CN117314871A - Band steel sorting degree grading method and device based on gray level co-occurrence matrix and decision tree - Google Patents

Abstract

The invention discloses a strip steel sorting degree grading method and device based on a grayscale co-occurrence matrix and a decision tree, and relates to the technical fields of metallurgical machinery and automation. Including: obtaining the image data and text data of the cold-rolled strip surface to be graded; inputting the image data and text data into the constructed cold-rolled strip surface quality sorting degree grading model; based on the image data, text data and cold-rolled The surface quality sorting degree classification model of strip steel is used to obtain the surface quality sorting degree classification results of cold-rolled strip steel. Through the present invention, the intelligent determination of surface quality of cold-rolled strip steel can be better realized, the accuracy of surface quality determination can be improved, and the results of the determination system can be closer to the quality inspection personnel.

Description

Strip sorting degree classification method and device based on gray level co-occurrence matrix and decision tree

Technical field

The invention relates to the technical fields of metallurgical machinery and automation, and in particular to a strip steel sorting degree grading method and device based on a grayscale co-occurrence matrix and a decision tree.

Background technique

As the automotive industry, high-end home appliances and other industries have increasingly stringent quality requirements for steel products, there are also increasing demands for personalized products. On the premise that the basic quality indicators are qualified, steel companies and downstream users have begun to pay attention to more detailed surface quality issues. Due to the long process involved, complex formation mechanisms, and high frequency of occurrence, it has a great impact on the final quality of the product, and the surface quality It is considered to be one of the most important and difficult to control quality indicators. Due to the complexity of surface quality, major steel companies attach great importance to the management of surface quality of cold-rolled strip steel, hoping to improve product quality and achieve greater commercial value.

Determining the surface quality of cold-rolled strip steel is currently a difficult problem. The cold-rolled strip steel production line is equipped with a surface inspection system (referred to as: surface inspection system), but it can only identify the types of defects and often does not have the ability to detect the strip surface. Quality sorting grading function. Different from other automatic determination methods of performance indicators based on parameter thresholds, due to the limitations of the surface inspection system such as complex working environment, high false alarm rate of defect detection, and inability to evaluate the severity of defects, the surface quality inspection process still requires manual labor. The off-line products are inspected one by one. The judgment method based on manual experience is difficult to be replaced by automated methods, which affects the efficiency and accuracy of the entire production process.

Contents of the invention

This invention aims at the existing defect information based on the surface quality detection system and combines the process ideas of manual steel judgment to realize the method of automatic judgment of the strip surface quality. Due to the large number and variety of surface defects of cold-rolled strip steel, various types of defects exist. The problem of inaccurate sorting degree grading under basic rules, the original data of defects has many dimensions, irregular format, and the problem of not unifying the dimensions of each data sample, because the sorting degree grading of defects is a multi-variable coupling classification problem. It is difficult to quantify the problem into specific classification rules, so the present invention is proposed.

In order to solve the above technical problems, the present invention provides the following technical solutions:

On the one hand, the present invention provides a strip sorting degree grading method based on a gray level co-occurrence matrix and a decision tree. The method is implemented by electronic equipment. The method includes:

S1. Obtain the image data and text data of the cold-rolled strip surface to be graded.

S2. Input the image data and text data into the constructed cold-rolled strip surface quality sorting degree model.

S3. According to the image data, text data and the cold-rolled strip surface quality sorting degree grading model, obtain the cold-rolled strip surface quality sorting degree grading results.

Optionally, the construction process of the cold-rolled strip surface quality sorting model in S2 includes:

S21. Obtain a data set; wherein, the data set includes defect image information on the surface of cold-rolled strip steel, text information on defects on the surface of cold-rolled strip steel, and quality sorting grade of defects on the surface of cold-rolled strip steel.

S22. According to the data set and the gray level co-occurrence matrix algorithm, calculate the eigenvalues of the defect image.

S23. Establish a cold-rolled strip surface quality feature vector based on the characteristic values and the defect text information on the cold-rolled strip surface.

S24. Based on the cold-rolled strip surface quality feature vector and data set, establish a decision rule model based on the decision tree algorithm.

S25. Construct a cold-rolled strip surface quality sorting degree classification model based on the classification function of the decision rule model.

Optionally, the defect text information on the cold-rolled strip surface in S21 includes:

The relative coordinates, defect area, defect length and defect width of the defect area of interest on the cold rolled strip surface.

