Real-time monitoring method of liquid material droplet characteristics based on vision and single-pixel calibration

Abstract

The application discloses a real-time monitoring method for liquid material drop characteristics based on vision and single-pixel calibration, which comprises the following steps: a camera is used to collect a group of images of a material drop and identify the outline in real time; the identified outline is scaled and vertically corrected; a fitting algorithm is used to establish a three-dimensional model; the number of three-dimensional pixels is calculated; the calibration quality of a single three-dimensional pixel is calculated; and the real-time quality of the material drop is calculated and monitored. The application belongs to the technical field of liquid material drop monitoring, particularly relates to an image processing and quality acquisition algorithm for liquid high-temperature material drops, and is suitable for monitoring the shape, volume and quality characteristics of liquid material drops. The application provides a three-camera vision-based three-dimensional modeling and single-pixel quality calibration algorithm which can acquire the quality of material drops in real time during the falling process of the material drops, can acquire the three-dimensional model and quality characteristics of each material drop in real time, has high precision, and is accurate and reliable in data, thereby providing data support for the monitoring of liquid material drops.

Description

Technical Field

[0001] This invention belongs to the field of liquid droplet monitoring technology, and specifically relates to an image processing and quality acquisition algorithm for high-temperature liquid droplets. It is applicable to the monitoring of the shape, volume and quality characteristics of liquid droplets, and specifically refers to a real-time monitoring method for liquid droplet characteristics based on vision and single-pixel calibration. Background Technology

[0002] In manufacturing industries such as steel, glass, and plastics, product quality control is achieved by controlling the mass of liquid droplets. This mass is determined mechanically by the action of a punch or shears in the feed channel, but precisely measuring the droplet mass during this process is difficult. This makes it challenging to achieve target quality control, creating a significant industry barrier. Furthermore, it's impossible to weigh the liquid droplets during their descent; monitoring can only be achieved through non-contact, indirect methods.

[0003] High-speed cameras are a type of industrial camera, generally referring to digital industrial cameras. They are typically installed on production lines to replace the human eye for measurement and judgment. They capture digital images of the target, convert them into image signals, and transmit them to a dedicated image processing system. High-speed camera vision analysis provides solutions for the industry; with sampling rates of hundreds of frames per second and high resolution, it can acquire complete images of each droplet's fall in real time. However, during free fall, the droplets are affected by aerodynamics and the initial shear stress of the shears, which can cause tilting and surface irregularities. Summary of the Invention

[0004] The problem the invention aims to solve:

[0005] To address the above issues and overcome the shortcomings of existing technologies, the present invention aims to provide a three-camera vision-based stereo modeling and single-pixel quality calibration algorithm that can acquire the mass of a droplet in real time during its descent. Addressing the issue that the droplet may tilt and develop surface irregularities due to aerodynamic forces and initial shear stress during free fall, the algorithm corrects these irregularities. Furthermore, multi-camera, multi-view stereo modeling is employed to increase the visible area, capture more surface irregularities, improve the accuracy of observation data and the algorithm, and comprehensively ensure that the calculated volume is not affected by technical limitations.

[0006] By marking a droplet and weighing it to obtain its volume and mass, the calibration average mass of a single 3D pixel is calculated and used as a common variable for subsequent droplet monitoring. This allows for the real-time acquisition of the mass of all droplets during the production process, and timely calibration and correction of the calibration average mass of a single 3D pixel during production.

[0007] Technical means to solve the problem:

[0008] The real-time monitoring method for liquid droplet features based on vision and single-pixel calibration proposed in this invention includes the following steps:

[0009] Step 1: Acquire a set of images of a single droplet in real time using a camera, and perform contour recognition on the images to obtain the recognized contour;

[0010] Step 2: Scale and vertically correct the identified contours from Step 1; obtain a new set of three images.

[0011] Step 3: Based on the three contour curves of the new image and the center line formed by the center points of the nth row contour points of the three new images, a fitting algorithm is used to establish a 3D model and calculate the number of pixels containing the 3D model.

[0012] Step 4: Weigh the calibration droplet;

[0013] Step 5: Calculate the calibration quality of a single 3D pixel based on the number and quality of the calibrated droplet.

[0014] Step 6: Calculate and monitor the real-time quality of the droplet based on the volume of the droplet measured subsequently and the calibration quality of a single 3D pixel.

