Centrifuge Liquid Flow Detection System Based on Fuzzy Recognition and Classification Technology

Abstract

The present invention discloses a centrifuge liquid flow detection system based on fuzzy recognition and classification technology, which relates to the technical field of centrifuge flow detection, and includes: collecting an image set, performing preprocessing and marking processing to obtain a segmented flow velocity; obtaining characteristic parameters; obtaining an optimized fuzzy model; obtaining the characteristic parameters of a real-time liquid image to perform flow detection on the real-time liquid. The present invention is used to solve the problem that in the existing methods for liquid flow detection, when the liquid surface is irregular or there is liquid fluctuation, the accuracy and stability of the data received by the sensor will be affected, resulting in a difference between the collected liquid flow and the actual liquid flow, thereby affecting the operation of the centrifuge equipment.

Description

Technical Field

[0001] The present invention relates to the technical field of centrifuge flow detection, and specifically to a centrifuge liquid flow detection system based on fuzzy recognition and classification technology. Background Art

[0002] In the aspect of liquid flow detection of centrifuges, methods such as the turbine method, the Karman vortex street method, and the velocity method are usually used. Liquid flow detection is to measure the flow velocity and flow rate of the liquid in the pipeline by using various flow meters; common flow meters include liquid mass flow meters and liquid flow meters; liquid mass flow meters use thermophysical properties for measurement and calculate the mass flow rate by detecting the change of heat by the fluid through sensors, and are applicable to fields such as chemical industry, petrochemical industry, metallurgy, food, and pharmaceuticals.

[0003] For the existing liquid flow detection of centrifuges, it is usually detected by using a liquid flow sensor or a flow velocity sensor, and the improvement direction is usually to improve the stability and convenience of sensor detection. For example, in the patent application with the publication number CN111989162A, a method for monitoring the lubricant flow rate on a centrifuge is disclosed. This solution monitors the lubricant flow rate through the bearing assembly by means of at least one measurement of the lubricant temperature at at least one measurement site on or in the device for supplying lubricant; or a remote data acquisition method is adopted. Although this method can improve the data acquisition efficiency during liquid flow detection, this improvement method is too dependent on the measurement results of the sensor. When the liquid surface is irregular or there is liquid fluctuation, the accuracy and stability of the data received by the sensor will be affected, resulting in a difference between the collected liquid flow rate and the actual liquid flow rate, and further affecting the operation of high-precision instruments such as centrifuges. For other improvements in liquid flow detection, it is usually to improve the acquisition period and acquisition duration of the flow sensor, or to avoid missed detection by gear rotation. This improvement method still depends on the measurement results of the sensor or can only collect whether there is liquid passing through, and cannot solve the problem that when the liquid surface is irregular or there is liquid fluctuation, the accuracy and stability of the data received by the sensor will be affected, resulting in a difference between the collected liquid flow rate and the actual liquid flow rate, and further affecting the operation of the centrifuge equipment. In view of this, it is necessary to improve the existing methods for liquid flow detection. Summary of the Invention

[0004] The present invention aims to solve at least one of the technical problems in the prior art to some extent. By proposing a centrifuge liquid flow rate detection system based on fuzzy recognition and classification technology, it is used to solve the problem that in the existing methods for liquid flow rate detection, when the liquid surface is irregular or there is liquid fluctuation, the accuracy and stability of the data received by the sensor will be affected, resulting in a difference between the collected liquid flow rate and the actual liquid flow rate, thereby affecting the operation of the centrifuge equipment.

[0005] To achieve the above object, the present application provides a centrifuge liquid flow rate detection system based on fuzzy recognition and classification technology, including an image acquisition and marking module, an image feature extraction module, a fuzzy recognition training module, and a flow rate detection module;

[0006] The image acquisition and marking module is used to obtain an acquisition image set, preprocess and mark the images in the acquisition image set, and obtain the segmentation flow rates corresponding to all liquids in the acquisition image set;

[0007] The image feature extraction module is used to obtain the characteristic parameters corresponding to the liquids in the acquisition image set based on the segmentation flow rates of the liquids, where the characteristic parameters include a first characteristic parameter and a second characteristic parameter;

[0008] The fuzzy recognition training module is used to train the training images with the characteristic parameters of the images in the acquisition image set, and obtain an optimized fuzzy model based on the training results;

[0009] The flow rate detection module is used to obtain the characteristic parameters of the real-time liquid image, and perform flow rate detection on the real-time liquid image using the optimized recognition model.