Optionally, in S22, based on the data set and the gray level co-occurrence matrix algorithm, the characteristic values of the defect image are calculated, including:

S221. Obtain the structural parameters of the gray level co-occurrence matrix determined based on pre-training; where the structural parameters include the generation step size, the image gray level and the generation direction.

S222. According to the gray-level co-occurrence matrix, calculate the gray-level co-occurrence matrix of each defect image in the data set, and obtain the energy value, entropy value, contrast and correlation of the aggregated defects on the surface of the cold-rolled strip steel.

Optionally, in S23, a cold-rolled strip surface quality feature vector is established based on the characteristic values and the defect text information on the cold-rolled strip surface, including:

According to the characteristic values and the defect text information on the surface of the cold-rolled strip, the aggregated defect characteristics on the surface of the cold-rolled strip are calculated, and the surface quality feature vector of the cold-rolled strip is obtained, as shown in the following formula (1):

Among them, F1 represents the number of all aggregated defects on a roll of cold-rolled strip steel, F2 _i represents the defect aspect ratio, F3 _i represents the defect area, W1 _i represents the energy value, W2 _i represents the entropy value, and W3 _i represents the contrast. W4 _i represents correlation.

Optionally, in S24, a decision rule model based on the decision tree algorithm is established based on the cold-rolled strip surface quality feature vector and data set, including:

S241. Use the grid search method to adjust the parameters of the decision tree algorithm, determine the value limit range of the parameters, and establish multiple decision tree models; the parameters include the maximum tree depth and the minimum number of samples required for branches.

S242. Calculate the feature standard Gini coefficient for the training set in the data set.

S243. According to the characteristic standard Gini coefficient, calculate the Gini coefficient for each cold-rolled strip surface quality feature vector to obtain the decision rule model based on the decision tree algorithm under the current parameters, and then obtain the decision based on the decision tree algorithm under different parameters. Rule model.

S244. Use the pre-pruning method to process the decision tree model under different parameters to obtain a decision rule model based on the decision tree algorithm.

Optionally, in S242, calculate the feature standard Gini coefficient for the training set in the data set, as shown in the following equation (2):

Among them, Gini(D) represents the probability that two samples are randomly selected from the data set and their categories are inconsistent, k represents the number of categories, and p( _xi ) represents the probability that category _xi appears.

Optionally, the Gini coefficient in S243 is as shown in the following equation (3):

Among them, Gini(D,A) represents the Gini coefficient under feature A, A represents the feature, D represents the data set, D1 and D2 represent the two parts that divide each feature training sample according to the value a _i of feature A, Gini(D1 ) represents the probability that two samples randomly selected from feature training sample D1 have inconsistent categories, and Gini(D2) represents two samples randomly selected from feature training sample D2.

Optionally, the pre-pruning method in S244 is used to process the decision tree model under different parameters to obtain a decision rule model based on the decision tree algorithm, including:

For the test set in the data set, a confusion matrix of the decision tree model under different parameters is established. By calculating the accuracy of the decision tree model under different parameter combinations, the optimal parameters are selected to obtain a decision rule model based on the decision tree algorithm.

On the other hand, the present invention provides a strip sorting degree grading device based on a gray level co-occurrence matrix and a decision tree, which device is used to implement a strip steel sorting degree grading method based on a gray level co-occurrence matrix and a decision tree. Devices include:

The acquisition module is used to acquire the image data and text data of the cold-rolled strip surface to be graded.

The input module is used to input image data and text data into the constructed cold-rolled strip surface quality sorting degree model.

The output module is used to obtain the cold-rolled strip surface quality sorting degree grading results based on the image data, text data and the cold-rolled strip surface quality sorting degree grading model.

Optionally, input modules are further used to:

S21. Obtain a data set; wherein, the data set includes defect image information on the surface of cold-rolled strip steel, text information on defects on the surface of cold-rolled strip steel, and quality sorting grade of defects on the surface of cold-rolled strip steel.

S22. According to the data set and the gray level co-occurrence matrix algorithm, calculate the eigenvalues of the defect image.

S23. Establish a cold-rolled strip surface quality feature vector based on the characteristic values and the defect text information on the cold-rolled strip surface.

S24. Based on the cold-rolled strip surface quality feature vector and data set, establish a decision rule model based on the decision tree algorithm.