[0015] Furthermore, the cameras mentioned in step one are three cameras, a set of images consists of three images, and the identified contours consist of three identified contours.

[0016] Preferably, the contour recognition method in step one involves binarizing the three images and then extracting the contours, specifically by extracting the coordinate values ​​of the contour points of each contour boundary line.

[0017] As a further preferred embodiment of the present invention, the coordinate values ​​of the contour points are the row and column values ​​of a continuous and closed string of boundary pixels.

[0018] As a further preferred embodiment of the present invention, the scaling and vertical correction method described in step two involves obtaining the top and bottom tangent points in the droplet identification contour of each image, and calculating the height value (i.e., the difference in the number of rows between the top and bottom tangent points) and width value (the difference in the number of columns between the leftmost and rightmost tangent points) of the droplet shape. The other two images are then scaled according to the height value of the first image to obtain a new set of three images, namely new image A, new image B, and new image C.

[0019] Furthermore, the specific steps in step three, which involve using a fitting algorithm to establish a stereo model and calculating the number of stereo pixels, include the following operations:

[0020] 1) Mobile Center Hierarchical Matching

[0021] To form a droplet 3D contour from the contour vector architecture of three new images of equal height, the six contour points of the same row number are subjected to 2D center matching. The coordinates of the contour points of the three new images corresponding to the nth row of pixels are as follows: In the nth row, the coordinates of the two contour points in new image A are... , The center point is The coordinates of the two contour points in the new image B are: , The center point is The coordinates of the two contour points in the new image C are: , The center point is Where n represents the number of rows and x represents the number of columns in which the contour points are located;

[0022] Three cameras, A, B, and C, are used. New images B and C are translated to the center point of new image A according to their center point coordinates. Simultaneously, the angles between cameras B and C and camera A are included in the two-dimensional coordinates of the new image points. The angle between camera A and the image points is... The angle of the image from camera B is... The angle of the image taken by camera C is... Then, the two-dimensional cross-section of the nth row is obtained, and the two contour points in the new image A are represented as: , The two contour points in the new image B are represented as follows: , The two contour points in the new image C are represented as follows: , ;in Not necessarily Twice as, when the three cameras are positioned at 120 degrees each. , ;

[0023] 2) Removing error data based on the double mean error detection elimination method

[0024] Before using the contour points of the new image in 1) to construct the ellipse, it is necessary to remove erroneous data or points with large errors;

[0025] Removal method:

[0026] (1) Calculate the distance from all points to the center point;

[0027] (2) Calculate the mean of all distances;

[0028] (3) Calculate the deviation of each distance from the mean and calculate the standard deviation;

[0029] (4) Remove points with a distance deviation greater than twice the standard deviation;

[0030] The formula for calculating standard deviation is as follows:

[0031] (1)

[0032] Let d be the standard deviation, and d be the distance of each point from the center point. This represents the average distance of each point from the center point, where n represents the number of rows.

[0033] 3) Least square fitting ellipse parameters

[0034] The nth layer can be regarded as a two-dimensional plane, and an ellipse can be fitted given six points and the center position.

[0035] Curve equation:

[0036] (2)

[0037] Given six points whose coordinates represent the coordinates of the contour points of the three new images corresponding to the nth pixel row, and six plane points obtained after angle normalization from these six vector points, and substituted into the curve normal equation, the objective function is:

[0038] (3)

[0039] Least squares adjustment processing makes To minimize this, we finally calculate the values ​​of the equation coefficients A, B, C, D, E, and F.

[0040] Based on the values ​​of A, B, C, D, E, and F, the ellipse parameters can be calculated using the ellipse calculation formula:

[0041] (4)

[0042] (5)

[0043] (6)

[0044] (7)

[0045] (8)

[0046] Where a is the length of the major semi-axis of the ellipse, and b is the length of the minor semi-axis of the ellipse. It is a parameter. It is a parameter. The angle between the line connecting the origin and a point on the ellipse and the positive x-axis is also called the angle of elevation.

[0047] Then, the area of ​​the pixel row containing the ellipse (i.e., the number of pixels it contains) can be calculated:

[0048] (i=0~N) (9)

[0049] in, This represents the area of ​​the pixel row containing the ellipse. Pi Let be the length of the major semi-axis of the n-row ellipse. Let be the length of the minor semi-axis of the n-row ellipse;

[0050] 4) Calculate the volume of the three-dimensional material droplet by layer-by-layer accumulation:

[0051] (10)

[0052] in, The volume of the three-dimensional droplet.