[0010] Further, the image acquisition and marking module includes an image acquisition and marking unit, and the image acquisition and marking unit is configured with an image acquisition and marking strategy, and the image acquisition and marking strategy includes:

[0011] Obtain images of various liquids at different flow rates based on big data, and store all the images in the acquisition image set; for all the images corresponding to any one liquid in the acquisition image set, sequentially record the images of the liquid with the flow rate increasing from small to large as the training images XT1 to training images XT k ; Obtain the training images corresponding to all liquids in the acquisition image set;

[0012] Use the preprocessing marking method to obtain the maximum marking area of the training images of the liquids; the preprocessing marking method is: for the training images XT1 to training images XT kFor any one of the training images in , the training image is grayscale processed, and the grayscale processed training image is denoted as the grayscale image; obtain the grayscale values of all pixel points in the grayscale image, and adjust the grayscale values of the pixel points with grayscale values greater than the determination value to the first standard value, and adjust the grayscale values of the pixel points with grayscale values less than or equal to the determination value to the second standard value.

[0013] Further, the image acquisition marking strategy further includes:

[0014] Denote the grayscale image after adjusting the grayscale values of the pixel points as the binary image; obtain the closed regions enclosed by the pixel points with the grayscale value of the second standard value in the binary image, and denote them as the image marking regions; obtain the areas of all the image marking regions, and denote the image marking region with the largest area as the largest marking region, and denote the area of the largest marking region as the largest marking area;

[0015] Obtain the largest marking areas corresponding to all the training images; establish a rectangular coordinate system, denoted as the flow velocity division coordinate system, where the unit of the X-axis of the flow velocity division coordinate system is flow velocity, and the unit of the Y-axis is area; based on the flow velocity of the liquid when each training image is taken and the largest marking area corresponding to each training image, mark corresponding points in the flow velocity division coordinate system, and connect the points with adjacent abscissas, and the obtained broken line graph is denoted as the flow velocity broken line graph;

[0016] Obtain the absolute values of the slopes of all the line segments in the flow velocity broken line graph, and denote the line segment corresponding to the maximum value among all the absolute values as the flow velocity segmentation line segment; denote the midpoint of the flow velocity segmentation line segment as the segmentation point, and denote the abscissa of the segmentation point as the segmentation flow velocity;

[0017] Obtain the segmentation points and segmentation flow velocities corresponding to all the liquids recorded in the image set.

[0018] Further, the image feature extraction module includes an image feature extraction unit, and the image feature extraction unit is configured with an image feature extraction strategy, and the image feature extraction strategy includes:

[0019] For any one kind of liquid in the acquired image set, for any one training image corresponding to the liquid, perform singular value decomposition on the training image using the singular value decomposition formula. The singular value decomposition formula is: , where XT is the training image, and both U and V are orthogonal matrices; Σ = diag(λ1, λ2,..., λ n ), that is, the singular value matrix with the singular values arranged from large to small as the diagonal elements; E i is the feature image; denote the singular value vector Q s obtained from the singular value decomposition formula as the first feature parameter of the training image, where the singular value vector Q s= (λ1, λ2,..., λ n ).

[0020] Furthermore, the image feature extraction strategy further includes:

[0021] Denote the number of rows of the pixel points of the training image as M, and the number of columns of the pixel points of the training image as N. Perform a cosine transform on the training image, and the algorithm for the cosine transform is: , where , I(m, n) is the pixel point data input based on the training image, u is a positive integer from 0 to M - 1, and v is a positive integer from 0 to N - 1; Denote the number P of non-zero coefficients obtained during the cosine transform e as the second feature parameter of the training image;

[0022] Obtain the first feature parameter and the second feature parameter corresponding to all the training images of all the liquids in the acquired image set.

[0023] Furthermore, the fuzzy recognition training module includes a fuzzy recognition training unit, and the fuzzy recognition training unit is configured with a fuzzy recognition training strategy, and the fuzzy recognition training strategy includes:

[0024] Obtain a TSK fuzzy model, set the number of neurons in the input layer of the TSK fuzzy model to n + 1, and set the number of neurons in the output layer to 1;

[0025] For any one training image in the acquired image set, use the first feature parameter and the second feature parameter of the training image as the input signals of the input layer of the TSK fuzzy model, and denote the output result of the neuron in the output layer as the fuzzy degree value of the training image;

[0026] Obtain the fuzzy degree values corresponding to all the training images, and denote the TSK fuzzy model at this time as the preferred fuzzy model.

[0027] Furthermore, the fuzzy recognition training strategy further includes:

[0028] For any one liquid in the acquired image set, establish a table with T rows × R columns, denoted as the parameter regularization table. Among them, except for the first cell in the top row of the parameter regularization table, fill in the first feature parameter, the second feature parameter, and the fuzzy degree value from left to right in sequence. Except for the first cell in the leftmost column of the parameter regularization table, fill in the liquid flow rates in all the images corresponding to the liquid in the acquired image set from top to bottom in sequence. Among them, the flow rates in the leftmost column of the parameter regularization table increase from top to bottom;

[0029] Fill in the first feature parameter, the second feature parameter, and the fuzzy degree value corresponding to all the training images of the liquid into the parameter regularization table; Obtain the parameter regularization tables corresponding to all the liquids recorded in the acquired image set.

