Analysis method and device for optimal setting values of multiple parameters based on high-frequency machine vision learning
Through the multi-parameter optimal setpoint analysis method of high-frequency machine vision learning, combined with random forest and Bayesian optimization algorithm, the problems of model overfitting and image quality reduction are solved, and the accuracy and efficiency of product quality detection are improved. .
Patent Information
- Application Number
- CN202411306635.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-09-19
- Publication Date
- 2024-11-29
- Estimated Expiration
- 2044-09-19
AI Technical Summary
In the product quality inspection, existing high-frequency machine vision learning technology has caused the model to be overfitted due to insufficient training data or improper feature selection, and image sensor noise and environmental interference have caused the quality of the acquired image to decline, which in turn affects the detection accuracy.
Using multi-parameter optimal setpoint analysis method based on high-frequency machine vision learning, the history and batch data of the same type of products are collected, and divided into test sets and training sets, the influencing factors are determined and correlation verification is performed, and random The forest algorithm builds a model, combines Bayesian optimization algorithm to optimize the model, and improves the accuracy of the image acquisition device through precise adjustment and cleaning mechanisms.
It improves the generalization ability and robustness of the model, enhances the acquisition accuracy and detection accuracy of image data, and ensures the accuracy and efficiency of product quality detection.
Smart Images

Figure CN118823523B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of image data processing, and in particular to an analysis method for optimal setting values of multiple parameters based on high-frequency machine vision learning. Background Art
[0002] High-frequency machine vision learning combines machine learning and computer vision technology, and can analyze and process large amounts of image data in real time. The analysis method of multi-parameter optimal setting values aims to optimize system performance by optimizing the combination of multiple parameters. Combining the two can achieve efficient decision support in complex application scenarios.
[0003] However, in the prior art, high-frequency machine learning and computer vision technologies are used to collect image information of products, and then the quality of the products is judged by comparing and analyzing the set parameters with the collected parameters. When quality inspection is performed on products produced by automated production, due to insufficient training data or improper feature selection, the model may overfit the training set, resulting in difficulty in setting the optimal parameters. At the same time, due to interference from factors such as noise and environmental interference in image sensors, the quality of the collected images is reduced. When collecting images of products on the production line, when the products are in motion following the production line, the images collected by the camera are prone to blurring, which also leads to a reduction in the quality of the collected images, and ultimately causes a decrease in the accuracy of product quality inspection. Summary of the invention
[0004] The purpose of the present invention is to provide an analysis method for multi-parameter optimal setting values based on high-frequency machine vision learning, so as to solve the problem proposed in the above background technology that due to insufficient training data or improper feature selection, the model may overfit the training set, resulting in difficulty in setting the optimal parameters. At the same time, due to the interference of factors such as noise and environmental interference in the image sensor, the quality of the collected image is reduced. When capturing images of products on the production line, when the products are in motion following the production line, the images captured by the camera are prone to blur, which also leads to a decrease in the quality of the collected image, and ultimately causes a decrease in the accuracy of product quality inspection.
[0005] To achieve the above object, the present invention provides the following technical solution: an analysis method for optimal setting values of multiple parameters based on high-frequency machine vision learning, comprising the following contents:
[0006] S1. Data source collection: The data source consists of historical data of the same type of products and the same batch data of the same type of products;
[0007] S2, dividing the collected data sources into a test set and a training set;
[0008] S3. Determine the influencing factors: Divide the influencing factors according to human factors, environmental factors and equipment factors;
[0009] S4. Conduct correlation verification on the impact factors. The correlation verification uses the variance inflation factor method for multicollinearity verification, and factors with a VIF value less than 10 are screened. The calculation formula of the VIF is as follows: ;
[0010] In the formula: is the coefficient of determination between the i-th impact factor and other impact factors;
[0011] S5. Model construction: Use the collected samples of the training set for model construction. The model construction uses the random forest algorithm, and the random forest algorithm is used to select an optimal feature from the training set sample data as the splitting attribute of this node;
[0012] S6. Model evaluation: Use the collected samples of the test set to evaluate the model constructed;
[0013] S7. Model optimization: Used to optimize the models that fail the model evaluation.
[0014] Preferably, in step S1, the ratio of the historical data of the same type of product in the data source to the same batch data of the same type of product is 1:1.