S25. Construct a cold-rolled strip surface quality sorting degree classification model based on the classification function of the decision rule model.

Optionally, the defect text information on the cold-rolled strip surface includes:

The relative coordinates, defect area, defect length and defect width of the defect area of interest on the cold rolled strip surface.

Optionally, input modules are further used to:

S221. Obtain the structural parameters of the gray level co-occurrence matrix determined based on pre-training; where the structural parameters include the generation step size, the image gray level and the generation direction.

S222. According to the gray-level co-occurrence matrix, calculate the gray-level co-occurrence matrix of each defect image in the data set, and obtain the energy value, entropy value, contrast and correlation of the aggregated defects on the surface of the cold-rolled strip steel.

Optionally, input modules are further used to:

According to the characteristic values and the defect text information on the surface of the cold-rolled strip, the aggregated defect characteristics on the surface of the cold-rolled strip are calculated, and the surface quality feature vector of the cold-rolled strip is obtained, as shown in the following formula (1):

Among them, F1 represents the number of all aggregated defects on a roll of cold-rolled strip steel, F2 _i represents the defect aspect ratio, F3 _i represents the defect area, W1 _i represents the energy value, W2 _i represents the entropy value, and W3 _i represents the contrast. W4 _i represents correlation.

Optionally, input modules are further used to:

S241. Use the grid search method to adjust the parameters of the decision tree algorithm, determine the value limit range of the parameters, and establish multiple decision tree models; the parameters include the maximum tree depth and the minimum number of samples required for branches.

S242. Calculate the feature standard Gini coefficient for the training set in the data set.

S243. According to the characteristic standard Gini coefficient, calculate the Gini coefficient for each cold-rolled strip surface quality feature vector to obtain the decision rule model based on the decision tree algorithm under the current parameters, and then obtain the decision based on the decision tree algorithm under different parameters. Rule model.

S244. Use the pre-pruning method to process the decision tree model under different parameters to obtain a decision rule model based on the decision tree algorithm.

Optionally, for the training set in the data set, calculate the feature standard Gini coefficient, as shown in the following equation (2):

Among them, Gini(D) represents the probability that two samples are randomly selected from the data set and their categories are inconsistent, k represents the number of categories, and p( _xi ) represents the probability that category _xi appears.

Optionally, the Gini coefficient is as shown in the following equation (3):

Among them, Gini(D,A) represents the Gini coefficient under feature A, A represents the feature, D represents the data set, D1 and D2 represent the two parts that divide each feature training sample according to the value a _i of feature A, Gini(D1 ) represents the probability that two samples randomly selected from feature training sample D1 have inconsistent categories, and Gini(D2) represents two samples randomly selected from feature training sample D2.

Optionally, use the pre-pruning method to process the decision tree model under different parameters to obtain a decision rule model based on the decision tree algorithm, including:

For the test set in the data set, a confusion matrix of the decision tree model under different parameters is established. By calculating the accuracy of the decision tree model under different parameter combinations, the optimal parameters are selected to obtain a decision rule model based on the decision tree algorithm.

On the one hand, an electronic device is provided. The electronic device includes a processor and a memory. At least one instruction is stored in the memory. The at least one instruction is loaded and executed by the processor to implement the above-mentioned grayscale-based symbiosis. Strip sorting degree classification method using matrix and decision tree.

On the one hand, a computer-readable storage medium is provided. At least one instruction is stored in the storage medium. The at least one instruction is loaded and executed by a processor to implement the above-mentioned strip classification based on gray-scale co-occurrence matrix and decision tree. Selection grading method.

Compared with the existing technology, the above technical solution has at least the following beneficial effects:

The above scheme provides a sorting and grading method for the surface quality of cold-rolled strip based on gray-scale co-occurrence matrix and decision tree rules, which can automatically classify the surface quality of cold-rolled strip and thereby reduce the surface quality of the strip. The missed detection rate reduces quality objections in downstream processes.

Description of the drawings

In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings needed to be used in the description of the embodiments will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can also be obtained based on these drawings without exerting creative efforts.

