Method, device, equipment and medium for judging sampling points during the refining reduction period of anode furnace

A method using image processing and a trained judgment model to identify anode furnace sampling points improves accuracy and efficiency by reducing reliance on operator experience.

CN118608874BActive Publication Date: 2024-12-03NORTHEASTERN UNIV CHINA
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411070308.3
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-08-06
Publication Date
2024-12-03
Estimated Expiration
2044-08-06

AI Technical Summary

Technical Problem

The selection of sampling points in the anode furnace refining process is largely dependent on operator experience, leading to delays and increased smelting costs due to subjective factors, reducing production efficiency and passive production.

Method used

A method involving image acquisition, preprocessing, feature extraction, and a trained judgment model to identify sampling points in the late reduction period based on furnace mouth flame images, utilizing color, texture, and shape features.

Benefits of technology

Accurately identifies sampling points, reducing smelting costs and improving production efficiency by avoiding delays caused by human subjective factors.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118608874B_ABST
    Figure CN118608874B_ABST
Patent Text Reader

Abstract

The present application relates to the field of metallurgy. The present invention discloses a method, device, equipment and medium for judging the sampling point of the reduction period of an anode furnace refining. The method comprises: obtaining a furnace mouth flame image; obtaining a processed furnace mouth flame image according to the furnace mouth flame image based on a pre-set image preprocessing model; obtaining color features, texture features and shape features in the processed furnace mouth flame image based on a pre-set image feature extraction model; inputting the color features, texture features and shape features into the trained reduction period sampling point judgment model to judge whether the time point corresponding to the furnace mouth flame image is a sampling point in the late reduction period. The present application identifies the sampling point in the late reduction period based on the reduction period sampling point judgment model by extracting color features, shape features and texture features from the flame image, thereby improving the accuracy of recognition, avoiding the increase in smelting costs caused by delaying the refining time due to subjective human factors, increasing the initiative of production and improving production efficiency.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the field of metallurgical technology, and in particular to a method, device, equipment and medium for determining sampling points during the refining reduction period of an anode furnace. Background Art

[0002] Copper mainly comes from pyrometallurgy. The pyrometallurgy copper smelting process mainly includes matte smelting, copper matte blowing, crude copper pyrometallurgy and anode copper electrolytic refining. The purpose of crude copper pyrometallurgy is to remove impurities and provide qualified anode copper for electrolytic refining. The process involved is the anode furnace refining process. The anode furnace refining process is mainly divided into four periods: insulation, oxidation, slag removal and reduction. The end point of the reduction period represents the end point of the entire refining. In the entire refining process, the main production cost comes from the consumption of fuel materials. At present, the end point of anode copper refining is still judged by manual multiple sampling. The selection of sampling points depends entirely on the work experience and work attitude of the operator. Human subjective factors delay the refining time, which will increase the smelting cost, cause passive production and reduce production efficiency.

[0003] At present, the research and application of anode furnace refining technology are still very scarce, and mainly focus on the improvement of anode furnace refining equipment. Moreover, due to the complex mechanism of the anode furnace refining process, it is difficult to establish a high-precision mechanism model to simulate and judge the anode furnace refining state. Therefore, it is particularly important to realize intelligent judgment of anode furnace refining sampling points. Summary of the invention

[0004] In view of the above situation, the embodiments of the present application provide a method, device, equipment and medium for determining the sampling points during the refining reduction period of an anode furnace, aiming to solve the above problem or at least partially solve the above problem.

[0005] In a first aspect, a method for determining sampling points during the refining reduction period of an anode furnace is provided, comprising:

[0006] Acquire furnace mouth flame image;

[0007] Based on the preset image preprocessing model, a processed furnace mouth flame image is obtained according to the furnace mouth flame image;

[0008] Based on a preset image feature extraction model, the color features, texture features, and shape features in the processed furnace mouth flame image are obtained;

[0009] The color feature, the texture feature, and the shape feature are input into the trained reduction period sampling point judgment model to judge whether the time point corresponding to the furnace mouth flame image is a sampling point in the late reduction period.

[0010] Preferably, the reduction period sampling point judgment model is obtained by training according to the following steps:

[0011] Acquire the flame image of furnace mouth during reduction period;

[0012] Based on a preset image preprocessing model, and according to the reduction period furnace mouth flame image, a reduction period furnace mouth flame processed image is obtained;

[0013] Based on a preset image feature extraction model, the color features, texture features, and shape features in each image after the reduction period furnace mouth flame treatment are obtained;

[0014] Marking the reduction phases of the color feature, the texture feature, and the shape feature, respectively, and obtaining a label sample set according to the marked color feature, the texture feature, and the shape feature;

[0015] The reduction period sampling point judgment model is obtained by training according to the label sample set.

[0016] Preferably, the step of obtaining the reduction period furnace mouth flame image comprises:

[0017] Acquire the furnace mouth flame image in the early stage of reduction period, acquire the furnace mouth flame image in the middle stage of reduction period, and acquire the furnace mouth flame image in the late stage of reduction period;

[0018] The step of marking the reduction phases of the color feature, the texture feature, and the shape feature, and obtaining a label sample set according to the marked color feature, the texture feature, and the shape feature, comprises:

[0019] Marking the color feature, texture feature, and shape feature corresponding to the furnace mouth flame image in the early stage of the reduction period as a first value;

[0020] Marking the color feature, texture feature, and shape feature corresponding to the furnace mouth flame image in the middle period of the reduction period as a second value;

[0021] The color feature, texture feature, and shape feature corresponding to the furnace mouth flame image in the late reduction period are marked as a third value.

[0022] Preferably, before obtaining the reduction period sampling point judgment model through training according to the label sample set, the method further comprises:

[0023] Dividing the labeled sample set into a training set and a test set according to a preset ratio;

[0024] The method of obtaining the reduction period sampling point judgment model according to the label sample set training includes:

[0025] Based on the TPE algorithm and the cross-validation algorithm, the pre-set LightGBM model is optimized for hyperparameters according to the training set to obtain the optimal hyperparameter combination;

[0026] The LightGBM model is optimized and trained according to the training set and the optimal hyperparameter combination to obtain the reduction period sampling point judgment model.

[0027] Preferably, after obtaining the reduction period sampling point judgment model through training according to the label sample set, the method further comprises:

[0028] Testing the reduction period sampling point judgment model according to the test set;

[0029] Based on the preset evaluation model, multiple evaluation indicators of the reduction period sampling point judgment model are calculated, wherein the evaluation indicators include: accuracy, precision, recall, F 1-score value, Matthews correlation coefficient and Kappa Statistics;

[0030] The reduction period sampling point judgment model is evaluated according to the plurality of evaluation indicators.