[0015] Preferably, in step S1, the image information of the same batch data of the same type of product is collected using the image acquisition device of the production line.
[0016] Preferably, in step S2, the ratio of the data samples of the test set and the training set is 3:7.
[0017] Preferably, in step S5, the random forest algorithm includes the following:
[0018] S51. Bootstrap sampling: Randomly draw k samples from the collected samples D of the training set with replacement as the training set K for a decision tree;
[0019] S52. Random feature selection: Randomly select m features from the candidate M features, where m < M as the alternative feature subset for decision-making under the current node; then select an optimal feature from the selected m features using the information gain strategy as the splitting attribute of this node. The information gain Gain(A, K) of feature A for the data set K is the difference between the entropy Ent(K) of the data set and the conditional entropy CEnt(A, K) of feature A. The formula is as follows: ;
[0020] In the formula: is the proportion of unqualified and qualified samples in the current data set K; represents the number of samples with the value of feature A being v;
[0021] S53, decision tree growth: During the decision tree formation process, each node repeats step S2 to split until there are no new features, and the split ends;
[0022] S54, constructing a random forest: constructing N decision trees according to steps S1 to S3;
[0023] S55. The final prediction result is classified by voting on the prediction results of all decision trees to determine an optimal feature parameter.
[0024] Preferably, in step S53, the decision tree growth adopts XGBoost algorithm to reduce the complexity of the decision tree and obtain the optimal objective function of the decision tree. , the optimal objective function as follows: ;
[0025] Where: , for all samples belonging to the jth leaf node The sum of is the first-order derivative of the loss function; , for all samples belonging to the jth leaf node The sum of is the second-order derivative of the loss function; It means that the i-th sample is located at the j-th leaf node; T is the number of leaf nodes; is the weight value of the leaf node; It is a hyperparameter used to control the intensity of the penalty;
[0026] The optimal objective function is used to measure the quality of the tree structure. The smaller the value, the better the tree structure.
[0027] Preferably, in step S6, the model evaluation uses a confusion matrix and a ROC curve to evaluate the model constructed by the model, the confusion matrix is used to calculate the accuracy, precision and recall rate of the acquired data to evaluate the model, and the ROC curve is used to quantitatively judge the accuracy of model detection.
[0028] Preferably, in step S7, the model optimization uses a Bayesian optimization algorithm to optimize the constructed model.
[0029] Preferably, in step S7, the Bayesian optimization algorithm optimizes the constructed model including the following contents:
[0030] S71. First, establish prior knowledge by approximating the target function through the Gaussian regression model;
[0031] S72, then by observing the output of the objective function at different input points, update this prior knowledge to form a posterior distribution;
[0032] S73. Based on the posterior distribution, use the PI acquisition function to perform the next sampling, add the sampled data to the data set, calculate the probability proxy function, and repeat the above steps. After iterating for a set number of times, select the global optimal solution.
[0033] The device of multi-parameter optimal setting value based on high-frequency machine vision learning comprises an installation mechanism, an adjustment mechanism and a cleaning mechanism, wherein the installation mechanism comprises a No. 1 slide rail, wherein the No. 1 slide rail is slidably connected to a No. 1 slider inside the No. 1 slide rail, and a No. 1 servo motor is fixedly connected to the top of the No. 1 slide rail, an No. 1 adjustment screw is fixedly connected to the output end of the No. 1 servo motor, and the No. 1 adjustment screw is threadedly connected to the No. 1 slider, a No. 2 slide rail is fixedly connected to the side of the No. 1 slider, a No. 2 slide rail is slidably connected to the No. 2 slider inside the No. 2 slide rail, and a No. 2 servo motor is fixedly connected to one end of the No. 2 slide rail, a No. 2 adjustment screw is fixedly connected to the output end of the No. 2 servo motor, and the No. 2 adjustment screw is threadedly connected to the No. 2 slider, and the No. 2 servo motor is fixedly connected to the No. 2 adjustment screw. The adjusting mechanism includes a No. 3 servo motor, which is fixedly connected to the side of the No. 2 slider, and the output end of the No. 3 servo motor is fixedly connected to a No. 1 connecting frame, the lower part of the No. 1 connecting frame is fixedly connected to a No. 4 servo motor, the output end of the No. 4 servo motor is fixedly connected to the No. 2 connecting frame, the side of the No. 2 connecting frame is fixedly connected to a No. 5 servo motor, the output end of the No. 5 servo motor is fixedly connected to a mounting frame, an image acquisition device is installed on the surface of the mounting frame, the cleaning mechanism includes a fixed frame, the fixed frame is fixedly connected to the bottom of the No. 2 slide rail, and one end of the fixed frame is fixedly connected to a mounting plate, a plug plate is plugged into the inside of the mounting plate, and a wiping cotton layer is fixedly connected to the upper part of the plug plate.