Figure 1 is a schematic flow chart of the strip sorting degree grading method based on the gray level co-occurrence matrix and decision tree provided by the embodiment of the present invention;

Figure 2 is a schematic flow chart of a cold-rolled strip surface quality sorting degree grading determination method based on gray value statistics and decision tree rules provided by an embodiment of the present invention;

Figure 3 is a decision tree provided by an embodiment of the present invention;

Figure 4 is a schematic diagram of an S1 level defect image provided by an embodiment of the present invention;

Figure 5 is a schematic diagram of an S2 level defect image provided by an embodiment of the present invention;

Figure 6 is a schematic diagram of an S4 level defect image provided by an embodiment of the present invention;

Figure 7 is a block diagram of a strip sorting degree grading device based on a gray level co-occurrence matrix and a decision tree provided by an embodiment of the present invention;

Figure 8 is a schematic structural diagram of an electronic device provided by an embodiment of the present invention.

Detailed ways

In order to make the purpose, technical solutions and advantages of the embodiments of the present invention more clear, the technical solutions of the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings of the embodiments of the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. Based on the described embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts fall within the scope of protection of the present invention.

As shown in Figure 1, an embodiment of the present invention provides a strip sorting degree grading method based on a gray-level co-occurrence matrix and a decision tree. This method can be implemented by electronic equipment. As shown in Figure 1, the flow chart of the strip sorting degree classification method based on gray level co-occurrence matrix and decision tree, the processing flow of this method can include the following steps:

S1. Obtain the image data and text data of the cold-rolled strip surface to be graded.

In a feasible implementation, data can be acquired through a surface detection system.

S2. Input the image data and text data into the constructed cold-rolled strip surface quality sorting degree model.

Among them, sorting is the process of selecting materials that can be downgraded from a pile of qualified and unqualified materials.

Optionally, as shown in Figure 2, the construction process of the cold-rolled strip surface quality sorting model in S2 may include the following steps S21-S25:

S21. Obtain the data set.

In a feasible implementation, image information and text information of aggregated defects on the surface of cold-rolled strip steel are collected. The text information can include: relative coordinates of the defect area of interest, defect area, defect length, defect width, etc., and obtain each roll Surface quality sorting grade of strip steel. Create a data set based on the above three-in-one data.

S22. Calculate the characteristic value of the defect image according to the data set and the gray level co-occurrence matrix algorithm, which may include the following steps S221-S222:

S221. Obtain the structural parameters of the gray level co-occurrence matrix determined based on pre-training; where the structural parameters include the generation step size d, the image gray level G, and the generation direction θ.

S222. According to the gray-level co-occurrence matrix, calculate the gray-level co-occurrence matrix of each defect image in the data set, and obtain the energy value, entropy value, contrast and correlation of the aggregated defects on the surface of the cold-rolled strip steel.

In a feasible implementation, each obtained defect image is grayscaled. According to the probability from the original image point with grayscale i to the point with grayscale j in the original image, d=(Dx,Dy) is The corresponding relationship between the two points i and j determines the positional relationship between the two pixels in the image. i, j=0,1,2,...L-1, L represents the gray level of the pixel, then the gray level The co-occurrence matrix can be expressed as P(i,j).

Furthermore, assuming that a certain point (x, y) of the defect image can be moved, then different (Ii, Ij) will be obtained. Then there are L×L combinations of (Ii, Ij), and the number of occurrences of each type can be counted. , arranged into a matrix. In this embodiment, the total number Z of all situations is:

Further, the grayscale matrix P _L×L is:

Among them, C(Ii,Ij) represents each element in P _L×L ; L represents the gray level of the pixel.

energy:

Entropy value:

Contrast:

Correlation:

Among them, P (i, j, d, θ) is the probability of the original image point with gray level i to the point with gray level j in the original image, and G is the gray level.

S23. Based on the eigenvalues and the defect text information on the surface of the cold-rolled strip, calculate all the aggregated defect characteristics on the surface of a roll of strip to form a eigenvector, as shown in the following equation (8):

Among them, F1 represents the number of all aggregated defects on a roll of cold-rolled strip steel, F2 _i represents the defect aspect ratio, F3 _i represents the defect area, W1 _i represents the energy value, W2 _i represents the entropy value, and W3 _i represents the contrast. W4 _i represents correlation.

S24. Based on the cold-rolled strip surface quality feature vector and data set, establish a decision rule model based on the decision tree algorithm.