[0031] Preferably, the evaluation of the reduction period sampling point judgment model according to the plurality of evaluation indicators includes:

[0032] Based on the preset evaluation model, calculate the multiple evaluation indicators of the SVM classifier, KNN model, LR classifier, decision tree DT, RF algorithm, AdaBoost model and Xgboost model;

[0033] According to the plurality of evaluation indicators, the reduction period sampling point judgment model, SVM classifier, KNN model, LR classifier, decision tree DT, RF algorithm, AdaBoost model and Xgboost model are compared.

[0034] Preferably, the color features, texture features, and shape features in the processed furnace mouth flame image are obtained based on a preset image feature extraction model, including:

[0035] Converting the processed furnace mouth flame image from RGB space to HSV space;

[0036] Based on the color moment algorithm, the color feature is extracted according to the processed furnace mouth flame image converted into the HSV space;

[0037] Extracting the texture feature based on the processed furnace mouth flame image converted to the HSV space based on the color co-occurrence matrix and the Gabor filter;

[0038] Based on the Hu invariant moment algorithm, the shape feature is extracted according to the processed furnace mouth flame image in RGB space.

[0039] In a second aspect, a device for determining sampling points during the refining reduction period of an anode furnace is provided, comprising:

[0040] An image acquisition module is used to obtain the furnace mouth flame image;

[0041] A preprocessing module, used for obtaining a processed furnace mouth flame image according to the furnace mouth flame image based on a preset image preprocessing model;

[0042] A feature extraction module, used to obtain color features, texture features, and shape features in the processed furnace mouth flame image based on a preset image feature extraction model;

[0043] The judgment module is used to input the color feature, the texture feature, and the shape feature into the reduction period sampling point judgment model obtained by training to judge whether the time point corresponding to the furnace mouth flame image is a sampling point in the late reduction period.

[0044] In a third aspect, a computer device is provided, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the steps of the method for determining the sampling point of the anode furnace refining reduction period as described in the first aspect are implemented.

[0045] In a fourth aspect, a computer-readable storage medium is provided, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the method for determining the sampling point of the anode furnace refining reduction period as described in the first aspect are implemented.

[0046] At least one of the above technical solutions adopted in the embodiments of the present application can achieve the following beneficial effects:

[0047] The present application collects furnace mouth flame images, extracts color features, shape features and texture features in the image, and the reduction period sampling point judgment model can accurately identify the sampling points in the late reduction period based on the color features, shape features and texture features, thereby improving the accuracy of recognition, avoiding the increase in smelting costs caused by delays in refining time due to human subjective factors, increasing production initiative and improving production efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0049] Figure 1 It is a schematic diagram of an application environment of a method for determining sampling points during the refining reduction period of an anode furnace in one embodiment of the present invention;

[0050] Figure 2 It is a flow chart of a method for determining sampling points during the refining reduction period of an anode furnace according to an embodiment of the present invention;

[0051] Figure 3 It is a schematic diagram of the training process of the reduction period sampling point judgment model in one embodiment of the present invention;

[0052] Figure 4 is a flowchart of hyperparameter optimization through a tree Parzen estimator (TPE) in one embodiment of the present invention;

[0053] Figure 5 It is a schematic diagram of a process of performing hyperparameter optimization through 10-fold cross training in one embodiment of the present invention;

[0054] Figure 6 Schematic diagram of the LightGBM model classification confusion matrix in one embodiment of the present invention;

[0055] Figure 7 It is a flow chart of a device for determining sampling points during the refining reduction period of an anode furnace according to an embodiment of the present invention;

[0056] Figure 8 is a schematic diagram of a structure of a computer device in one embodiment of the present invention;

[0057] Fig. 9 It is another structural schematic diagram of a computer device in one embodiment of the present invention. DETAILED DESCRIPTION

[0058] In order to make the purpose, technical solution and advantages of the present application clearer, the technical solution of the present application will be clearly and completely described below in combination with the specific embodiments of the present application and the corresponding drawings. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present application.

[0059] It should be noted that the terms "first", "second", etc. in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that such use is interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein. In addition, the term "including" and its variants are to be interpreted as open-ended terms meaning "including but not limited to".

[0060] The technical solutions provided by various embodiments of the present application are described in detail below in conjunction with the accompanying drawings.

[0061] As mentioned above, the current selection of sampling points in the late reduction period is mainly based on the experience of operators, which may delay refining time and increase smelting costs due to subjective factors, resulting in passive production and reduced production efficiency. In order to solve this technical problem, the present application embodiment provides a method for determining sampling points in the reduction period of an anode furnace refining.

[0062] The method for determining the sampling point of the anode furnace refining reduction period provided by the embodiment of the present invention can be applied in the following aspects: Figure 1 In an application environment, the device side communicates with the server side through a network. The server side can obtain the furnace mouth flame image through the device side, and based on the preset image preprocessing model, obtain the processed furnace mouth flame image according to the furnace mouth flame image; based on the preset image feature extraction model, obtain the color features, texture features, and shape features in the processed furnace mouth flame image; input the color features, texture features, and shape features into the trained reduction period sampling point judgment model to judge whether the time point corresponding to the furnace mouth flame image is a late reduction period sampling point. The present application collects the furnace mouth flame image, extracts the color features, shape features, and texture features in the image, and the reduction period sampling point judgment model can accurately identify the late reduction period sampling point according to the color features, shape features, and texture features, thereby improving the accuracy of recognition, avoiding the increase in smelting costs due to delays in refining time due to subjective human factors, increasing production initiative, improving production efficiency, and meeting production capacity requirements.

[0063] The device side may be, but is not limited to, various personal computers, laptops, smart phones, tablet computers, and portable wearable devices. The server side may be implemented by an independent server or a server cluster consisting of multiple servers. The present invention is described in detail below through specific embodiments.

[0064] See also Figure 2 As shown, Figure 2 A flow chart of a method for determining sampling points during the refining reduction period of an anode furnace provided in an embodiment of the present invention includes the following steps:

[0065] S10: Acquire a furnace mouth flame image. Specifically, use an industrial camera to acquire a furnace mouth flame image of the anode furnace.

[0066] S20: Based on a preset image preprocessing model, a processed furnace mouth flame image is obtained according to the furnace mouth flame image.