[0030] Further, the flow rate detection module includes a real-time detection unit, and the real-time detection unit is configured with a real-time detection strategy, and the real-time detection strategy includes:

[0031] Denote the liquid used for flow velocity detection as the real-time liquid; use a camera to take a picture of the real-time liquid, and denote the obtained image as the real-time liquid image; based on the image acquisition marking module, obtain the maximum marking area in the real-time liquid image, and denote it as the real-time maximum area;

[0032] Obtain the flow velocity line graph corresponding to the real-time liquid in the acquired image set, and denote it as the comparison line graph; denote the point on the comparison line graph whose abscissa is equal to the real-time maximum area as the real-time point. When the ordinate of the real-time point is greater than the segmentation flow velocity, denote the row where the cell with a liquid flow velocity greater than the segmentation flow velocity in the leftmost column of the parameter regularization table of the real-time liquid as the row to be compared; when the ordinate of the real-time point is less than or equal to the segmentation flow velocity, denote the row where the cell with a liquid flow velocity less than or equal to the segmentation flow velocity in the leftmost column of the parameter regularization table of the real-time liquid as the row to be compared.

[0033] Further, the real-time detection strategy further includes:

[0034] Based on the image feature extraction module and the optimized fuzzy model, obtain the first feature parameter, the second feature parameter, and the fuzzy degree value corresponding to the real-time liquid image; respectively obtain the absolute values of the differences between the first feature parameter, the second feature parameter, and the fuzzy degree value of each row in all rows to be compared and the first feature parameter, the second feature parameter, and the fuzzy degree value of the real-time liquid image, and denote them as the first difference, the second difference, and the fuzzy difference respectively; denote the sum of the first difference, the second difference, and the fuzzy difference as the comparison difference.

[0035] Further, the real-time detection strategy further includes:

[0036] Obtain the comparison differences corresponding to all rows to be compared, and denote the row corresponding to the minimum value of all comparison differences as the minimum error row; denote the liquid flow velocity in the leftmost cell of the minimum error row as the liquid flow velocity of the real-time liquid; denote the product of the liquid flow velocity of the real-time liquid and the detection time of the real-time liquid as the liquid flow rate of the real-time liquid.

[0037] Advantages of the present invention: First, the present invention collects an image set, preprocesses and marks the images in the collected image set to obtain the segmented flow rate; then, based on the segmented flow rate, characteristic parameters corresponding to the liquid in the collected image set are obtained. The advantage of this is that by obtaining the segmented flow rate, it is possible to preliminarily distinguish between images of the same liquid with relatively fast and slow flow rates, preventing the influence on the analysis results caused by considering low-flow-rate images when analyzing high-flow-rate liquids or considering high-flow-rate images when analyzing low-flow-rate liquids; by obtaining the characteristic parameters, it is possible to obtain the characteristics of the same liquid at different flow rates, which helps to analyze the liquid to be detected at different flow rates during subsequent analysis and makes the analysis results more accurate.

[0038] The present invention also uses the characteristic parameters to train the training images and obtains an optimized fuzzy model based on the training results; finally, the characteristic parameters of the real-time liquid image are obtained, and the optimized recognition model is used to detect the flow rate of the real-time liquid image. The advantage of this is that by obtaining the optimized fuzzy model, it is possible to effectively handle the blurring and uncertainty problems on the liquid surface, not only providing a reliable liquid flow rate estimation, but also reducing the dependence on sensor data, improving the robustness and stability of the flow rate detection system, and preventing the problem that the accuracy and stability of the data received by the sensor are affected, resulting in a difference between the collected liquid flow rate and the actual liquid flow rate, thereby affecting the operation of high-precision instruments such as centrifuges. Brief Description of the Drawings

[0039] Figure 1 is the principle block diagram of the system of the present invention;

[0040] Figure 2 is the schematic diagram of the flow rate broken line graph of the present invention;

[0041] Figure 3 is the schematic diagram of the position of the real-time point of the present invention. Detailed Embodiments

[0042] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0043] Please refer to Figure 1 as shown, the present application provides a liquid flow rate detection system for fuzzy recognition and classification technology, including an image acquisition and marking module, an image feature extraction module, a fuzzy recognition and training module, and a flow rate detection module;

[0044] The image acquisition and marking module is used to obtain an acquisition image set, preprocess and mark the images in the acquisition image set, and obtain the segmentation flow rates corresponding to all liquids in the acquisition image set; the image acquisition and marking module includes an image acquisition and marking unit, and the image acquisition and marking unit is configured with an image acquisition and marking strategy, and the image acquisition and marking strategy includes:

[0045] Obtain images of various liquids at different flow rates based on big data, and store all the images in the acquisition image set; for the images of any one liquid corresponding to all flow rates in the acquisition image set, the images of the liquid are sequentially recorded as training images XT1 to training image XT according to the ascending order of the liquid flow rate in the image k ; Obtain the training images corresponding to all liquids in the acquisition image set;

[0046] In the specific implementation process, images corresponding to different flow rates can be obtained for the liquid whose flow rate needs to be detected as required. For example, if the liquids whose flow rates need to be detected are water and glycerol, then images of water and glycerol at different flow rates can be obtained based on big data; if the flow rate range of water during flow rate detection is 2 m / s to 5 m / s, then images of water at flow rates of 2 m / s, 3 m / s, 4 m / s, and 5 m / s can be obtained and stored in the acquisition image set for subsequent analysis;

[0047] Use the preprocessing marking method to obtain the maximum marking area of the training image of the liquid; the preprocessing marking method is: for any one of the training images from training image XT1 to training image XT k in, perform grayscale processing on the training image, and record the grayscale processed training image as a grayscale image; obtain the grayscale values of all pixel points in the grayscale image, and adjust the grayscale values of the pixel points whose grayscale values are greater than the determination value to the first standard value, and adjust the grayscale values of the pixel points whose grayscale values are less than or equal to the determination value to the second standard value;

[0048] In the specific implementation process, the first standard value is 255 and the second standard value is 0; by adjusting the grayscale values of all pixel points in the training image to 255 or 0, it helps to obtain a more obvious contour in the training image XT, thereby helping to preliminarily divide all liquid images XT of the liquid;

[0049] Record the grayscale image after adjusting the grayscale values of the pixel points as a binary image; obtain the closed regions enclosed by the pixel points with the grayscale value of the second standard value in the binary image, and record them as image marking regions; obtain the areas of all image marking regions, and record the image marking region with the largest area as the largest marking region, and record the area of the largest marking region as the maximum marking area;

[0050] Obtain the maximum marked area corresponding to all training images; establish a rectangular coordinate system, denoted as the flow velocity division coordinate system, where the unit of the X-axis of the flow velocity division coordinate system is flow velocity, and the unit of the Y-axis is area; based on the flow velocity of the liquid at the time of shooting each training image and the maximum marked area corresponding to each training image, mark corresponding points in the flow velocity division coordinate system, and connect the points with adjacent abscissas. The obtained line graph is denoted as the flow velocity line graph; in the specific implementation process, for example, in a data processing, the liquid to be analyzed is water, and all training images XT of water are images at flow velocities of 2 m / s, 3 m / s, 4 m / s, and 5 m / s respectively. The maximum marked areas obtained through analysis are 5 cm², 4 cm², 3 cm², and 1 cm² respectively. The obtained flow velocity line graph is as Figure 2 shown, where the broken line ZZ1 is the flow velocity line graph;

[0051] Obtain the absolute values of the slopes of all line segments in the flow velocity line graph, and denote the line segment corresponding to the maximum value among all absolute values as the flow velocity segmentation line segment; denote the midpoint of the flow velocity segmentation line segment as the segmentation point, and denote the abscissa of the segmentation point as the segmentation flow velocity; in the specific implementation process, through the analysis of the broken line ZZ1, it can be obtained that the line segment formed by point DD1 and point DD2 is the line segment with the largest absolute value of the slope, that is, the flow velocity segmentation line segment, and the midpoint of the line segment formed by point DD1 and point DD2 is the segmentation point; by obtaining the flow velocity segmentation line segment, the point with the largest difference between different flow velocities of the liquid can be obtained, which helps to initially distinguish the images with faster and slower flow velocities of the same liquid, thus assisting subsequent analysis;

[0052] Obtain the segmentation points and segmentation flow velocities corresponding to all liquids recorded in the image set.

[0053] The image feature extraction module is used to obtain the characteristic parameters corresponding to the liquid in the acquisition image set based on the segmentation flow velocity of the liquid, where the characteristic parameters include the first characteristic parameter and the second characteristic parameter; the image feature extraction module includes an image feature extraction unit, and the image feature extraction unit is configured with an image feature extraction strategy, and the image feature extraction strategy includes:

[0054] For any liquid in the acquisition image set, for any training image corresponding to the liquid, use the singular value decomposition formula to perform singular value decomposition on the training image. The singular value decomposition formula is: , where XT is the training image, and both U and V are orthogonal matrices; Σ = diag(λ1, λ2,..., λ n ), that is, the singular value matrix with the singular values arranged from large to small as diagonal elements; E i is the feature image; denote the singular value vector Q s obtained from the singular value decomposition formula as the first characteristic parameter of the training image, where the singular value vector Qs = (λ1, λ2,..., λ n ); In the specific implementation process, the value of n can be obtained according to the matrix corresponding to Σ during actual decomposition. For example, during a data processing, λ1 to λ n are respectively (5, 0, 0), (0, 4, 0), (0, 0, 3), (6, 0, 0), (0, 1, 0), (0, 0, 2), (7, 0, 0), and (0, 1, 1). Then, through analysis, the value of n is 8, and the singular value vector Q s = [(5, 0, 0), (0, 4, 0), (0, 0, 3), (6, 0, 0), (0, 1, 0), (0, 0, 2), (7, 0, 0), (0, 1, 1)];