[0053] Preferably, step five, which involves calculating the calibration quality of a single stereo pixel, includes the following steps:

[0054] First, in step four, weigh the calibration droplet product to obtain its quality. The calibration quality of a single stereo pixel is:

[0055] (11)

[0056] in, Calibrate the quality of a single stereo pixel; The volume of each 3D droplet.

[0057] As a further preferred embodiment of the present invention, the calculation and monitoring of the real-time mass of the droplets in step six, wherein the real-time mass of each droplet is calculated as follows:

[0058] (12)

[0059] in, The real-time quality of each droplet. For the volume of the droplet measured later, To calibrate the quality of a single stereo pixel.

[0060] Preferably, the three cameras are three cameras with the same resolution that are intersected at arbitrary angles.

[0061] The beneficial effects achieved by the present invention using the above solution are as follows:

[0062] This invention employs three cameras of the same resolution, each positioned at an arbitrary angle, to capture real-time images of falling droplets from the molten metal furnace inlet on the production line. The three images are matched, and through contour recognition, scaling, vertical correction, and coaxial stereo fitting, the number of stereo pixels contained within the solid image is obtained. Simultaneously, a marked droplet is precisely weighed, and its weight and the number of stereo pixels are used to calculate the marking mass of a single stereo pixel. Subsequent droplets can then have their mass calculated using the marking mass of a single stereo pixel and the real-time calculated number of stereo pixels contained within the solid image. This algorithm enables real-time acquisition and monitoring of the shape and mass of all droplets on the production line.

[0063] This invention provides a three-camera vision-based stereo modeling and single-pixel quality calibration algorithm that can acquire the mass of a droplet in real time during its descent. Addressing the issue that droplets may tilt and develop surface irregularities due to aerodynamic forces and initial shear stress during free fall, the algorithm corrects these issues. Furthermore, multi-camera, multi-view stereo modeling is employed to increase the visible area and capture more surface irregularities, thereby improving the accuracy of both observation data and the algorithm. This comprehensive approach ensures that the calculated volume is not affected by technical limitations.

[0064] By marking a droplet and weighing it to obtain its volume and mass, the calibration average mass of a single 3D pixel is calculated and used as a common variable for subsequent droplet monitoring. This allows for the real-time acquisition of the mass of all droplets during the production process, and timely calibration and correction of the calibration average mass of a single 3D pixel during production.

[0065] This solution can acquire the three-dimensional model and mass characteristics of each droplet in real time with high accuracy and reliable data, providing data support for the monitoring of liquid droplets. Attached Figure Description

[0066] Figure 1 This is a diagram illustrating the data acquisition process of the three cameras used in this invention to acquire images.

[0067] Figure 2 These are three images captured by a camera during the process of liquid droplets falling;

[0068] Figure 3 Three new images after scaling and vertical correction;

[0069] Figure 4 This is a schematic diagram of the extracted contour after binarization.

[0070] Figure 5 These are outline images of three pictures captured by a camera during the dripping of liquid material;

[0071] Figure 6Outline images of the three new images after scaling and vertical correction;

[0072] Figure 7 These are the coordinates of the six contour vector points and the center point corresponding to the nth pixel row;

[0073] Figure 8 Center matching map for the contour points in the nth row;

[0074] Figure 9 A two-dimensional planar ellipse representing the cross-section of a three-dimensional droplet;

[0075] Figure 10 This is a stereo model fitted based on three cameras;

[0076] Figure 11 This is a graph showing the real-time quality monitoring of the material droplets.

[0077] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used together with the embodiments of the invention to explain the invention and do not constitute a limitation thereof. Detailed Implementation

[0078] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0079] In the description of this invention, it should be understood that the terms "upper", "lower", "front", "rear", "left", "right", "top", "bottom", "inner", "outer", etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this invention.

[0080] like Figures 1-11 As shown, this invention proposes a real-time monitoring method for liquid droplet features based on vision and single-pixel calibration, comprising the following steps:

[0081] Step 1, see appendix Figure 1 Data acquisition diagram of three cameras, showing the real-time acquisition of a set of images of a single droplet by each camera (see attached diagram). Figure 2 and attached Figure 3 The image is then used to perform contour recognition to obtain the recognized contours.