[0067] Specifically, the image preprocessing includes cropping and segmentation. Cropping is to crop the furnace mouth flame image to a preset size, which is 1200×800, so that the flame in the furnace mouth flame image is as close to the middle of the image as possible; segmentation is to extract the flame from the cropped furnace mouth flame image using adaptive threshold segmentation, except for the area where the flame is located, all is black. After cropping and segmentation, a flame image with the background removed can be obtained, that is, the processed furnace mouth flame image.

[0068] S30: Based on a preset image feature extraction model, the color features, texture features, and shape features in the processed furnace mouth flame image are obtained.

[0069] Specifically, step S30 includes:

[0070] S31: Convert the processed furnace mouth flame image from RGB space to HSV space.

[0071] Specifically, for any coordinate point in the flame image, its RGB color space is (R, G, B), and its HSV color space is (H, S, V). First, we need to convert the R (red), G (green), and B (blue) values ​​from the range of 0 to 255 to the range of 0 to 1 using the following formula:

[0072]

[0073] Then calculate the H (hue), S (saturation), and V (brightness) values ​​in the HSV color space. The calculation process is as follows:

[0074]

[0075] If the H´ value is less than 0, the following calculation is required:

[0076]

[0077] Finally, in the open source computer vision library Opencv (Open Source Computer Vision Library), each value is converted to between 0 and 255 to obtain the final H, S, and V:

[0078]

[0079]

[0080]

[0081] S32: Based on the color moment algorithm, extract the color feature according to the processed furnace mouth flame image converted into the HSV space.

[0082] Specifically, the first-order moment:

[0083] Second moment:

[0084] Third moment:

[0085] in, Indicates the total number of pixels, Indicates The color channel is in image pixel values; E i Indicates i The mean of the color channels; σ i Indicates i The standard deviation of each color channel; S i Indicates i The first-order moment can reflect the brightness of the flame, the second-order moment can reflect the color distribution of the flame, and the third-order moment can reflect the symmetry of the flame color distribution.

[0086] The mean of the three channels R, G, and B of the flame image is calculated as the supplement of the color feature in the RGB space.

[0087] S33: extracting the texture feature according to the processed furnace mouth flame image converted into the HSV space based on the color co-occurrence matrix and the Gabor filter.

[0088] Specifically, in terms of the overall image, the color co-occurrence matrix (CCM) is used to extract texture features from the processed furnace mouth flame image. In the HSV space, the H, S and V channels are separated and spatially quantized to obtain the pseudo grayscale images of each color channel, and the color co-occurrence matrix of each pseudo grayscale image is calculated respectively. According to the color co-occurrence matrix, the secondary statistic Haralick feature is extracted as a description of the texture of the original color image. Here, the angular second moment (AngularSecond Moment, ASM), entropy (entropy, ENT), inverse difference moment (Inverse Different Moment, IDM), contrast (contrast, CON) and correlation (correlation, COR) are selected for characterization, and the feature calculation is as follows:

[0089]

[0090] The sum of the squares of all elements of the co-occurrence matrix is ​​the angular second moment.

[0091]

[0092] The entropy value represents the disorder of the image texture, that is, the information value.

[0093]

[0094] The inverse difference moment represents the uniformity of the local texture of the image.

[0095]

[0096] Contrast refers to the depth of the image texture grooves.

[0097]

[0098] in,

[0099]

[0100] The correlation coefficient represents the correlation of local textures.

[0101] Among them, the above formula Line Column elements express; Indicates the grayscale quantization level of each channel; and are pixel coordinates; is the standard deviation; is the mean; is the image pixel value.

[0102] In terms of local image, Gabor filter can effectively extract local texture information in different scales and directions. The real part of Gabor filter is used to process the local texture features of the furnace flame image, and the grayscale image is filtered to capture the texture features. The mathematical expression is:

[0103]

[0104] in, and are pixel coordinates, is the standard deviation of the filter, is the ellipticity of the filter, is the filter frequency, is the phase shift.

[0105] S34: Based on the Hu invariant moment algorithm, the shape feature is extracted according to the processed furnace mouth flame image in the RGB space.

[0106] Specifically, the Hu invariant moment algorithm is used to extract the grayscale distribution. The processed furnace flame image is The order normalized central moment is defined as follows:

[0107]

[0108] In the formula, A is the center distance, subscript pq represents the order of the moment, where p and q are positive integers, representing the powers in the X and Y directions respectively;

[0109] in,

[0110] ; ; ; Represents pixel coordinates in an image.

[0111] The following set of invariant moments can be derived using the second-order and third-order normalized central moments, which remain unchanged when the image is translated, rotated, and scaled:

[0112]

[0113] In the formula, I represents the moment invariant; y represents the pixel coordinate in the image.

[0114] S40: Inputting the color feature, the texture feature, and the shape feature into the trained reduction period sampling point judgment model to judge whether the time point corresponding to the furnace mouth flame image is a sampling point in the late reduction period.

[0115] Specifically, when it is determined that it is a sampling point in the late reduction period, sampling is performed, and whether it is the reduction end point is determined based on the sampling result, thereby avoiding errors in manual judgment of sampling points, thereby reducing production costs, improving production efficiency, and meeting production capacity requirements.

[0116] Preferably, see Figure 3 As shown, the reduction period sampling point judgment model is trained and obtained according to the following steps:

[0117] S00: Acquire the flame image of the furnace mouth during the reduction period.

[0118] Specifically, an industrial camera is used to collect the furnace mouth flame image in the early stage of the reduction period, the furnace mouth flame image in the middle stage of the reduction period, and the furnace mouth flame image in the late stage of the reduction period.

[0119] Specifically, in one embodiment, the time of the late reduction period is extremely short, and the late reduction period is determined to be the reduction period sampling point, and sampling can be performed, or the late reduction period can be determined to be the reduction end point.

[0120] Specifically, in one embodiment, the flame images of each stage of the anode furnace refining and reduction are confirmed by workers at the refining site.

[0121] S01: Based on a preset image preprocessing model, and according to the reduction period furnace mouth flame image, a reduction period furnace mouth flame processed image is obtained.

[0122] Specifically, the image preprocessing includes cropping and segmentation. Cropping is to crop the furnace mouth flame image in the early stage of the reduction period, the furnace mouth flame image in the middle stage of the reduction period, and the furnace mouth flame image in the late stage of the reduction period to a preset size, which is 1200×800, so that the flame in the furnace mouth flame image is located in the middle of the image as much as possible; segmentation is to extract the flame from the cropped furnace mouth flame image in the early stage of the reduction period, the furnace mouth flame image in the middle stage of the reduction period, and the furnace mouth flame image in the late stage of the reduction period respectively by using adaptive threshold segmentation, and all areas except the area where the flame is located are black. After cropping and segmentation, a flame image with the background removed can be obtained, that is, the processed image of the furnace mouth flame in the reduction period.