In a feasible implementation, the obtained feature vectors are used as leaf nodes of the decision tree to build a decision tree. In order to ensure the accuracy of the decision tree model and avoid overfitting, a pre-pruning method is used to stop the construction of the tree in advance. Perform pruning. The decision tree algorithm uses the cart decision tree algorithm, and the selected feature standard is the Gini coefficient. The model construction may include the following steps S241-S244:

S241. Use the grid search method to adjust the parameters of the decision tree algorithm, determine the value limit range of the parameters, and establish multiple decision tree models; the parameters include the maximum tree depth and the minimum number of samples required for branches.

S242. For the training set in the data set, calculate the feature standard Gini coefficient, as shown in the following equation (9):

Among them, Gini(D) represents the probability that two samples are randomly selected from the data set and their categories are inconsistent, k represents the number of categories, and p( _xi ) represents the probability that category _xi appears.

S243. According to the feature standard Gini coefficient, for each feature, calculate the Gini coefficient under the feature to obtain the decision rule model based on the decision tree algorithm under the current parameters, and then obtain the decision rule model based on the decision tree algorithm under different parameters.

Among them, for each feature A, each possible value a can be obtained. Divide each feature training sample into two parts, D1 and D2, according to the value ai. The Gini coefficient under feature A is obtained from the Gini coefficient formula, as shown in the following equation (10):

Among them, Gini(D,A) represents the Gini coefficient under feature A, A represents the feature, D represents the data set, D1 and D2 represent the two parts that divide each feature training sample according to the value a _i of feature A, Gini(D1 ) represents the probability that two samples randomly selected from feature training sample D1 have inconsistent categories, and Gini(D2) represents two samples randomly selected from feature training sample D2.

Furthermore, for the data set D, the Gini coefficient of each feature is calculated to obtain the decision tree model under the current maximum tree depth and the minimum number of samples required for branches.

S244. For the test set in the data set, establish a confusion matrix of the decision tree model under different parameters. By calculating the accuracy of the decision tree model under different parameter combinations, select the optimal parameters, implement pre-pruning of the decision tree, and obtain the decision-based Decision rule model of tree algorithm.

S25. Construct a cold-rolled strip surface quality sorting degree classification model based on the classification function of the decision rule model.

S3. According to the image data, text data and the cold-rolled strip surface quality sorting degree grading model, obtain the cold-rolled strip surface quality sorting degree grading results.

In a feasible implementation, the surface detector data is used as input, the surface quality feature vector is obtained through the above steps, and the surface quality sorting degree classification of cold-rolled strip steel is realized through the decision tree model classification function.

Specifically, taking the defects detected by the meter inspection system of a cold rolling continuous production line of a steel company as an example, the specific implementation of this patent will be introduced below in conjunction with the content of the present invention, which specifically includes the following steps:

Step 1: Obtain the strip image with clustered defects and the geometric information of the defect area through the existing surface inspection system. Combined with manual experience, each surface defect sample is marked with three levels according to the severity of the defect. Among them, a total of Collect 50 rolls of S1 grade strip data, 50 rolls of S2 grade strip data, and 50 rolls of S4 grade strip data. The number of samples for each grade is relatively even.

Step 2: Calculate the energy value, entropy value, correlation, and contrast of each aggregated defect through the gray level co-occurrence matrix, and through calculation, obtain the energy value, entropy value, correlation, and contrast of the surface of each strip steel strip. The steel surface characteristic value results are shown in Table 1:

Table 1

Step 3: After obtaining the geometric information of surface defects and gray level co-occurrence matrix statistical information of each roll of strip, calculate the surface feature vector of each roll of strip. The feature vectors of 150 rolls of strip are shown in Table 2 below:

Table 2

Step 4: Use the obtained feature vectors as leaf nodes of the decision tree to build a decision tree. In order to ensure the accuracy of the decision tree model and avoid overfitting, the pre-pruning method is used to prune the tree by stopping the construction of the tree in advance. The decision tree algorithm uses the cart decision tree algorithm, and the feature standard selected is the Gini coefficient. The decision tree shape is shown in Figure 3. The model accuracy under different parameter combinations is shown in Table 3 below:

table 3

Among them, the model accuracy reaches the highest when the maximum tree depth is 5 and the minimum number of samples required for branches is 1, so it is selected as the pre-pruning structural parameter of the decision tree.

Step 5: Using the surface detector data as input, the surface quality feature vector is obtained through steps 2 and 3, and the surface quality sorting degree classification of cold-rolled strip steel is realized through the decision tree model classification function.