[0082] Step 2: Scale and vertically correct the identified contours from Step 1; obtain a new set of three images; Step 3: Based on the three contour curves of the new images and the center line formed by the center points of the nth row contour points of the three new images, use a fitting algorithm to build a stereo model and calculate the number of stereo pixels.

[0083] Step 4: Weigh the calibration droplet;

[0084] Step 5: Calculate the calibration quality of a single 3D pixel based on the number and quality of the calibrated droplet.

[0085] Step 6: Calculate and monitor the real-time quality of the droplet based on the volume of the droplet measured subsequently and the calibration quality of a single 3D pixel.

[0086] As one embodiment of this solution, such as Figure 1 As shown, the cameras mentioned in step one refer to three cameras, which are three cameras with the same resolution that are positioned at arbitrary angles. A set of images consists of three images; the recognized contours consist of three recognized contours.

[0087] As one embodiment of this solution, such as Figure 4 As shown, the contour recognition method described in step one involves extracting contours after binarizing three images, such as... Figure 5 As shown, this specifically involves extracting the coordinate values ​​of the contour points of each contour boundary line.

[0088] Preferably, the coordinate values ​​of the contour points are the row and column values ​​of a continuous and closed string of boundary pixels.

[0089] Following the above embodiments, as another embodiment of this solution, such as... Figure 5 As shown, the scaling and vertical correction method described in step two involves obtaining the top and bottom tangent points in the droplet recognition contour of each image, and calculating the height value (i.e., the difference in the number of rows between the top and bottom tangent points) and width value (the difference in the number of columns between the tangent points of the leftmost and rightmost points) of the droplet shape. Figure 6 As shown, the other two images are scaled proportionally according to the height value of the first image to obtain a new set of three images, namely new image A, new image B, and new image C.

[0090] The specific steps for establishing a 3D model using a fitting algorithm and calculating the number of 3D pixels described in step three include the following operations:

[0091] 1) Mobile Center Hierarchical Matching

[0092] To form a 3D droplet profile from the contour vector architecture of three new images of equal height, the six contour points of equal row number are matched in two dimensions, such as... Figure 7As shown, the coordinates of the contour points of the three new images corresponding to the nth row of pixels are as follows: In the nth row, the coordinates of the two contour points in the new image A are... , The center point is The coordinates of the two contour points in the new image B are: , The center point is The coordinates of the two contour points in the new image C are: , The center point is Where n represents the number of rows and x represents the number of columns in which the contour points are located;

[0093] Three cameras, A, B, and C, are used. New images B and C are translated to the center point of new image A according to their center point coordinates. Simultaneously, the angles between cameras B and C and camera A are included in the two-dimensional coordinates of the new image points. The angle between camera A and the image points is... The angle of the image from camera B is... The angle of the image taken by camera C is... Then obtain the two-dimensional cross-section of the nth row, such as Figure 8 As shown ( Figure 8 (As shown in the schematic diagram): The two contour points in the new image A are represented as follows: , The two contour points in the new image B are represented as follows: , The two contour points in the new image C are represented as follows: , ;in Not necessarily Twice as, when the three cameras are positioned at 120 degrees each. , .

[0094] 2) Removing error data based on the double mean error detection elimination method

[0095] Before using the contour points of the new image in 1) to construct the ellipse, it is necessary to remove erroneous data or points with large errors;

[0096] Removal method:

[0097] (1) Calculate the distance from all points to the center point;

[0098] (2) Calculate the mean of all distances;

[0099] (3) Calculate the deviation of each distance from the mean and calculate the standard deviation;

[0100] (4) Remove points with a distance deviation greater than twice the standard deviation;

[0101] The formula for calculating standard deviation is as follows:

[0102] (1)

[0103] Let d be the standard deviation, and d be the distance of each point from the center point. This represents the average distance of each point from the center point, where n represents the number of rows.

[0104] 3) Least square fitting ellipse parameters

[0105] The nth layer can be regarded as a two-dimensional plane, and an ellipse can be fitted given six points and the center position.