[0123] S02: Based on a preset image feature extraction model, the color features, texture features, and shape features in each image after the reduction period furnace mouth flame treatment are obtained.

[0124] Specifically, step S02 includes:

[0125] S021: Convert the image after the reduction period furnace mouth flame treatment from RGB space to HSV space.

[0126] Specifically, for any coordinate point in the flame image, its RGB color space is (R, G, B), and its HSV color space is (H, S, V). First, we need to convert the R, G, and B values ​​to between 0 and 1 using the following formula:

[0127]

[0128] Then calculate the H, S, and V values ​​in the HSV color space. The calculation process is as follows:

[0129]

[0130] If the H´ value is less than 0, the following calculation is required:

[0131]

[0132] Finally, in the open source computer vision library Opencv (Open Source Computer Vision Library), each value is converted to between 0 and 255 to obtain the final H, S, and V:

[0133]

[0134]

[0135]

[0136] S32: Based on the color moment algorithm, the color feature is extracted according to the image of the furnace mouth flame after the reduction period is converted into the HSV space.

[0137] Specifically, the first-order moment:

[0138] Second moment:

[0139] Third moment:

[0140] in, Indicates the total number of pixels, Indicates The color channel is in image pixel values; E i Indicates i The mean of the color channels; σ i Indicates i The standard deviation of each color channel; S i Indicates i The first-order moment can reflect the brightness of the flame, the second-order moment can reflect the color distribution of the flame, and the third-order moment can reflect the symmetry of the flame color distribution.

[0141] The mean of the three channels R, G, and B of the flame image is calculated as the supplement of the color feature in the RGB space.

[0142] S33: Based on the color co-occurrence matrix and Gabor filter, the texture features are extracted according to the image after the reduction period furnace mouth flame treatment converted to the HSV space.

[0143] Specifically, in terms of the overall image, the color co-occurrence matrix (CCM) is used to extract texture features from the image after the flame treatment of the furnace mouth in the reduction period. In the HSV space, the H, S and V channels are separated and spatially quantized to obtain the pseudo grayscale image of each color channel, and the color co-occurrence matrix of each pseudo grayscale image is calculated respectively. The secondary ensemble is extracted based on the color co-occurrence matrix.

[0144] The Haralick feature is used as the description of the original color image texture. Here, the angular second moment (AngularSecond

[0145] Moment (ASM), entropy (ENT), inverse different moment (IDM), contrast (CON) and correlation (COR) are used for characterization. The feature calculation is as follows:

[0146]

[0147] The sum of the squares of all elements of the co-occurrence matrix is ​​the angular second moment.

[0148]

[0149] The entropy value represents the disorder of the image texture, that is, the information value.

[0150]

[0151] The inverse difference moment represents the uniformity of the local texture of the image.

[0152]

[0153] Contrast refers to the depth of the image texture grooves.

[0154]

[0155] in,

[0156]

[0157] The correlation coefficient represents the correlation of local textures.

[0158] Among them, the above formula Line Column elements express; Indicates the grayscale quantization level of each channel; and are pixel coordinates; is the standard deviation; is the mean; is the image pixel value.

[0159] In terms of local image, Gabor filter can effectively extract local texture information in different scales and directions. The real part of Gabor filter is used to process the local texture features of the furnace flame image, and the grayscale image is filtered to capture the texture features. The mathematical expression is:

[0160]

[0161] in, and are pixel coordinates, is the standard deviation of the filter, δ is the ellipticity of the filter, is the filter frequency, is the phase shift.

[0162] S34: Based on the Hu invariant moment algorithm, the shape feature is extracted according to the image of the furnace mouth after the reduction period flame treatment in the RGB space.

[0163] Specifically, the Hu invariant moment algorithm is used to extract the grayscale distribution. The processed furnace flame image is The order normalized central moment is defined as follows:

[0164]

[0165] In the formula, A is the center distance, subscript pq represents the order of the moment, where p and q are positive integers, representing the powers in the X and Y directions respectively;

[0166] in,

[0167] ; ; ; Represents pixel coordinates in an image.

[0168] The following set of invariant moments can be derived using the second-order and third-order normalized central moments, which remain unchanged when the image is translated, rotated, and scaled:

[0169]

[0170] In the formula, I represents the moment invariant; y represents the pixel coordinate in the image.

[0171] S03: Marking the reduction phases of the color feature, the texture feature, and the shape feature, respectively, and obtaining a label sample set according to the marked color feature, the texture feature, and the shape feature.

[0172] Specifically, the color feature, texture feature, and shape feature corresponding to the furnace mouth flame image in the early stage of the reduction period are marked as a first value;

[0173] Marking the color feature, texture feature, and shape feature corresponding to the furnace mouth flame image in the middle period of the reduction period as a second value;

[0174] The color feature, texture feature, and shape feature corresponding to the furnace mouth flame image in the late reduction period are marked as a third value.

[0175] Preferably, the first value is 0, the second value is 1, and the third value is 2.

[0176] S04: The reduction period sampling point judgment model is obtained by training according to the label sample set.

[0177] Specifically, before obtaining the reduction period sampling point judgment model through training according to the label sample set in step S04, the method further includes:

[0178] The labeled sample set is divided into a training set and a test set according to a preset ratio.

[0179] Specifically, the preset ratio is 7:3, that is, the ratio of the training set to the test set is 7:3.

[0180] Specifically, the step S04 of obtaining the reduction period sampling point judgment model by training according to the label sample set includes:

[0181] S041: Based on the TPE algorithm and the cross-validation algorithm, the hyperparameters of the pre-set LightGBM model are optimized according to the training set to obtain the optimal hyperparameter combination.

[0182] S042: Optimize and train the LightGBM model according to the training set and the optimal hyperparameter combination to obtain the reduction period sampling point judgment model.

[0183] In this embodiment, through performance comparison among many existing models such as LightGBM model, support vector machine (SVM) classifier, K Nearest Neighbors (KNN) model, Logistic Regression (LR) classifier, decision tree (DT), random forest (RF) algorithm, AdaBoost model, Xgboost model, etc., the recovery period sampling point judgment model based on the LightGBM model structure training is finally selected.

[0184] Specifically, the LightGBM model builds learners in series, improves the effect of weak classifiers, and combines them to obtain strong classifiers. The base learner of the LightGBM model uses a decision tree. Each iteration builds a new decision tree through the residual between the true value. The objective function is expanded using the Taylor formula to obtain:

[0185]

[0186]

[0187] in, Is the current The predicted value of the step; It is The predicted value of the step; is the new prediction function added; is the regularization term; is the first-order derivative of the loss function; is the second-order derivative of the loss function; l is the cross entropy loss function.