[0034] Compared with the prior art, the present invention has the following beneficial effects:
[0035] 1. In the present invention, by collecting historical data of the same type of products and data of the same batch of products of the same type, the diversity of image data samples is improved, thereby improving the generalization ability of the model, enhancing the robustness of the model, and improving the accuracy of the model's prediction and judgment. When collecting image data and building a model, the influencing factors are fully considered, and by verifying the correlation of the influencing factors, different variables can be screened before modeling, and those variables that contribute more to the model can be selected, thereby improving the efficiency and accuracy of the model, and can also guide the optimization and adjustment of the model. Finally, the model is constructed through the random forest algorithm, and the optimal parameters are determined;
[0036] 2. In the present invention, when constructing a model, the constructed model is evaluated through model evaluation, and the accuracy of the model detection is quantitatively judged, which plays a role in detecting the established model. When the established model fails to reach the optimal setting value, the model that fails the model evaluation is optimized through model optimization, and the optimized model is cyclically evaluated through the model evaluation step, so that the obtained parameter setting value is the optimal setting value;
[0037] 3. In the present invention, the accuracy of image data acquisition of the image acquisition device is improved by accurately adjusting the installation position of the image acquisition device on the production line. When dust adheres to the lens of the image acquisition device during the acquisition process, the No. 2 servo motor drives the No. 2 slider to slide along the No. 2 slide rail, so that the lens of the image acquisition device passes through the surface of the wiping cotton layer. The lens of the image acquisition device is cleaned by wiping the cotton layer to ensure the accuracy of image data acquisition of the image acquisition device. BRIEF DESCRIPTION OF THE DRAWINGS
[0038] Figure 1 A flow chart of a method and device for analyzing optimal setting values of multiple parameters based on high-frequency machine vision learning of the present invention;
[0039] Figure 2 It is a schematic diagram of the three-dimensional structure of the installation mechanism in the analysis method and device of the optimal setting value of multiple parameters based on high-frequency machine vision learning of the present invention;
[0040] Figure 3 It is a schematic diagram of the three-dimensional structure of the cleaning mechanism in the analysis method of multi-parameter optimal setting values based on high-frequency machine vision learning and the device of the present invention;
[0041] Figure 4 It is a schematic diagram of the three-dimensional structure of the adjustment mechanism in the analysis method of multi-parameter optimal setting values based on high-frequency machine vision learning of the present invention and the device;
[0042] Figure 5 It is a front view structural schematic diagram of an image acquisition device in an analysis method for optimal setting values of multiple parameters based on high-frequency machine vision learning of the present invention and a device.
[0043] In the figure: 1. Installation mechanism; 11. Slide rail No. 1; 12. Slider No. 1; 13. Servo motor No. 1; 14. Adjustment screw No. 1; 15. Slide rail No. 2; 16. Slider No. 2; 17. Servo motor No. 2; 18. Adjustment screw No. 2; 2. Adjustment mechanism; 21. Servo motor No. 3; 22. Connecting frame No. 1; 23. Servo motor No. 4; 24. Connecting frame No. 2; 25. Servo motor No. 5; 26. Installation frame; 27. Image acquisition device; 3. Cleaning mechanism; 31. Fixing frame; 32. Installation plate; 33. Insertion plate; 34. Wiping cotton layer. DETAILED DESCRIPTION
[0044] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the implementation regulations described are only part of the embodiments of the present invention, not 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.
[0045] Reference Figure 1 As shown: The analysis method of multi-parameter optimal setting values based on high-frequency machine vision learning includes the following contents:
[0046] Step 1. Collect data sources: The data source consists of historical data of the same type of products and data of the same batch of products of the same type. The ratio of historical data of the same type of products to data of the same batch of products of the same type in the data source is 1:1. The image information of the same batch of products of the same type is collected using the image acquisition device of the production line.