[0106] Curve equation:

[0107] (2)

[0108] Given six points whose coordinates represent the coordinates of the contour points of the three new images corresponding to the nth pixel row, and six plane points obtained after angle normalization from these six vector points, and substituted into the curve normal equation, the objective function is:

[0109] (3)

[0110] Least squares adjustment processing makes To minimize this, we finally calculate the values ​​of the equation coefficients A, B, C, D, E, and F.

[0111] Based on the values ​​of A, B, C, D, E, and F, according to... Figure 9 The parameters of an ellipse can be calculated using the formula shown below.

[0112] (4)

[0113] (5)

[0114] (6)

[0115] (7)

[0116] (8)

[0117] Where a is the length of the major semi-axis of the ellipse, and b is the length of the minor semi-axis of the ellipse. It is a parameter. It is a parameter. The angle between the line connecting the origin and a point on the ellipse and the positive x-axis is also called the angle of elevation.

[0118] Then, the area of ​​the pixel row containing the ellipse (i.e., the number of pixels it contains) can be calculated:

[0119] (i=0~N) (9)

[0120] in, This represents the area of ​​the pixel row containing the ellipse. Pi Let be the length of the major semi-axis of the n-row ellipse. Let be the length of the minor semi-axis of the n-row ellipse;

[0121] 4) Calculate the volume of the three-dimensional material droplet by layer-by-layer accumulation:

[0122] (10)

[0123] in, The volume of the three-dimensional droplet.

[0124] As a preferred embodiment of this solution, the calculation of the calibration quality of a single stereo pixel in step five includes the following steps:

[0125] First, in step four, weigh the calibration droplet product to obtain its quality. The calibration quality of a single stereo pixel is:

[0126] (11)

[0127] in, Calibrate the quality of a single stereo pixel; The volume of each 3D droplet.

[0128] Step six describes the calculation and monitoring of the real-time mass of the droplets, where the real-time mass of each droplet is calculated as follows:

[0129] (12)

[0130] in, The real-time quality of each droplet. For the volume of the droplet measured later, To calibrate the quality of a single stereo pixel.

[0131] The above is the overall operation flow of the present invention. Referring to an embodiment of the present invention, see below. Figure 10 The figure shown is a three-dimensional model of the droplet obtained by the three-camera fitting method proposed in this invention; as shown Figure 11 The figure shown is a graph illustrating the real-time quality monitoring of the droplet quality according to the present invention.

[0132] Among them, the algorithm proposed in this invention for establishing a three-dimensional droplet from three images, the volume calculation method for obtaining the number of three-dimensional pixels of the three-dimensional model by accumulation, the calculation of the three-dimensional pixel calibration mass based on the three-dimensional volume and weighing mass of the calibrated droplet, and the calculation of the mass of subsequent droplets, and the method for real-time monitoring of droplet mass and shape, etc., are all novel methods in the industry and have not been disclosed in the industry at present.

[0133] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

[0134] The present invention and its embodiments have been described above. This description is not restrictive, and the accompanying drawings are only one embodiment of the present invention; the actual structure is not limited thereto. In conclusion, if those skilled in the art are inspired by this description and design similar structures and embodiments without departing from the spirit of the invention, such designs should fall within the protection scope of the present invention.

