be seen that the fitted side cross-sectional profile coincides perfectly with the remaining droplet boundary in the burette background image; the comparison diagram between the approximate droplet cross-sectional profile and the approximate droplet binary image is shown in Fig.10 As shown, from Fig.10 It can be seen that fitting the side cross-sectional profile also corrects the error caused by the image processing algorithm, thereby verifying the reliability of the method described in the present invention;

[0092] S5. In order to make the fitted approximate droplet side cross-sectional profile more intuitive, the approximate droplet cross-sectional profile map is filled to obtain the approximate droplet side cross-sectional filling map, and a three-dimensional model of the approximate droplet is drawn based on the approximate droplet side cross-sectional filling map; the obtained three-dimensional model of the approximate droplet is as shown in FIG. Fig.11 As shown;

[0093] S6. The three-dimensional model approximating the liquid droplet is regarded as a segmented rotating body. The actual volume of each segment of the rotating body is first calculated by using the rotating body volume calculation formula and the camera parameters. Then, the actual volumes of all rotating bodies are summed up. The sum result is the actual volume of the liquid droplet.

[0094] Specifically, the volume calculation formula of the rotating body is as follows:

[0095]

[0096] In the above formula, V and v represent the actual capacity and pixel capacity of the rotating body respectively; u and f represent the object distance and focal length of the camera respectively; Q(x) represents the optimal mathematical expression of the segmented area corresponding to the rotating body located at n≤x≤m, and n and m both represent segment boundary points.

[0097] The present invention obtains and calculates the volume of a burette droplet through a complete set of image processing operation processes; the method overcomes the subjectivity of traditional manual droplet volume measurement that is prone to errors and greatly improves the speed and efficiency of burette droplet volume measurement.

[0098] Embodiment 2:

[0099] See also Fig.12 A droplet volume estimation system based on machine vision includes an image acquisition module, an image processing module, a differential processing module, a cross-section profile fitting module, a three-dimensional model drawing module, and a volume calculation module. The image acquisition module is used to acquire a burette background image and a droplet critical point image through a camera; the image processing module is used to process the burette background image and the droplet critical point image using an image processing algorithm to obtain a burette background binary image and a droplet critical point binary image; specifically, the image processing algorithm refers to graying, denoising, gray enhancement, image segmentation, opening operation removal, image expansion, image erosion, and image filling in sequence; wherein the denoising The method refers to an image denoising method based on wavelet transform, the grayscale enhancement refers to an image brightening algorithm based on grayscale transformation enhancement, and the image segmentation refers to an image segmentation algorithm based on maximum inter-class variance; the difference processing module is used to perform difference processing on the burette background binary image and the droplet critical point binary image to obtain an approximate droplet binary image; the cross-sectional profile fitting module is used to fit one side of the approximate droplet binary image to obtain the contour of one side of the approximate droplet binary image, and then obtain the contour of the other side of the approximate droplet binary image through symmetrical mapping, and finally obtain the approximate droplet cross-sectional profile image; specifically, the cross-sectional profile fitting module fits one side of the approximate droplet binary image according to the following steps:

[0100] S41, use the slope single variable point analysis method to find the approximate droplet binary value Figure 1 The segmentation boundary point on the side is determined, and the segmentation area is determined based on the segmentation boundary point;

[0101] S42, determining the best mathematical expressions for these segmented regions by polynomial interpolation, least squares fitting, and multivariate linear regression in turn;

[0102] S43. Based on the best mathematical expressions of all segmented regions, draw the approximate droplet binary Figure 1 side profile;

[0103] The three-dimensional model drawing module is used to fill the approximate droplet cross-sectional contour map to obtain the approximate droplet side cross-sectional filling map, and draw the approximate droplet three-dimensional model based on the approximate droplet side cross-sectional filling map; the volume calculation module is used to regard the approximate droplet three-dimensional model as a rotating body, and obtain the actual volume of the droplet through the rotating body volume calculation formula and camera parameters; specifically, the volume calculation module calculates the actual volume of the droplet according to the following steps:

[0104] The three-dimensional model approximating the droplet is regarded as a segmented rotating body. The actual volume of each segment of the rotating body is calculated by the rotating body volume calculation formula and the camera parameters. Then, the actual volumes of all rotating bodies are summed up, and the sum result is the actual volume of the droplet. The rotating body volume calculation formula includes:

[0105]

[0106] In the above formula, V and v represent the actual capacity and pixel capacity of the rotating body respectively; u and f represent the object distance and focal length of the camera respectively; Q(x) represents the optimal mathematical expression of the segmented area corresponding to the rotating body located at n≤x≤m, and n and m both represent segment boundary points.

[0107] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0108] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0109] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0110] Although the preferred embodiments of the present application have been described, those skilled in the art may make other changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications falling within the scope of the present application.

[0111] Obviously, those skilled in the art can make various changes and modifications to the present application without departing from the spirit and scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the claims of the present application and their equivalents, the present application is also intended to include these modifications and variations.