[0047] Step 2: Divide the collected data sources into a test set and a training set, with the data sample ratio of the test set to the training set being 3:7.
[0048] Step 3: Determine the influencing factors: Divide the influencing factors according to human factors, environmental factors and equipment factors.
[0049] Step 4: Verify the correlation of the influencing factors. The correlation verification uses the variance inflation factor method to verify multicollinearity and screen factors with VIF values less than 10. The calculation formula of VIF is as follows: ;
[0050] Where: is the determination coefficient between the ith influencing factor and other influencing factors.
[0051] Step 5: Model construction: Use the collected samples of the training set to build the model. The model construction adopts the random forest algorithm. The random forest algorithm is used to select an optimal feature from the training set sample data as the split attribute of the node. The random forest algorithm includes the following contents:
[0052] S51, self-service sampling: randomly select k samples with replacement from the collected samples D of the training set as the training set K of a decision tree;
[0053] S52. Feature Random Selection: Randomly select m features from the M candidate features, where m < M, as the alternative feature subset for decision-making under the current node; then use the information gain strategy to select an optimal feature from the selected m features as the splitting attribute of this node. The information gain Gain(A, K) of feature A for dataset K is the difference between the entropy Ent(K) of the dataset and the conditional entropy CEnt(A, K) of feature A. The formula is as follows: ;
[0054] In the formula: is the proportion of unqualified and qualified samples in the current dataset K; represents the number of samples with the value of v for feature A;
[0055] S53. Decision Tree Growth: During the formation of the decision tree, each node repeats step S2 for splitting until there are no new features, and then the splitting ends;
[0056] S54. Construct Random Forest: Construct N decision trees according to steps S1 to S3;
[0057] S55. The final prediction result is classified by voting on the prediction results of all decision trees to determine an optimal feature parameter;
[0058] The decision tree growth uses the XGBoost algorithm to reduce the complexity of the decision tree and obtain the optimal objective function of the decision tree , and the optimal objective function is as follows: ;
[0059] In the formula: , is the sum of all samples belonging to the j-th leaf node ; is the first derivative of the loss function; , is the sum of all samples belonging to the j-th leaf node ; is the second derivative of the loss function; indicates that the i-th sample is located in the j-th leaf node; T is the number of leaf nodes; is the weight value of the leaf node; is a hyperparameter used to control the penalty strength;
[0060] The optimal objective function is a standard for measuring the quality of the tree structure, and the smaller the value, the better the tree structure.
[0061] Step 6. Model evaluation: Use the collected samples of the test set to evaluate the model constructed by the model. The model evaluation uses the confusion matrix and ROC curve to evaluate the model constructed by the model. The confusion matrix is used to calculate the accuracy, precision, and recall rate of the acquired data to evaluate the model. The ROC curve is used to quantitatively judge the accuracy of model detection.
[0062] Step 7: Model optimization: used to optimize models that fail the model evaluation. Model optimization uses the Bayesian optimization algorithm to optimize the constructed model. The Bayesian optimization algorithm optimizes the constructed model and includes the following contents:
[0063] S71. First, establish prior knowledge by approximating the target function through the Gaussian regression model;
[0064] S72, then by observing the output of the objective function at different input points, update this prior knowledge to form a posterior distribution;
[0065] S73. Based on the posterior distribution, use the PI acquisition function to perform the next sampling, add the sampled data to the data set, calculate the probability proxy function, and repeat the above steps. After iterating for a set number of times, select the global optimal solution.