Claims

1. A real-time monitoring method for liquid droplet features based on vision and single-pixel calibration, characterized in that, Includes the following steps: Step 1: Acquire a set of images of a single droplet in real time using a camera, and perform contour recognition on the images to obtain the recognized contour; Step 2: Scale and vertically correct the identified contours from Step 1; obtain a new set of three images. Step 3: Based on the three contour curves of the new image and the center line formed by the center points of the nth row contour points of the three new images, a fitting algorithm is used to establish a 3D model and calculate the number of pixels containing the 3D model. Step 4: Weigh the calibration droplet; Step 5: Calculate the calibration quality of a single 3D pixel based on the number and quality of the calibrated droplet. Step 6: Calculate and monitor the real-time quality of the droplet based on the volume of the droplet measured subsequently and the calibration quality of a single 3D pixel; The cameras mentioned in step one are three cameras, a set of images consists of three images, and the recognition contours consist of three recognition contours; The contour recognition method described in step one involves binarizing the three images and then extracting the contours, specifically by extracting the coordinate values ​​of the contour points of each contour boundary line. The coordinate values ​​of the contour points are the row and column values ​​of a continuous and closed boundary pixel string; The scaling and vertical correction method described in step two involves obtaining the top and bottom tangent points of each droplet recognition contour, calculating the height and width values ​​of the droplet shape, scaling the other two images according to the height value of the first image, and obtaining a new set of three images, namely new image A, new image B, and new image C. The specific steps for establishing a 3D model using a fitting algorithm and calculating the number of 3D pixels described in step three include the following operations: 1) Mobile Center Hierarchical Matching To form a droplet 3D contour from the contour vector architecture of three new images of equal height, the six contour points of the same row number are subjected to 2D center matching. The coordinates of the contour points of the three new images corresponding to the nth row of pixels are as follows: In the nth row, the coordinates of the two contour points in new image A are... , The center point is The coordinates of the two contour points in the new image B are: , The center point is The coordinates of the two contour points in the new image C are: , The center point is Where n represents the number of rows and x represents the number of columns in which the contour points are located; Three cameras, A, B, and C, are used. New images B and C are translated to the center point of new image A according to their center point coordinates. Simultaneously, the angles between cameras B and C and camera A are included in the two-dimensional coordinates of the new image points. The angle between camera A and the image points is... The angle of the image from camera B is... ; The angle of the image from camera C is Then, the two-dimensional cross-section of the nth row is obtained, and the two contour points in the new image A are represented as: , The two contour points in the new image B are represented as follows: , The two contour points in the new image C are represented as follows: , ; 2) Removing error data based on the double mean error detection elimination method Before using the contour points of the new image in 1) to construct the ellipse, it is necessary to remove erroneous data or points with large errors; Removal method: (1) Calculate the distance from all points to the center point; (2) Calculate the mean of all distances; (3) Calculate the deviation of each distance from the mean and calculate the standard deviation; (4) Remove points with a distance deviation greater than twice the standard deviation; The formula for calculating standard deviation is as follows: （1） is the standard deviation, d is the distance of each point from the center point; is the mean of the distance of each point from the center point, n represents the number of rows; 3) Least square fitting ellipse parameters The nth layer can be regarded as a two-dimensional plane, and an ellipse can be fitted given six points and the center position. Curve equation: （2） Given six points whose coordinates represent the coordinates of the contour points of the three new images corresponding to the nth pixel row, and six plane points obtained after angle normalization from these six vector points, and substituted into the curve normal equation, the objective function is: （3） The least square adjustment process is such that The minimum, final calculated equation coefficients A, B, C, D, E, F values; Based on the values ​​of A, B, C, D, E, and F, the ellipse parameters can be calculated using the ellipse calculation formula: （4） （5） （6） （7） （8） Where a is the length of the major semi-axis of the ellipse, and b is the length of the minor semi-axis of the ellipse. It is a parameter. It is a parameter. The angle between the line connecting the origin and a point on the ellipse and the positive x-axis is also called the angle of elevation. Then, the area of ​​the pixel row containing the ellipse can be calculated: (i=0~N) (9) in, This represents the area of ​​the pixel row containing the ellipse. Pi Let be the length of the major semi-axis of the n-row ellipse. Let be the length of the minor semi-axis of the n-row ellipse; 4) Calculate the volume of the three-dimensional material droplet by layer-by-layer accumulation: （10） wherein, is the volume of the stereoscopic material drop.

2. The real-time monitoring method for liquid droplet features based on vision and single-pixel calibration according to claim 1, characterized in that, Step five, which involves calculating the calibration quality of a single stereo pixel, includes the following steps: First, in step four, weigh the calibration droplet product to obtain its quality. The calibration quality of a single stereo pixel is then: （11） wherein, is the quality of the single stereoscopic pixel; is the volume of each stereoscopic drop.

3. The method of real-time monitoring of liquid droplet characteristics based on vision and single-pixel calibration according to claim 2, characterized in that, Step six describes the calculation and monitoring of the real-time mass of the droplets, where the real-time mass of each droplet is calculated as follows: （12） wherein, is the real-time mass of each droplet, is the volume of the droplet for the subsequent measurement, is the single voxel calibrated mass.

4. The real-time monitoring method for liquid droplet features based on vision and single-pixel calibration according to claim 3, characterized in that, The three cameras are three cameras with the same resolution that are intersected at arbitrary angles.

Images