Claims

1. A method for estimating droplet volume based on machine vision, characterized in that: include: S1, collecting the burette background image and the droplet critical point image through the camera; S2, using an image processing algorithm to process the burette background image and the droplet critical point image to obtain a burette background binary image and a droplet critical point binary image; S3, performing difference processing on the burette background binary image and the droplet critical point binary image to obtain an approximate droplet binary image; S4, fitting one side of the approximate droplet binary image to obtain the contour of one side of the approximate droplet binary image, and then obtaining the contour of the other side of the approximate droplet binary image by symmetrical drawing, and finally obtaining the approximate droplet cross-sectional contour image; S5, filling the approximate liquid droplet cross-section contour map to obtain an approximate liquid droplet side cross-section filling map, and drawing a three-dimensional model of the approximate liquid droplet based on the approximate liquid droplet side cross-section filling map; S6. The three-dimensional model approximating the droplet is regarded as a solid of revolution, and the actual volume of the droplet is calculated by the volume calculation formula of the solid of revolution and the camera parameters.

2. The method for estimating droplet volume based on machine vision according to claim 1, characterized in that: In S4, fitting one side of the approximate droplet binary image is specifically performed as follows: S41, using the slope single variable point analysis method to find the segmentation boundary point on one side of the approximate droplet binary image, and determining the segmentation area based on the segmentation boundary point; S42, determining the best mathematical expressions for these segmented regions by polynomial interpolation, least squares fitting, and multivariate linear regression in turn; S43. Based on the optimal mathematical expressions of all segmented areas, draw the contour of one side of the approximate droplet binary image.

3. The method for estimating droplet volume based on machine vision according to claim 2, characterized in that: In S5, the three-dimensional model approximating the droplet is regarded as a segmented rotating body, and the actual volume of each segment of the rotating body is calculated by the rotating body volume calculation formula and the camera parameters, and then the actual volumes of all rotating bodies are summed up, and the sum result is the actual volume of the droplet.

4. The method for estimating droplet volume based on machine vision according to claim 3, characterized in that: The formula for calculating the volume of the rotating body includes: In the above formula, V and V represent the actual capacity and pixel capacity of the rotating body respectively; u and f represent the object distance and focal length of the camera respectively; Q(x) represents the optimal mathematical expression of the segmented area corresponding to the rotating body located in n≤x≤m, and n and m both represent segment boundary points.

5. A method for estimating droplet volume based on machine vision according to any one of claims 1 to 4, characterized in that: In S2, the image processing algorithm refers to grayscale, denoising, grayscale enhancement, image segmentation, opening operation to remove impurities, image expansion, image erosion, and image filling in sequence. The denoising refers to an image denoising method based on wavelet transform, the grayscale enhancement refers to an image brightening algorithm based on grayscale transform enhancement, and the image segmentation refers to an image segmentation algorithm based on maximum inter-class variance.

6. A droplet volume estimation system based on machine vision, characterized in that: include: An image acquisition module, used for acquiring a burette background image and a droplet critical point image through a camera; An image processing module is used to process the burette background image and the droplet critical point image using an image processing algorithm to obtain a burette background binary image and a droplet critical point binary image; A differential processing module is used to perform differential processing on the burette background binary image and the droplet critical point binary image to obtain an approximate droplet binary image; A cross-sectional profile fitting module is used to fit one side of the approximate droplet binary image to obtain the contour of one side of the approximate droplet binary image, and then obtain the contour of the other side of the approximate droplet binary image by symmetrical drawing, and finally obtain the approximate droplet cross-sectional profile image; A three-dimensional model drawing module is used to fill the approximate droplet cross-section contour map to obtain the approximate droplet side cross-section filling map, and draw a three-dimensional model of the approximate droplet based on the approximate droplet side cross-section filling map; The volume calculation module is used to regard the three-dimensional model of the approximate droplet as a rotating body, and calculate the actual volume of the droplet through the rotating body volume calculation formula and camera parameters.

7. The machine vision-based droplet volume estimation system according to claim 6, characterized in that: The cross-sectional profile fitting module fits one side of the approximate droplet binary image according to the following steps: S41, using the slope single variable point analysis method to find the segmentation boundary point on one side of the approximate droplet binary image, and determining the segmentation area based on the segmentation boundary point; S42, determining the best mathematical expressions for these segmented regions by polynomial interpolation, least squares fitting, and multivariate linear regression in turn; S43. Based on the optimal mathematical expressions of all segmented areas, draw the contour of one side of the approximate droplet binary image.

8. The machine vision-based droplet volume estimation system according to claim 7, characterized in that: The volume calculation module calculates the actual volume of the droplet according to the following steps: The three-dimensional model approximating the droplet is regarded as a segmented rotating body. The actual volume of each segment of the rotating body is calculated by the rotating body volume calculation formula and the camera parameters. Then the actual volumes of all rotating bodies are summed up, and the sum result is the actual volume of the droplet.

9. The machine vision-based droplet volume estimation system according to claim 8, characterized in that: The formula for calculating the volume of the rotating body includes: In the above formula, v and v represent the actual capacity and pixel capacity of the rotating body respectively; u and f represent the object distance and focal length of the camera respectively; Q(x) represents the position corresponding to the rotating body at n≤x≤m 的 The best mathematical expression of the segmented area, n and m both represent the segmentation boundaries.

10. A machine vision-based droplet volume estimation system according to any one of claims 6 to 9, characterized in that: The image processing algorithm refers to grayscale, denoising, grayscale enhancement, image segmentation, opening operation to remove impurities, image expansion, image erosion, and image filling in sequence; wherein the denoising refers to an image denoising method based on wavelet transform, the grayscale enhancement refers to an image brightening algorithm based on grayscale transform enhancement, and the image segmentation refers to an image segmentation algorithm based on maximum inter-class variance.

Images