[0066] like Figure 2-Figure 5 As shown: a device for multi-parameter optimal setting values based on high-frequency machine vision learning, comprising an installation mechanism 1, an adjustment mechanism 2 and a cleaning mechanism 3, wherein the installation mechanism 1 comprises a No. 1 slide rail 11, wherein the No. 1 slide rail 11 is internally slidably connected to a No. 1 slider 12, and the top of the No. 1 slide rail 11 is fixedly connected to a No. 1 servo motor 13, and the output end of the No. 1 servo motor 13 is fixedly connected to a No. 1 adjustment screw rod 14, and the No. 1 adjustment screw rod 14 is threadedly connected to the No. 1 slider 12, and the side of the No. 1 slider 12 is fixedly connected to a No. 2 slide rail 15, and the No. 2 slide rail 15 is internally slidably connected to a No. 2 slider 16, and one end of the No. 2 slide rail 15 is fixedly connected to a No. 2 servo motor 17, and the output end of the No. 2 servo motor 17 is fixedly connected to a No. 2 adjustment screw rod 18, and the No. 2 adjustment screw rod 18 is threadedly connected to the No. 2 slider 16, and the adjustment mechanism 3 is fixedly connected to the No. 2 servo motor 17. The section mechanism 2 includes a No. 3 servo motor 21, which is fixedly connected to the side of the No. 2 slider 16, and the output end of the No. 3 servo motor 21 is fixedly connected to the No. 1 connecting frame 22, the lower part of the No. 1 connecting frame 22 is fixedly connected to the No. 4 servo motor 23, the output end of the No. 4 servo motor 23 is fixedly connected to the No. 2 connecting frame 24, the side of the No. 2 connecting frame 24 is fixedly connected to the No. 5 servo motor 25, the output end of the No. 5 servo motor 25 is fixedly connected to the mounting frame 26, and the surface of the mounting frame 26 is installed with an image acquisition device 27, the cleaning mechanism 3 includes a fixed frame 31, the fixed frame 31 is fixedly connected to the bottom of the No. 2 slide rail 15, and one end of the fixed frame 31 is fixedly connected to a mounting plate 32, a plug-in plate 33 is plugged into the mounting plate 32, and a wiping cotton layer 34 is fixedly connected to the upper part of the plug-in plate 33.
[0067] When the present invention is used, by collecting historical data of the same type of products and data of the same batch of the same type of products, the diversity of image data samples is improved, thereby improving the generalization ability of the model, enhancing the robustness of the model, and improving the accuracy of the prediction and judgment of the model. Influencing factors are fully considered when collecting image data and building models, and by verifying the correlation of influencing factors, different variables can be screened before modeling, and those variables that contribute more to the model can be selected, thereby improving the efficiency and accuracy of the model. It can also guide the optimization and adjustment of the model, and finally the model is built through the random forest algorithm, and the optimal parameters are judged;
[0068] When constructing a model, the constructed model is evaluated through model evaluation, and the accuracy of model detection is quantitatively judged, which plays a role in detecting the established model. When the established model fails to reach the optimal setting value, the model that fails the model evaluation is optimized through model optimization, and the optimized model is cyclically evaluated through the model evaluation step to make the obtained parameter setting value the optimal setting value;
[0069] When the image acquisition device of the production line is used to acquire image information for the same batch of data of the same type of products, the image acquisition device 27 is used to acquire image data of the same batch of products of the same type. When the image acquisition device 27 is installed on the production line, the No. 1 servo motor 13 drives the No. 1 adjusting screw 14 to rotate, so that the No. 1 slider 12 slides along the No. 1 slide rail 11, thereby adjusting the height of the No. 2 slide rail 15, and then adjusting the installation height of the image acquisition device 27. Similarly, the No. 2 servo motor 17 drives the No. 2 adjusting screw 18 to rotate, driving the No. 2 slider 16 to slide along the No. 2 slide rail 15, thereby adjusting the image acquisition device 27. The left and right positions of the image acquisition device 27 are adjusted to ensure that the image acquisition device 27 is accurately installed on the production line, thereby improving the accuracy of image data acquisition by the image acquisition device 27. Similarly, the No. 1 connecting frame 22 is driven to move by the No. 3 servo motor 21, the No. 2 connecting frame 24 is driven to move by the No. 4 servo motor 23, and the No. 5 servo motor 25 drives the mounting frame 26 to move, thereby further accurately adjusting the installation position of the image acquisition device 27, so that the image acquisition device 27 is in the best image data acquisition position. At the same time, by adjusting the position of the image acquisition device 27, it can accurately acquire image data for different products.
[0070] When dust accumulates on the lens of the image acquisition device 27 during the acquisition process, the second servo motor 17 drives the second slider 16 to slide along the second slide rail 15, so that the lens of the image acquisition device 27 passes through the surface of the wiping cotton layer 34. The lens of the image acquisition device 27 is cleaned by the wiping cotton layer 34 to ensure the accuracy of the image data acquisition of the image acquisition device 27. The wiping cotton layer 34 is installed by plugging the plug plate 33 with the mounting plate 32, which is convenient for cleaning or replacing the wiping cotton layer 34.