[0188] Combining regularization and objective function, we can get:

[0189]

[0190] in, No. The sample set of leaf nodes, , , is the number of leaf nodes, is the regularization coefficient.

[0191] The objective function is Taking the first-order derivative gives The weight corresponding to each leaf node is:

[0192]

[0193] Substituting into the objective function can be simplified to:

[0194]

[0195] That is, add a new tree in each iteration, using Generate and add this tree to the model Among them is the learning rate.

[0196] Specifically, in one embodiment, a 10-fold cross validation is performed based on the training set, and the cross entropy loss function is calculated, wherein TPE is used to determine the LightGBM model hyperparameters, including the base model, leaf nodes, maximum depth, and reduction rate, etc. The hyperparameters that need to be optimized are shown in Table 1, including pruning hyperparameters: max_depth, num_leaves, feature and data processing hyperparameters: colsample_bytree, Boosting process control hyperparameters: n_estimators, learning_rate, reg_alpha, and reg_lambda. The hyperparameters that minimize the 10-fold average cross entropy loss are selected as the optimal hyperparameters.

[0197] Table 1 Hyperparameter meaning table

[0198]

[0199] Specifically, we use the Tree Parzen Estimator (TPE) and cross-validation to optimize the hyperparameters of the LightGBM model. The core of TPE is the proxy model and the acquisition function. The proxy model models the target function, and the input is the observed function value. After training, we can get an estimate of the target function. The acquisition function determines the location of the next observation of the target function value, that is, the location of the next sampling point. The acquisition function is derived from the proxy model. In TPE, Available and To express, yes The following divisions are made:

[0200]

[0201] In each trial, TPE builds Gaussian mixture models for the hyperparameters associated with the best target value and the remaining hyperparameters and , substituting into the EI acquisition function, after a series of derivations, the following relationship is obtained:

[0202]

[0203] Select The corresponding hyperparameters at the maximum are used as the next set of search values.

[0204] See also Figure 4 As shown in the figure, the training process of the reduction period sampling point judgment model includes: defining the hyperparameter space and creating a multi-classification cross entropy loss function; using the tree Parzen estimator (TPE) and cross validation to optimize the hyperparameters of the LightGBM model to obtain the optimal parameters (optimal hyperparameter combination); training the LightGBM model, training the LightGBM model with the optimal parameters and training set; model prediction experiment, testing the trained LightGBM model according to the test set, and obtaining the reduction period sampling point judgment model. Among them, the steps of optimization through TPE are as follows:

[0205] (1) Randomly select observation values, which are the set of hyperparameters to be optimized;

[0206] (2) Sort and group the obtained observations;

[0207] (3) Parzen estimator models the density distribution of two groups, that is, Gaussian mixture models are established based on the sorted and grouped observations. and ;

[0208] (4) Extract and evaluate the sample hyperparameters and obtain the next sampling point according to the EI acquisition function;

[0209] (5) Return the improved hyperparameter set, that is, add the new observations to the parameter adjustment record;

[0210] (6) Evaluate the objective function obj;

[0211] (7) Repeat the above steps until the number of iterations is met.

[0212] Specifically, when using TPE to optimize LightGBM hyperparameters, the following is done: perform 10-fold cross validation in the training set and calculate the cross entropy loss function. The cross validation is as follows: Figure 5 As shown, the training set data is divided into 10 groups, each subset data is used as a validation set, and the remaining 9 groups of subset data are used as training subsets. In this way, 10 rounds of training (K1, K2...K10) are performed, and the cross entropy loss (E1, E2...E10) of each round is calculated. The average cross entropy loss is obtained based on the cross entropy loss (E1, E2...E10) of 10 rounds of training , according to the average cross entropy loss The optimal hyperparameter combination is obtained, and then the entire training set is substituted into the optimal hyperparameter combination to train the LightGBM model to obtain the optimal LightGBM model (TPE-LightGBM), which is the reduction period sampling point judgment model.

[0213] Preferably, after obtaining the reduction period sampling point judgment model through training according to the label sample set, the method further comprises:

[0214] Testing the reduction period sampling point judgment model according to the test set;

[0215] Specifically, the test set is input into the reduction period sampling point judgment model for testing to realize the recognition of the sampling points in the later stage of anode furnace refining reduction. The recognition results are as follows: Figure 6 As shown, it can be seen that the reduction period sampling point judgment model can well realize the identification of the three periods of anode furnace refining reduction, namely, early reduction period, middle reduction period and late reduction period.

[0216] Figure 6 In , the sum of each row represents the number of true samples of that category, and the sum of each column represents the number of samples predicted to be of that category.

[0217] Based on the preset evaluation model, multiple evaluation indicators of the reduction period sampling point judgment model are calculated, wherein the evaluation indicators include: accuracy, precision, recall, F 1-score value, Matthews correlation coefficient and Kappa statistic;

[0218] Specifically, accuracy Accuracy Calculated by the following formula:

[0219]

[0220] Accuracy Precision Calculated by the following formula:

[0221]

[0222] Recall Recall Calculated by the following formula:

[0223]

[0224] F 1-score is calculated by the following formula:

[0225]

[0226] Matthews correlation coefficient MCC Calculated by the following formula:

[0227]

[0228] Kappa The statistic is calculated as follows:

[0229]

[0230] In the above formula, is the overall accuracy; is the accidental consistency error; T (Ture) represents the correct prediction; F (False) indicates a prediction error; P (Positive) represents 1; N (Negative) represents 0; TP It means the prediction is 1, the prediction is correct, that is, the actual value is 1; FP Indicates that the prediction is 1, the prediction is wrong, that is, the actual 0; FN Indicates that the prediction is 0, the prediction is wrong, that is, the actual 1; TN It means the prediction is 0, and the prediction is correct, which means the actual value is 0.

[0231] The reduction period sampling point judgment model is evaluated according to the plurality of evaluation indicators.

[0232] Specifically, in one embodiment, a reference value of each evaluation index is preset, the calculated evaluation index is compared with the reference value of the corresponding evaluation index, and the performance of the reduction period sampling point judgment model is evaluated based on the comparison result.

[0233] Preferably, the method further comprises:

[0234] Based on the preset evaluation model, calculate the multiple evaluation indicators of the SVM classifier, KNN model, LR classifier, decision tree DT, RF algorithm, AdaBoost model and Xgboost model;

[0235] According to the plurality of evaluation indicators, the reduction period sampling point judgment model, SVM classifier, KNN model, LR classifier, decision tree DT, RF algorithm, AdaBoost model and Xgboost model are compared.