Among them, the decision tree classification function is:

Furthermore, 45 coils of strip steel data are used as input to extract their feature vectors. The final output of the classification system is the surface quality sorting grade of cold rolled strip steel, namely S1 level (as shown in Figure 4) and S2 level (as shown in Figure 5). (shown in Figure 6) or S4 level (shown in Figure 6), the prediction results are shown in Table 4:

Table 4

The prediction results are statistically graded. The samples with consistent prediction results and sample labels at each level are S1 level 16 volumes, S2 level 9 volumes, and S4 level 17 volumes. The S1 level prediction accuracy is 100%, and the S2 level prediction accuracy is 100%. The prediction accuracy of S3 level is 90% and 89.47%. There is no serious misjudgment, that is, S4 level defects are misjudged as S1 level defects. This data is lower than the manual misjudgment rate of defect images (3%).

In this embodiment, in order to further understand the cold-rolled strip surface quality sorting method based on the gray-scale co-occurrence matrix and decision tree described in the embodiment of the present invention, the present invention is applied to the cold rolling continuous production line of a steel enterprise , starting from the day shift production on the first Monday of each month, 1,000 coils of strip steel participating in the automatic judgment are continuously selected, and the judgment results are used as samples for error evaluation. The actual application effect analysis results are shown in Table 5 below:

table 5

In the embodiment of the present invention, a cold-rolled strip surface quality sorting method based on gray-scale co-occurrence matrix and decision tree rules is provided, which can automatically classify the cold-rolled strip surface quality, thereby reducing the strip surface quality. The missed detection rate of steel surface quality reduces quality objections in downstream processes.

As shown in Figure 7, the embodiment of the present invention provides a strip sorting and grading device 700 based on a gray-level co-occurrence matrix and a decision tree. The device 700 is used to implement strip sorting based on a gray-level co-occurrence matrix and a decision tree. Selectivity grading method, the device 700 includes:

The acquisition module 710 is used to acquire image data and text data of the cold-rolled strip surface to be graded.

The input module 720 is used to input image data and text data into the constructed cold-rolled strip surface quality sorting degree classification model.

The output module 730 is used to obtain the cold-rolled strip surface quality sorting degree grading results based on the image data, text data and the cold-rolled strip surface quality sorting degree grading model.

Optionally, the input module 720 is further used for:

S21. Obtain a data set; wherein, the data set includes defect image information on the surface of cold-rolled strip steel, text information on defects on the surface of cold-rolled strip steel, and quality sorting grade of defects on the surface of cold-rolled strip steel.

S22. According to the data set and the gray level co-occurrence matrix algorithm, calculate the eigenvalues of the defect image.

S23. Establish a cold-rolled strip surface quality feature vector based on the characteristic values and the defect text information on the cold-rolled strip surface.

S24. Based on the cold-rolled strip surface quality feature vector and data set, establish a decision rule model based on the decision tree algorithm.

S25. Construct a cold-rolled strip surface quality sorting degree classification model based on the classification function of the decision rule model.

Optionally, the defect text information on the cold-rolled strip surface includes:

The relative coordinates, defect area, defect length and defect width of the defect area of interest on the cold rolled strip surface.

Optionally, the input module 720 is further used for:

S221. Obtain the structural parameters of the gray level co-occurrence matrix determined based on pre-training; where the structural parameters include the generation step size, the image gray level and the generation direction.

S222. According to the gray-level co-occurrence matrix, calculate the gray-level co-occurrence matrix of each defect image in the data set, and obtain the energy value, entropy value, contrast and correlation of the aggregated defects on the surface of the cold-rolled strip steel.

Optionally, the input module 720 is further used for:

According to the characteristic values and the defect text information on the surface of the cold-rolled strip, the aggregated defect characteristics on the surface of the cold-rolled strip are calculated, and the surface quality feature vector of the cold-rolled strip is obtained, as shown in the following formula (1):

Among them, f1 represents the number of all aggregated defects on a roll of cold-rolled strip steel, F2 _i represents the defect aspect ratio, F3 _i represents the defect area, W1 _i represents the energy value, W2 _i represents the entropy value, W3 _i represents the contrast, W4 _i represents correlation.