[0071] Although the present invention has been described in detail with reference to the aforementioned embodiments, it is still possible for those skilled in the art to modify the technical solutions described in the aforementioned embodiments, or to make equivalent substitutions for some of the technical features therein. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.
Claims
1. An analysis method for optimal setting values of multiple parameters based on high-frequency machine vision learning, characterized in that: It includes the following: S1. Collect data sources: The data sources consist of historical data of the same type of products and data of the same batch of the same type of products; S2. Divide the collected data sources into a test set and a training set; S3. Determine influencing factors: Divide the influencing factors according to human factors, environmental factors, and equipment factors; S4. Verify the correlation of influencing factors: The correlation verification uses the variance inflation factor method to perform multicollinearity verification, and screens factors with VIF values less than 10. The calculation formula of VIF is as follows: ; Where: is the determination coefficient between the ith influencing factor and other influencing factors; S5. Model construction: Use the collected samples of the training set for model construction. The model construction uses the random forest algorithm, and the random forest algorithm is used to select an optimal feature from the training set sample data as the splitting attribute of the node; S6. Model evaluation: Use the collected samples of the test set to evaluate the model constructed in the model construction; S7. Model optimization: Used to optimize the models that fail the model evaluation; In step S1, the data of the same batch of the same type of products is used for image information collection by the image acquisition device of the production line; In step S5, the random forest algorithm includes the following: S51. Bootstrap sampling: Randomly draw k samples from the collected samples D of the training set with replacement as the training set K for a decision tree; S52. Random feature selection: Randomly select m features from the candidate M features, where m < M, as the alternative feature subset for decision-making under the current node; Then, the information gain strategy is used to select an optimal feature from the selected m features as the splitting attribute of the node. The information gain Gain (A, K) of feature A for data set K is the difference between the entropy Ent (K) of the data set and the conditional entropy CEnt (A, K) of feature A. The formula is as follows: ; Where: is the ratio of unqualified to qualified samples in the current data set K; Indicates the number of samples whose feature A takes the value v; S53. Decision tree growth: During the formation of the decision tree, each node repeats step S2 for splitting until there are no new features, and the splitting ends; S54. Construct a random forest: Construct N decision trees according to steps S1 to S3; S55. The final prediction result is classified by voting on the prediction results of all decision trees to determine an optimal feature parameter; When the image acquisition device of the production line is used to acquire image information for the same batch of data of the same type of products, the image acquisition device (27) is used to acquire image data for the same batch of products of the same type. When the image acquisition device (27) is installed on the production line, the first servo motor (13) drives the first adjusting screw (14) to rotate, so that the first slide block (12) slides along the first slide rail (11), thereby adjusting the height of the second slide rail (15), and then adjusting the installation height of the image acquisition device (27). Similarly, the second servo motor (17) drives the second adjusting screw (18) to rotate, driving the second slide block (16) to slide along the second slide rail (15), thereby adjusting the image acquisition device (27). The left and right positions of the image acquisition device (27) are adjusted so that the installation position of the image acquisition device (27) on the production line is accurate, thereby improving the accuracy of image data acquisition by the image acquisition device (27). Similarly, the third servo motor (21) drives the first connecting frame (22) to move, the fourth servo motor (23) drives the second connecting frame (24) to move, and the fifth servo motor (25) drives the mounting frame (26) to move, thereby further accurately adjusting the installation position of the image acquisition device (27) so that the image acquisition device (27) is in the best image data acquisition position. At the same time, by adjusting the position of the image acquisition device (27), it can accurately acquire image data for different products.
2. The method for analyzing multi-parameter optimal setting values based on high-frequency machine vision learning according to claim 1, characterized in that: In step S1, the ratio of historical data of the same type of product to the same batch data of the same type of product in the data source is 1:
1.
3. The method for analyzing multi-parameter optimal setting values based on high-frequency machine vision learning according to claim 2, characterized in that: In step S2, the data sample ratio of the test set and the training set is 3:
7.