[0236] Specifically, the model is compared with seven other classification models optimized by TPE: support vector machine (SVM) classifier, KNearest Neighbors (KNN) model, Logistic Regression (LR) classifier, decision tree (DT), random forest (RF) algorithm, AdaBoost model and Xgboost model. The comparison results are shown in Table 2:

[0237] Table 2 Comparison of model evaluation indicators

[0238]

[0239] The comparison of various indicators showed that the indicators of the reduction period sampling point judgment model were better than those of the other seven classifiers, including Acc, Pre-macro, Rec-macro, F 1 -macro, MCC and Kappa reached 0.990588, 0.990844, 0.989599, 0.990192, 0.985695 and 0.985661 respectively. In terms of unidirectional indicators, Acc, Pre-macro and Rec-macro indicators show that the reduction period sampling point judgment model can realize the accurate identification of the early, middle and late reduction of anode furnace refining reduction, F 1The -macro index shows the robustness of the reduction period sampling point judgment model, and the MCC index shows that the comprehensive prediction ability of the reduction period sampling point judgment model is very good. In addition, the Kappa of the reduction period sampling point judgment model is 0.985661, indicating that its prediction results and actual results are almost completely consistent, while the Kappa values ​​of DT and LR are 0.805783 and 0.870441. Compared with other machine learning methods, the reduction period sampling point judgment model proposed in the invention has the best performance in all aspects.

[0240] It should be noted that, from many existing models such as support vector machine (SVM) classifier, K Nearest Neighbors (KNN) model, Logistic Regression (LR) classifier, decision tree (DT), random forest (RF) algorithm, AdaBoost model, Xgboost model, LightGBM model, etc., through performance comparison, the final reduction period sampling point judgment model of the embodiment of the present application obtained by training with the LightGBM model structure was finally selected. The multiple evaluation indicators in Table 2 can clearly prove that the reduction period sampling point judgment model of the present application has better performance. The whole process involves the selection of evaluation indicators, the calculation and comparison of the performance of various model indicators, which requires the inventor to pay more intellectual labor, so as to finally select the reduction period sampling point judgment model with the best performance in various aspects, so as to improve the accuracy and effectiveness of the intelligent judgment of the sampling point.

[0241] In addition, the present invention uses TPE to optimize the hyperparameters of the LightGBM model to obtain a reduction period sampling point judgment model. Since the core of TPE is to estimate the objective function and estimate the position of the next sampling point in real time, it can effectively reduce invalid searches and redundant calculations, and more reliably find the global optimal solution.

[0242] It can be seen that in the above scheme, by collecting the furnace mouth flame image and extracting the color features, shape features and texture features in the image, the reduction period sampling point judgment model can accurately identify the sampling points in the late reduction period according to the color features, shape features and texture features, thereby improving the accuracy of recognition, avoiding the increase in smelting costs caused by delays in refining time due to human subjective factors, increasing production initiative, and improving production efficiency.

[0243] It should be understood that the order of execution of the steps in the above embodiment does not necessarily mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiment of the present invention.

[0244] In one embodiment, a device for determining the sampling point of the anode furnace refining reduction period is provided, and the device for determining the sampling point of the anode furnace refining reduction period corresponds one-to-one to the method for determining the sampling point of the anode furnace refining reduction period in the above embodiment. Figure 7 As shown, the anode furnace refining reduction period sampling point determination device includes an image acquisition module 101, a preprocessing module 102, a feature extraction module 103, and a determination module 104.

[0245] In the application process of the reduction period sampling point judgment model, each functional module is described in detail as follows:

[0246] An image acquisition module 101 is used to acquire a furnace mouth flame image;

[0247] A preprocessing module 102 is used to obtain a processed furnace mouth flame image according to the furnace mouth flame image based on a preset image preprocessing model;

[0248] A feature extraction module 103 is used to obtain color features, texture features, and shape features in the processed furnace mouth flame image based on a preset image feature extraction model;

[0249] The judgment module 104 is used to input the color feature, the texture feature, and the shape feature into the trained reduction period sampling point judgment model to judge whether the time point corresponding to the furnace mouth flame image is a sampling point in the late reduction period.

[0250] The preprocessing module 102 is further used to crop the furnace mouth flame image into a preset size so that the flame in the furnace mouth flame image is located in the middle of the image, and extract the flame from the cropped furnace mouth flame image using adaptive threshold segmentation.

[0251] The feature extraction module 103 is also used to convert the processed furnace mouth flame image from the RGB space to the HSV space; based on the color moment algorithm, extract the color feature according to the processed furnace mouth flame image converted to the HSV space; based on the color co-occurrence matrix and the Gabor filter, extract the texture feature according to the processed furnace mouth flame image converted to the HSV space; based on the Hu invariant moment algorithm, extract the shape feature according to the processed furnace mouth flame image in the RGB space.

[0252] The anode furnace refining reduction period sampling point determination device also includes: a label marking module 105, a data division module 106, a model training module 107, a model classification module 108 and a model evaluation module 109.

[0253] During the training process of the reduction period sampling point judgment model, each functional module is described in detail as follows:

[0254] The image acquisition module 101 is also used to obtain the flame image of the furnace mouth in the reduction period, specifically, to obtain the flame image of the furnace mouth in the early reduction period, the flame image of the furnace mouth in the middle reduction period, and the flame image of the furnace mouth in the late reduction period.

[0255] The preprocessing module 102 is further used to obtain a processed image of the furnace mouth flame in the reduction period according to the furnace mouth flame image in the reduction period based on a preset image preprocessing model.

[0256] The preprocessing module 102 is specifically used to crop the furnace mouth flame image in the early stage of the reduction period, the furnace mouth flame image in the middle stage of the reduction period, and the furnace mouth flame image in the late stage of the reduction period into a preset size, so that the flame in the furnace mouth flame image in the early stage of the reduction period, the furnace mouth flame image in the middle stage of the reduction period, and the furnace mouth flame image in the late stage of the reduction period is located in the middle of the image, and use adaptive threshold segmentation to extract the flame from the cropped furnace mouth flame image in the early stage of the reduction period, the furnace mouth flame image in the middle stage of the reduction period, and the furnace mouth flame image in the late stage of the reduction period.

[0257] The feature extraction module 103 is also used to obtain the color features, texture features, and shape features in each image after the reduction period furnace mouth flame treatment based on a preset image feature extraction model.