[0055] Denote the number of rows of the pixel points of the training image as M, and the number of columns of the pixel points of the training image as N. Perform cosine transform on the training image. The algorithm of the cosine transform is: , where, , I(m, n) is the pixel point data input based on the training image, u is a positive integer from 0 to M - 1, and v is a positive integer from 0 to N - 1; Denote the number P e of non-zero coefficients obtained during the cosine transform as the second characteristic parameter of the training image. In the specific implementation process, for example, during a data processing, the number of non-zero coefficients obtained through cosine transform is 4, then set the value of M e to 4; Thus, the first characteristic parameter can be obtained as [(5, 0, 0), (0, 4, 0), (0, 0, 3), (6, 0, 0), (0, 1, 0), (0, 0, 2), (7, 0, 0), (0, 1, 1)], and the second characteristic parameter is 4;

[0056] Obtain the first characteristic parameter and the second characteristic parameter corresponding to all the training images of all liquids in the collected image set.

[0057] The fuzzy recognition training module is used to train the training images using the characteristic parameters of the images in the collected image set, and obtain an optimal fuzzy model based on the training results. The fuzzy recognition training module includes a fuzzy recognition training unit, and the fuzzy recognition training unit is configured with a fuzzy recognition training strategy. The fuzzy recognition training strategy includes:

[0058] Obtain a TSK fuzzy model, set the number of neurons in the input layer of the TSK fuzzy model to n + 1, and set the number of neurons in the output layer to 1; In the specific implementation process, for example, during a data processing, when the value of n is obtained as 8, the number of neurons in the input layer of the TSK fuzzy model can be set to 9 for the input of the first characteristic parameter and the second characteristic parameter;

[0059] In this embodiment, the hidden layer of the TSK fuzzy model contains 15 neurons; the maximum number of iterations is 50,000, the objective function is the mean square error function, the minimum mean square error is set to 0.01, the activation function is the Sigmoid function, and the learning step size is 0.1; all feature elements are normalized to -1 to 1 and then input. The label of the clear image block is 0, and the blurred image is marked as 1; in specific implementation, the above parameters can be adjusted according to the actually obtained TSK fuzzy model to meet the actual analysis requirements;

[0060] For any training image in the collected image set, the first feature parameter and the second feature parameter of the training image are used as the input signals of the input layer of the TSK fuzzy model, and the output result of the neuron in the output layer is recorded as the blur degree value of the training image; in the specific implementation process, for example, in a data processing, the values input by the neurons in the input layer in the TSK fuzzy model obtained based on this embodiment are (5, 0, 0), (0, 4, 0), (0, 0, 3), (6, 0, 0), (0, 1, 0), (0, 0, 2), (7, 0, 0), (0, 1, 1), and 4. After calculation, the value of the neuron in the output layer is 10, then 10 can be recorded as the blur degree value;

[0061] Obtain the blur degree values corresponding to all training images, and record the TSK fuzzy model at this time as the preferred fuzzy model;

[0062] For any liquid in the collected image set, establish a table with T rows × R columns, denoted as the parameter regularization table. Among them, except for the first cell in the top row of the parameter regularization table, the first feature parameter, the second feature parameter, and the blur degree value are filled in from left to right in sequence. Except for the first cell in the leftmost column of the parameter regularization table, the liquid flow rates in all the images corresponding to the liquid in the collected image set are filled in from top to bottom in sequence. Among them, the flow rates in the leftmost column of the parameter regularization table increase from top to bottom; in the specific implementation process, for example, in a data processing, the obtained parameter regularization table is shown in Table 1: Parameter Regularization Table;

[0063] Table 1: Parameter Regularization Table

[0064] Liquid flow rate First characteristic parameter Second characteristic parameter Fuzziness value 2 m / s （5,0,0），（0,4,0），（0,0，3），（6,0,0），（0,1,0），（0,0,2），（7,0,0），（0,1,1） 4 10 3 m / s （6,0,0），（0,5,0），（0,0，4），（7,0,0），（0,2,0），（0,0,4），（7,0,0），（0,0,1） 8 20 4 m / s （8,0,0），（0,9,0），（0,0，7），（8,0,0），（0,2,0），（0,0,2），（7,0,0），（0,4,1） 16 40 5 m / s （10,0,0），（0,8,0），（0,0，8），（7,0,0），（0,4,2），（0,0,2），（7,0,0），（0,1,1） 30 70