Optionally, the input module 720 is further used for:

S241. Use the grid search method to adjust the parameters of the decision tree algorithm, determine the value limit range of the parameters, and establish multiple decision tree models; the parameters include the maximum tree depth and the minimum number of samples required for branches.

S242. Calculate the feature standard Gini coefficient for the training set in the data set.

S243. According to the characteristic standard Gini coefficient, calculate the Gini coefficient for each cold-rolled strip surface quality feature vector to obtain the decision rule model based on the decision tree algorithm under the current parameters, and then obtain the decision based on the decision tree algorithm under different parameters. Rule model.

S244. Use the pre-pruning method to process the decision tree model under different parameters to obtain a decision rule model based on the decision tree algorithm.

Optionally, for the training set in the data set, calculate the feature standard Gini coefficient, as shown in the following equation (2):

Among them, Gini(D) represents the probability that two samples are randomly selected from the data set and their categories are inconsistent, k represents the number of categories, and p( _xi ) represents the probability that category _xi appears.

Optionally, the Gini coefficient under the characteristic is as shown in the following equation (3):

Among them, Gini(D,A) represents the Gini coefficient under feature A, A represents the feature, D represents the data set, D1 and D2 represent the two parts that divide each feature training sample according to the value a _i of feature A, Gini(D1 ) represents the probability that two samples randomly selected from feature training sample D1 have inconsistent categories, and Gini(D2) represents two samples randomly selected from feature training sample D2.

Optionally, use the pre-pruning method to process the decision tree model under different parameters to obtain a decision rule model based on the decision tree algorithm, including:

For the test set in the data set, a confusion matrix of the decision tree model under different parameters is established. By calculating the accuracy of the decision tree model under different parameter combinations, the optimal parameters are selected to obtain a decision rule model based on the decision tree algorithm.

In the embodiment of the present invention, a cold-rolled strip surface quality sorting method based on gray-scale co-occurrence matrix and decision tree rules is provided, which can automatically classify the cold-rolled strip surface quality, thereby reducing the strip surface quality. The missed detection rate of steel surface quality reduces quality objections in downstream processes.

FIG. 8 is a schematic structural diagram of an electronic device 800 provided by an embodiment of the present invention. The electronic device 800 may vary greatly due to different configurations or performance, and may include one or more processors (central processing units, CPUs) 801 and one or more memories 802, wherein at least one instruction is stored in the memory 802, and at least one instruction is loaded and executed by the processor 801 to implement the following strip sorting degree classification method based on gray level co-occurrence matrix and decision tree :

S1. Obtain the image data and text data of the cold-rolled strip surface to be graded.

S2. Input the image data and text data into the constructed cold-rolled strip surface quality sorting degree model.

S3. According to the image data, text data and the cold-rolled strip surface quality sorting degree grading model, obtain the cold-rolled strip surface quality sorting degree grading results.

In an exemplary embodiment, a computer-readable storage medium is also provided, such as a memory including instructions. The instructions can be executed by a processor in a terminal to complete the above-mentioned strip sorting degree based on gray-level co-occurrence matrix and decision tree. Grading method. For example, computer-readable storage media may be ROM, random access memory (RAM), CD-ROM, magnetic tape, floppy disk, optical data storage device, etc.

Those of ordinary skill in the art can understand that all or part of the steps to implement the above embodiments can be completed by hardware, or can be completed by instructing relevant hardware through a program. The program can be stored in a computer-readable storage medium. The above-mentioned The storage media mentioned can be read-only memory, magnetic disks or optical disks, etc.

The above are only preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention shall be included in the protection of the present invention. within the range.

Claims (10)