[0258] The feature extraction module 103 is specifically used to convert the image of the furnace mouth flame in the reduction period from the RGB space to the HSV space; based on the color moment algorithm, extract the color feature according to the image of the furnace mouth flame in the reduction period converted to the HSV space; based on the color co-occurrence matrix and the Gabor filter, extract the texture feature according to the image of the furnace mouth flame in the reduction period converted to the HSV space; based on the Hu invariant moment algorithm, extract the shape feature according to the image of the furnace mouth flame in the reduction period in the RGB space.

[0259] The label marking module 105 is used to mark the stages of the reduction period of the color feature, the texture feature, and the shape feature, and obtain a label sample set according to the marked color feature, the texture feature, and the shape feature.

[0260] The label marking module 105 is specifically used to mark the color features, texture features, and shape features corresponding to the furnace mouth flame image in the early reduction period as 0; mark the color features, texture features, and shape features corresponding to the furnace mouth flame image in the middle reduction period as 1; and mark the color features, texture features, and shape features corresponding to the furnace mouth flame image in the late reduction period as 2.

[0261] The data partitioning module 106 is used to partition the labeled sample set into a training set and a test set according to a preset ratio.

[0262] The model training module 107 is used to optimize the hyperparameters of the pre-set LightGBM model based on the training set based on the TPE algorithm and the cross-validation algorithm to obtain the optimal hyperparameter combination; optimize the training of the LightGBM model based on the training set and the optimal hyperparameter combination to obtain the reduction period sampling point judgment model.

[0263] The model classification module 108 is used to test the reduction period sampling point judgment model according to the test set.

[0264] The model evaluation module 109 is used to calculate multiple evaluation indicators of the reduction period sampling point judgment model based on a preset evaluation model, wherein the evaluation indicators include: accuracy, precision, recall, F1-score value, Matthews correlation coefficient and Kappa statistic.

[0265] The model evaluation module 109 is also used to calculate multiple evaluation indicators of the SVM classifier, KNN model, LR classifier, decision tree DT, RF algorithm, AdaBoost model and Xgboost model based on a pre-set evaluation model; and compare the reduction period sampling point judgment model, SVM classifier, KNN model, LR classifier, decision tree DT, RF algorithm, AdaBoost model and Xgboost model according to the multiple evaluation indicators.

[0266] The present invention provides an intelligent judgment method and device for sampling points in the reduction period of an anode furnace refining. By collecting a furnace mouth flame image and extracting color features, shape features and texture features in the image, a reduction period sampling point judgment model can accurately identify sampling points in the late reduction period according to color features, shape features and texture features, thereby improving recognition accuracy, avoiding delays in refining time due to human subjective factors and increasing smelting costs, increasing production initiative and improving production efficiency.

[0267] The specific definition of the sampling point judgment device for the refining reduction period of the anode furnace can be found in the definition of the sampling point judgment method for the refining reduction period of the anode furnace above, which will not be repeated here. Each module in the above-mentioned sampling point judgment device for the refining reduction period of the anode furnace can be implemented in whole or in part by software, hardware and a combination thereof. The above-mentioned modules can be embedded in or independent of the processor in the computer device in the form of hardware, or can be stored in the memory of the computer device in the form of software, so that the processor can call and execute the operations corresponding to the above modules.

[0268] In one embodiment, a computer device is provided. The computer device may be a server, and its internal structure diagram may be as follows: Figure 8As shown. The computer device includes a processor, a memory, a network interface and a database connected via a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile and / or volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The network interface of the computer device is used to communicate with an external device end through a network connection. When the computer program is executed by the processor, the functions or steps of the service end side of the sampling point judgment method for the refining reduction period of the anode furnace are realized.

[0269] In one embodiment, a computer device is provided. The computer device may be a device end, and its internal structure diagram may be as follows: Fig. 9 As shown. The computer device includes a processor, a memory, a network interface, a display screen and an input device connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The network interface of the computer device is used to communicate with an external server through a network connection. When the computer program is executed by the processor, the functions or steps of the device side of the sampling point judgment method for the refining reduction period of the anode furnace are realized.

[0270] In one embodiment, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the following steps when executing the computer program:

[0271] Acquire furnace mouth flame image;

[0272] Based on the preset image preprocessing model, a processed furnace mouth flame image is obtained according to the furnace mouth flame image;

[0273] Based on a preset image feature extraction model, the color features, texture features, and shape features in the processed furnace mouth flame image are obtained;

[0274] The color feature, the texture feature, and the shape feature are input into the trained reduction period sampling point judgment model to judge whether the time point corresponding to the furnace mouth flame image is a sampling point in the late reduction period.

[0275] You can also implement the following steps:

[0276] Acquire the flame image of furnace mouth during reduction period;

[0277] Based on a preset image preprocessing model, and according to the reduction period furnace mouth flame image, a reduction period furnace mouth flame processed image is obtained;

[0278] Based on a preset image feature extraction model, the color features, texture features, and shape features in each image after the reduction period furnace mouth flame treatment are obtained;

[0279] Marking the reduction phases of the color feature, the texture feature, and the shape feature, respectively, and obtaining a label sample set according to the marked color feature, the texture feature, and the shape feature;

[0280] The reduction period sampling point judgment model is obtained by training according to the label sample set.

[0281] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored, and when the computer program is executed by a processor, the following steps are implemented:

[0282] Acquire furnace mouth flame image;

[0283] Based on the preset image preprocessing model, a processed furnace mouth flame image is obtained according to the furnace mouth flame image;

[0284] Based on a preset image feature extraction model, the color features, texture features, and shape features in the processed furnace mouth flame image are obtained;

[0285] The color feature, the texture feature, and the shape feature are input into the trained reduction period sampling point judgment model to judge whether the time point corresponding to the furnace mouth flame image is a sampling point in the late reduction period.

[0286] You can also implement the following steps:

[0287] Acquire the flame image of furnace mouth during reduction period;

[0288] Based on a preset image preprocessing model, and according to the reduction period furnace mouth flame image, a reduction period furnace mouth flame processed image is obtained;

[0289] Based on a preset image feature extraction model, the color features, texture features, and shape features in each image after the reduction period furnace mouth flame treatment are obtained;

[0290] Marking the reduction phases of the color feature, the texture feature, and the shape feature, respectively, and obtaining a label sample set according to the marked color feature, the texture feature, and the shape feature;

[0291] The reduction period sampling point judgment model is obtained by training according to the label sample set.

[0292] It should be noted that the above functions or steps that can be implemented by the computer-readable storage medium or computer device can refer to the relevant descriptions on the server side and the device side in the aforementioned method embodiment. To avoid repetition, they will not be described one by one here.