[0065] In this embodiment, the value of R is fixed at 4, and the value of T can be determined according to the number of liquid flow rates in all the images corresponding to the liquid. For example, if the liquid flow rates in all the images corresponding to water are 2 m / s, 3 m / s, 4 m / s, and 5 m / s respectively, then the value of T can be set to 5;

[0066] Fill the first characteristic parameter, the second characteristic parameter, and the blurring degree value corresponding to all training images of the liquid into the parameter regularization table; obtain the parameter regularization tables corresponding to all liquids recorded in the acquisition image set; in the specific implementation process, by obtaining the parameter regularization table, the first characteristic parameter, the second characteristic parameter, and the blurring degree value corresponding to the same liquid at different flow rates can be obtained, which helps to quickly, efficiently, and accurately obtain the flow rate of the liquid to be detected during the flow rate detection of the liquid.

[0067] The flow rate detection module is used to obtain the characteristic parameters of the real-time liquid image and perform flow rate detection on the real-time liquid image using the preferred recognition model; the flow rate detection module includes a real-time detection unit, and the real-time detection unit is configured with a real-time detection strategy, and the real-time detection strategy includes:

[0068] The liquid used for flow rate detection is denoted as the real-time liquid; use a camera to capture the real-time liquid and denote the obtained image as the real-time liquid image; based on the image acquisition marking module, obtain the maximum marked area in the real-time liquid image, denoted as the real-time maximum area;

[0069] In the specific implementation process, for example, during a data processing, the real-time liquid is water and the real-time maximum area is 2 cm². Through analysis, it can be obtained that the position of the real-time point is Figure 3 the point DD3 in, and at this time, through the above analysis of this embodiment, it can be obtained that the real-time point coincides with the segmentation point, that is, the ordinate of the real-time point is equal to the segmentation flow rate. The row where the cell with a flow rate less than XX1 in the leftmost column of the parameter regularization table corresponding to water is located can be denoted as the row to be compared;

[0070] Obtain the flow rate broken line graph corresponding to the real-time liquid in the acquisition image set, denoted as the comparison broken line graph; denote the point with the abscissa equal to the real-time maximum area in the comparison broken line graph as the real-time point. When the ordinate of the real-time point is greater than the segmentation flow rate, the row where the cell with a liquid flow rate greater than the segmentation flow rate in the leftmost column of the parameter regularization table of the real-time liquid is located is denoted as the row to be compared; when the ordinate of the real-time point is less than or equal to the segmentation flow rate, the row where the cell with a liquid flow rate less than or equal to the segmentation flow rate in the leftmost column of the parameter regularization table of the real-time liquid is located is denoted as the row to be compared;

[0071] Obtain the first characteristic parameter, the second characteristic parameter, and the fuzziness value corresponding to the real-time liquid image based on the image feature extraction module and the optimized fuzzy model; respectively obtain the absolute values of the differences between the first characteristic parameter, the second characteristic parameter, and the fuzziness value of each row in all rows to be compared and the first characteristic parameter, the second characteristic parameter, and the fuzziness value of the real-time liquid image, and denote them as the first difference, the second difference, and the fuzziness difference respectively; denote the sum of the first difference, the second difference, and the fuzziness difference as the comparison difference; in the specific implementation process, for example, in a data processing, the first characteristic parameter, the second characteristic parameter, and the fuzziness value corresponding to a row to be compared are obtained as (5,0,0), (0,4,0), (0,0,3), (6,0,0), (0,1,0), (0,0,2), (7,0,0) and (0,1,1), 4, and 10 respectively; the first characteristic parameter, the second characteristic parameter, and the fuzziness value of the real-time liquid image are (5,0,0), (0,4,0), (0,0,3), (6,0,0), (0,1,0), (0,0,2), (7,0,0) and (0,1,1), 3, and 15 respectively. Then, through calculation, the comparison difference is 6;

[0072] Obtain the comparison differences corresponding to all rows to be compared, and denote the row corresponding to the minimum value of all comparison differences as the row with the minimum error; denote the liquid flow rate in the leftmost cell of the row with the minimum error as the liquid flow rate of the real-time liquid; denote the product of the liquid flow rate of the real-time liquid and the detection time of the real-time liquid as the liquid flow of the real-time liquid; in the specific implementation, by obtaining the row with the minimum error, the flow rate with the smallest difference from the real-time liquid flow rate recorded in the parameter regularization table can be obtained, so that the obtained liquid flow rate of the real-time liquid is more accurate, and further the obtained liquid flow of the real-time liquid is more accurate.