4. The method for analyzing multi-parameter optimal setting values based on high-frequency machine vision learning according to claim 3 is characterized in that: In step S53, the decision tree growth adopts XGBoost algorithm to reduce the complexity of the decision tree and obtain the optimal objective function of the decision tree. , the optimal objective function as follows: ; Where: , for all samples belonging to the jth leaf node The sum of is the first-order derivative of the loss function; For all samples belonging to the jth leaf node The sum of is the second-order derivative of the loss function; It means that the i-th sample is located at the j-th leaf node; T is the number of leaf nodes; is the weight value of the leaf node; It is a hyperparameter used to control the intensity of the penalty; The optimal objective function is used to measure the quality of the tree structure. The smaller the value, the better the tree structure.
5. The method for analyzing multi-parameter optimal setting values based on high-frequency machine vision learning according to claim 4 is characterized in that: In step S6, the model evaluation uses confusion matrix and ROC curve to evaluate the model constructed by the model. The confusion matrix is used to calculate the accuracy, precision and recall rate of the acquired data to evaluate the model, and the ROC curve is used to quantitatively judge the accuracy of model detection.
6. The method for analyzing multi-parameter optimal setting values based on high-frequency machine vision learning according to claim 5, characterized in that: In step S7, the model optimization uses a Bayesian optimization algorithm to optimize the constructed model.
7. The method for analyzing multi-parameter optimal setting values based on high-frequency machine vision learning according to claim 6 is characterized in that: In step S7, the Bayesian optimization algorithm optimizes the constructed model including the following contents: S71. First, establish prior knowledge by approximating the target function through the Gaussian regression model; S72, then by observing the output of the objective function at different input points, update this prior knowledge to form a posterior distribution; S73. Based on the posterior distribution, use the PI acquisition function to perform the next sampling, add the sampled data to the data set, calculate the probability proxy function, and repeat the above steps. After iterating for a set number of times, select the global optimal solution.
8. A device for multi-parameter optimal setting values based on high-frequency machine vision learning, characterized in that The method for analyzing the optimal setting values of multiple parameters based on high-frequency machine vision learning as described in any one of claims 1 to 7 comprises a mounting mechanism (1), an adjusting mechanism (2) and a cleaning mechanism (3), wherein the mounting mechanism (1) comprises a first slide rail (11), wherein the first slide rail (11) is internally slidably connected to a first slider (12), and a first servo motor (13) is fixedly connected to the top of the first slide rail (11), and an output end of the first servo motor (13) is fixedly connected to a first adjusting screw rod (14), and the first adjusting screw rod (14) is threadedly connected to the first slider (12), and a second slide rail (15) is fixedly connected to the side of the first slider (12), and a second slide rail (15) is internally slidably connected to a second slider (16), and one end of the second slide rail (15) is fixedly connected to a second servo motor (17), and an output end of the second servo motor (17) is fixedly connected to a second adjusting screw rod (18), and the second adjusting screw rod (18) is threadedly connected to the second slider (16). The regulating mechanism (2) comprises a No. 3 servo motor (21), the No. 3 servo motor (21) being fixedly connected to a side surface of a No. 2 slider (16), and an output end of the No. 3 servo motor (21) being fixedly connected to a No. 1 connecting frame (22), a lower portion of the No. 1 connecting frame (22) being fixedly connected to a No. 4 servo motor (23), an output end of the No. 4 servo motor (23) being fixedly connected to a No. 2 connecting frame (24), and a side surface of the No. 2 connecting frame (24) being fixedly connected to a No. 5 servo motor (25). The output end of the No. 5 servo motor (25) is fixedly connected to a mounting frame (26), and an image acquisition device (27) is installed on the surface of the mounting frame (26). The cleaning mechanism (3) includes a fixing frame (31), and the fixing frame (31) is fixedly connected to the bottom of the No. 2 slide rail (15), and one end of the fixing frame (31) is fixedly connected to a mounting plate (32), and a plug-in plate (33) is plugged into the inside of the mounting plate (32), and a wiping cotton layer (34) is fixedly connected to the upper part of the plug-in plate (33).
Citation Information
Patent Citations
Method for measuring hearth temperature of coal-fired power plant boiler
CN110532674A
Method for predicting end-point quality of metallurgical process based on stacking integration strategy
CN118197463A