1. A strip sorting degree classification method based on gray level co-occurrence matrix and decision tree, characterized in that the method includes: S1. Obtain the image data and text data of the cold-rolled strip surface to be graded; S2. Input the image data and text data into the constructed cold-rolled strip surface quality sorting degree classification model; S3. According to the image data, text data and the cold-rolled strip surface quality sorting degree grading model, obtain the cold-rolled strip surface quality sorting degree grading results. 2. The method according to claim 1, characterized in that the construction process of the cold-rolled strip surface quality sorting degree classification model in S2 includes: S21. Obtain a data set; wherein the data set includes defect image information on the surface of the cold-rolled strip, text information on the defects on the surface of the cold-rolled strip, and quality sorting grade of defects on the surface of the cold-rolled strip; S22. Calculate the eigenvalues of the defect image according to the data set and the gray level co-occurrence matrix algorithm; S23. Establish a cold-rolled strip surface quality feature vector based on the characteristic values and the defect text information on the cold-rolled strip surface; S24. Establish a decision rule model based on the decision tree algorithm based on the cold-rolled strip surface quality feature vector and data set; S25. Construct a cold-rolled strip surface quality sorting degree classification model based on the classification function of the decision rule model. 3. The method according to claim 2, characterized in that the defect text information on the surface of the cold-rolled strip steel in S21 includes: The relative coordinates, defect area, defect length and defect width of the defect area of interest on the cold rolled strip surface. 4. The method according to claim 2, characterized in that, in S22, calculating the characteristic value of the defect image according to the data set and the gray level co-occurrence matrix algorithm includes: S221. Obtain the structural parameters of the gray level co-occurrence matrix determined based on pre-training; wherein the structural parameters include generation step size, image gray level and generation direction; S222. According to the gray-level co-occurrence matrix, calculate the gray-level co-occurrence matrix of each defect image in the data set, and obtain the energy value, entropy value, contrast and correlation of the aggregated defects on the surface of the cold-rolled strip steel. 5. The method according to claim 2, characterized in that, in S23, establishing a cold-rolled strip surface quality feature vector based on the characteristic value and the defect text information of the cold-rolled strip surface includes: According to the characteristic values and the defect text information on the surface of the cold-rolled strip, the aggregated defect characteristics on the surface of the cold-rolled strip are calculated, and the surface quality feature vector of the cold-rolled strip is obtained, as shown in the following formula (1): Among them, F1 represents the number of all aggregated defects on a roll of cold-rolled strip steel, F2 _i represents the defect aspect ratio, F3 _i represents the defect area, W1 _i represents the energy value, W2 _i represents the entropy value, and W3 _i represents the contrast. W4 _i represents correlation. 6. The method according to claim 2, characterized in that, in S24, a decision rule model based on a decision tree algorithm is established according to the cold-rolled strip surface quality feature vector and the data set, including: S241. Use the grid search method to adjust the parameters of the decision tree algorithm, determine the value limit range of the parameters, and establish multiple decision tree models; wherein the parameters include the maximum tree depth and the minimum number of samples required for branches; S242. Calculate the feature standard Gini coefficient for the training set in the data set; S243. According to the characteristic standard Gini coefficient, calculate the Gini coefficient for each cold-rolled strip surface quality feature vector to obtain the decision rule model based on the decision tree algorithm under the current parameters, and then obtain the decision rule model based on the decision tree algorithm under different parameters. Decision rule model; S244. Use a pre-pruning method to process the decision tree model under different parameters to obtain a decision rule model based on the decision tree algorithm. 7. The method according to claim 6, characterized in that, in the S242, for the training set in the data set, the characteristic standard Gini coefficient is calculated, as shown in the following formula (2): Among them, Gini(D) represents the probability that two samples are randomly selected from the data set and their categories are inconsistent, k represents the number of categories, and p( _xi ) represents the probability that category _xi appears. 8. The method according to claim 6, characterized in that the Gini coefficient in S243 is represented by the following formula (3): Among them, Gini(D,A) represents the Gini coefficient under feature A, A represents the feature, D represents the data set, D1 and D2 represent the two parts that divide each feature training sample according to the value a _i of feature A, Gini(D1 ) represents the probability that two samples randomly selected from feature training sample D1 have inconsistent categories, and Gini(D2) represents two samples randomly selected from feature training sample D2. 9. The method according to claim 6, wherein the pre-pruning method in S244 is used to process the decision tree model under different parameters to obtain a decision rule model based on the decision tree algorithm, including: For the test set in the data set, a confusion matrix of the decision tree model under different parameters is established. By calculating the accuracy of the decision tree model under different parameter combinations, the optimal parameters are selected to obtain a decision rule model based on the decision tree algorithm. 10. A strip sorting and grading device based on gray level co-occurrence matrix and decision tree, characterized in that the device includes: The acquisition module is used to acquire the image data and text data of the cold-rolled strip surface to be graded; An input module for inputting the image data and text data into the constructed cold-rolled strip surface quality sorting degree classification model; The output module is used to obtain the cold-rolled strip surface quality sorting degree grading results based on the image data, text data and cold-rolled strip surface quality sorting degree grading model.