[0073] Working principle: First, collect an image set, preprocess and label the images in the collected image set to obtain the segmented flow rate; then obtain the characteristic parameters corresponding to the liquid in the collected image set based on the segmented flow rate, also use the characteristic parameters to train the training images, and obtain the optimized fuzzy model based on the training results; finally, obtain the characteristic parameters of the real-time liquid image, and use the optimized recognition model to detect the flow rate of the real-time liquid image.

[0074] Through the description of the above embodiments, the embodiments of the present invention can be provided as a method, a system or a computer program product. Based on such understanding, the above technical solution, in essence, or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., including several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0075] In the embodiments provided in the present application, it should be understood that the disclosed system or method can be implemented in other ways. The above-described embodiments are merely illustrative. For example, the division of modules or units is only a logical function division, and there may be other division methods in actual implementation. For another example, multiple modules or units can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the displayed or discussed coupling or direct coupling or communication connection between each other can be through some communication interfaces. The indirect coupling or communication connection of systems, modules, and units can be electrical, mechanical, or other forms.

[0076] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements for some of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of each embodiment of the present application.

Claims

1. A centrifuge liquid flow detection system based on fuzzy recognition and classification technology, characterized in that: It includes image acquisition and marking module, image feature extraction module, fuzzy recognition training module and flow detection module; The image acquisition and marking module is used to acquire an acquired image set, preprocess and mark the images in the acquired image set, and acquire the segmented flow velocities corresponding to all liquids in the acquired image set; The image feature extraction module is used to obtain the feature parameters corresponding to the liquid in the collected image set based on the segmented flow velocity of the liquid, wherein the feature parameters include a first feature parameter and a second feature parameter; The fuzzy recognition training module is used to train the training images using the feature parameters of the images in the collected image set, and obtain the optimal fuzzy model based on the training results; The flow detection module is used to obtain characteristic parameters of the real-time liquid image and use the optimal recognition model to perform flow detection on the real-time liquid image; The image acquisition and marking module includes an image acquisition and marking unit, and the image acquisition and marking unit is configured with an image acquisition and marking strategy, and the image acquisition and marking strategy includes: Based on big data, images of various liquids at different flow rates are obtained, and all images are stored in a collection of images; for images of all flow rates corresponding to any liquid in the collection of images, the images are recorded in order from small to large as training images XT1 to training images XT2 based on the flow rates of the liquids in the images. k ; Obtain training images corresponding to all liquids in the collected image set; Use the preprocessing labeling method to obtain the maximum labeling area of ​​the liquid training image; the preprocessing labeling method is: for training images XT1 to training images XT k any training image in the grayscale image, grayscale the training image, and record the grayscale processed training image as the grayscale image; obtain the grayscale values ​​of all pixels in the grayscale image, and adjust the grayscale values ​​of pixels whose grayscale values ​​are greater than the judgment value to the first standard value, and adjust the grayscale values ​​of pixels whose grayscale values ​​are less than or equal to the judgment value to the second standard value; Image acquisition and labeling strategies also include: The grayscale image after adjusting the grayscale value of the pixel is recorded as a binary image; the closed area in the binary image that is surrounded by the pixel points whose grayscale value is the second standard value is obtained, and recorded as the image marked area; the area of ​​all the image marked areas is obtained, and the image marked area with the largest area is recorded as the maximum marked area, and the area of ​​the maximum marked area is recorded as the maximum marked area; Obtain the maximum marked area corresponding to all training images; establish a plane rectangular coordinate system, recorded as a flow velocity division coordinate system, wherein the unit of the X-axis of the flow velocity division coordinate system is the flow velocity, and the unit of the Y-axis is the area; based on the liquid flow velocity of each training image of the liquid when it is captured and the maximum marked area corresponding to each training image, perform corresponding punctuation in the flow velocity division coordinate system, and connect adjacent punctuation points on the horizontal axis to obtain a line graph recorded as a flow velocity line graph; Obtain the absolute values ​​of the slopes of all line segments in the velocity line graph, and record the line segment corresponding to the maximum value among all absolute values ​​as the velocity segmentation line segment; record the midpoint of the velocity segmentation line segment as the segmentation point, and mark the horizontal coordinate of the segmentation point as the segmentation velocity; Get the segmentation points and segmentation flow rates corresponding to all liquids recorded in the image set.

2. The centrifuge liquid flow detection system based on fuzzy recognition and classification technology according to claim 1 is characterized in that: The image feature extraction module includes an image feature extraction unit, and the image feature extraction unit is configured with an image feature extraction strategy, and the image feature extraction strategy includes: For any liquid in the collected image set and any training image corresponding to the liquid, the singular value decomposition formula is used to perform singular value decomposition on the training image. The singular value decomposition formula is: , where XT is the training image, U and V are both orthogonal matrices; Σ=diag(λ1,λ2,...,λ n ), which is the singular value matrix whose diagonal elements are composed of singular values ​​arranged from large to small; E i is the feature image; The singular value vector Q obtained by the singular value decomposition formula s Denoted as the first feature parameter of the training image, where the singular value vector Q s =(λ1,λ2,...,λ n ).

3. The centrifuge liquid flow detection system based on fuzzy recognition and classification technology according to claim 2 is characterized in that: Image feature extraction strategies also include: The number of rows of pixels in the training image is recorded as M, and the number of columns of pixels in the training image is recorded as N. The training image is cosine transformed. The algorithm of cosine transformation is: ,in, , I (m, n) is the pixel data based on the training image input, u is a positive integer from 0 to M-1, and v is a positive integer from 0 to N-1; the number of non-zero coefficients obtained in the cosine transformation process P e Denoted as the second characteristic parameter of the training image; The first characteristic parameters and the second characteristic parameters corresponding to all training images of all liquids in the collected image set are obtained.

4. The centrifuge liquid flow detection system based on fuzzy recognition and classification technology according to claim 3 is characterized in that: The fuzzy recognition training module includes a fuzzy recognition training unit, and the fuzzy recognition training unit is configured with a fuzzy recognition training strategy. The fuzzy recognition training strategy includes: Obtain the TSK fuzzy model, set the number of neurons in the input layer of the TSK fuzzy model to n+1, and set the number of neurons in the output layer to 1; For any training image in the collected image set, the first characteristic parameter and the second characteristic parameter of the training image are used as input signals of the input layer of the TSK fuzzy model, and the output result of the neurons in the output layer is recorded as the fuzziness value of the training image; Obtain the blur degree values ​​corresponding to all training images, and record the TSK blur model at this time as the preferred blur model.

5. The centrifuge liquid flow detection system based on fuzzy recognition and classification technology according to claim 4 is characterized in that: Fuzzy recognition training strategies also include: For any liquid in the collected image set, a table of T rows × R columns is established, which is recorded as a parameter regularization table, wherein the top row of the parameter regularization table is filled with the first characteristic parameter, the second characteristic parameter and the blur value from left to right except the first cell, and the leftmost column of the parameter regularization table is filled with the liquid flow rate in all images corresponding to the liquid in the collected image set from top to bottom except the first cell, wherein the flow rate in the leftmost column of the parameter regularization table increases from top to bottom; Fill the first characteristic parameter, the second characteristic parameter and the blur degree value corresponding to all the training images of the liquid into the parameter regularization table; obtain the parameter regularization table corresponding to all the liquids recorded in the collected image set.

6. The centrifuge liquid flow detection system based on fuzzy recognition and classification technology according to claim 5 is characterized in that: The traffic detection module includes a real-time detection unit, which is configured with a real-time detection strategy. The real-time detection strategy includes: The liquid used for flow rate detection is recorded as real-time liquid; the real-time liquid is photographed using a camera, and the obtained image is recorded as a real-time liquid image; the maximum marked area in the real-time liquid image is obtained based on the image acquisition and marking module, and recorded as the real-time maximum area; Obtain a flow rate line graph corresponding to the real-time liquid in the collected image set, and record it as a comparison line graph; record the point whose horizontal coordinate is equal to the real-time maximum area in the comparison line graph as a real-time point, and when the vertical coordinate of the real-time point is greater than the segmentation flow rate, record the row where the cell whose liquid flow rate is greater than the segmentation flow rate in the leftmost column of the real-time liquid parameter regularization table is located as a row to be compared; when the vertical coordinate of the real-time point is less than or equal to the segmentation flow rate, record the row where the cell whose liquid flow rate is less than or equal to the segmentation flow rate in the leftmost column of the real-time liquid parameter regularization table is located as a row to be compared.

7. The centrifuge liquid flow detection system based on fuzzy recognition and classification technology according to claim 6 is characterized in that: Real-time detection strategies also include: Based on the image feature extraction module and the preferred fuzzy model, the first feature parameter, the second feature parameter and the fuzziness degree value corresponding to the real-time liquid image are obtained; the absolute value of the difference between the first feature parameter, the second feature parameter and the fuzziness degree value of each row in all the rows to be compared and the first feature parameter, the second feature parameter and the fuzziness degree value of the real-time liquid image are obtained respectively, and recorded as the first difference, the second difference and the fuzzy difference respectively; the sum of the first difference, the second difference and the fuzzy difference is recorded as the comparison difference.

8. The centrifuge liquid flow detection system based on fuzzy recognition and classification technology according to claim 7 is characterized in that: Real-time detection strategies also include: Obtain the comparison differences corresponding to all rows to be compared, and record the row corresponding to the minimum value of all comparison differences as the minimum error row; record the liquid flow rate in the leftmost cell of the minimum error row as the liquid flow rate of the real-time liquid; and record the product of the liquid flow rate of the real-time liquid and the detection time of the real-time liquid as the liquid flow rate of the real-time liquid.

Images