[0293] Those of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be implemented by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application may include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0294] Those skilled in the art can clearly understand that for the convenience and simplicity of description, only the division of the above-mentioned functional units and modules is used as an example. In actual applications, the above-mentioned functions can be distributed and completed by different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above.

[0295] The embodiments described above are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. Such modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included in the protection scope of the present invention.

Claims

1. A method for determining sampling points during the refining reduction period of an anode furnace, characterized in that: include: Acquire furnace mouth flame image; Based on a preset image preprocessing model, a processed furnace mouth flame image is obtained according to the furnace mouth flame image, wherein the preset image preprocessing model is used for preprocessing the image, including cropping and segmentation; Based on a preset image feature extraction model, the color features, texture features, and shape features in the processed furnace mouth flame image are obtained, wherein the preset image feature extraction model is used to convert the processed furnace mouth flame image from the RGB space to the HSV space; based on the color moment algorithm, the color features are extracted according to the processed furnace mouth flame image converted to the HSV space; based on the color co-occurrence matrix and the Gabor filter, the texture features are extracted according to the processed furnace mouth flame image converted to the HSV space; based on the Hu invariant moment algorithm, the shape features are extracted according to the processed furnace mouth flame image in the RGB space; The color feature, the texture feature, and the shape feature are input into the trained reduction period sampling point judgment model to judge whether the time point corresponding to the furnace mouth flame image is a sampling point in the late reduction period, wherein the reduction period sampling point judgment model is trained according to the following steps: Acquire the flame image of furnace mouth during reduction period; Based on a preset image preprocessing model, and according to the reduction period furnace mouth flame image, a reduction period furnace mouth flame processed image is obtained; Based on a preset image feature extraction model, the color features, texture features, and shape features in each image after the reduction period furnace mouth flame treatment are obtained; Marking the reduction phases of the color feature, the texture feature, and the shape feature, respectively, and obtaining a label sample set according to the marked color feature, the texture feature, and the shape feature; The reduction period sampling point judgment model is obtained by training the label sample set; Before obtaining the reduction period sampling point judgment model through training according to the label sample set, the method further includes: Dividing the labeled sample set into a training set and a test set according to a preset ratio; The method of obtaining the reduction period sampling point judgment model according to the label sample set training includes: Define the hyperparameter space and create a multi-classification cross entropy loss function; Based on the TPE algorithm and the cross-validation algorithm, the pre-set LightGBM model is optimized for hyperparameters according to the training set to obtain the optimal hyperparameter combination, wherein the TPE algorithm is a tree Parzen estimator; The LightGBM model is optimized and trained according to the training set and the optimal hyperparameter combination, and the trained LightGBM model is tested according to the test set to obtain the reduction period sampling point judgment model.

2. The method for determining the sampling points of the anode furnace refining reduction period according to claim 1, characterized in that: The step of obtaining the reduction period furnace mouth flame image comprises: Acquire the furnace mouth flame image in the early stage of reduction period, acquire the furnace mouth flame image in the middle stage of reduction period, and acquire the furnace mouth flame image in the late stage of reduction period; The step of marking the reduction phases of the color feature, the texture feature, and the shape feature, and obtaining a label sample set according to the marked color feature, the texture feature, and the shape feature, comprises: Marking the color feature, texture feature, and shape feature corresponding to the furnace mouth flame image in the early stage of the reduction period as a first value; Marking the color feature, texture feature, and shape feature corresponding to the furnace mouth flame image in the middle period of the reduction period as a second value; The color feature, texture feature, and shape feature corresponding to the furnace mouth flame image in the late reduction period are marked as a third value.

3. The method for determining the sampling points of the anode furnace refining reduction period according to claim 1, characterized in that: After obtaining the reduction period sampling point judgment model through training according to the label sample set, the method further includes: Testing the reduction period sampling point judgment model according to the test set; Based on a preset evaluation model, multiple evaluation indicators of the reduction period sampling point judgment model are calculated, wherein the evaluation indicators include: at least one of accuracy, precision, recall, F1-score, Matthews correlation coefficient and Kappa statistic; The reduction period sampling point judgment model is evaluated according to the plurality of evaluation indicators.

4. The method for determining the sampling points during the refining reduction period of the anode furnace according to claim 3, characterized in that: The step of evaluating the reduction period sampling point judgment model according to the plurality of evaluation indicators includes: Based on a preset evaluation model, calculate multiple evaluation indicators of the SVM classifier, KNN model, LR classifier, decision tree DT, RF algorithm, AdaBoost model and / or Xgboost model; According to the plurality of evaluation indicators, the reduction period sampling point judgment model, SVM classifier, KNN model, LR classifier, decision tree DT, RF algorithm, AdaBoost model and Xgboost model are compared.

5. A device for determining sampling points during the refining reduction period of an anode furnace, characterized in that: include: An image acquisition module is used to obtain the furnace mouth flame image; A preprocessing module, used for obtaining a processed furnace mouth flame image according to the furnace mouth flame image based on a preset image preprocessing model, wherein the preset image preprocessing model is used for preprocessing the image, including cropping and segmentation; A feature extraction module, for obtaining color features, texture features, and shape features in the processed furnace mouth flame image based on a preset image feature extraction model, wherein the preset image feature extraction model is used to convert the processed furnace mouth flame image from RGB space to HSV space; based on a color moment algorithm, extract the color features according to the processed furnace mouth flame image converted to the HSV space; based on a color co-occurrence matrix and a Gabor filter, extract the texture features according to the processed furnace mouth flame image converted to the HSV space; based on a Hu invariant moment algorithm, extract the shape features according to the processed furnace mouth flame image in the RGB space; A judgment module, used for inputting the color feature, the texture feature, and the shape feature into the reduction period sampling point judgment model obtained by training, and judging whether the time point corresponding to the furnace mouth flame image is a sampling point in the late reduction period; During the training process of the reduction period sampling point judgment model, The image acquisition module is also used to obtain the flame image of the furnace mouth during the reduction period; The preprocessing module is also used to obtain a reduction period furnace mouth flame processed image according to the reduction period furnace mouth flame image based on a preset image preprocessing model; The feature extraction module is also used to obtain the color features, texture features, and shape features in each image after the reduction period furnace mouth flame treatment based on a preset image feature extraction model; A label marking module, used to mark the stages of the reduction period of the color feature, the texture feature, and the shape feature, and obtain a label sample set according to the marked color feature, the texture feature, and the shape feature; A data partitioning module, a model training module and a model classification module are used to train the reduction period sampling point judgment model according to the label sample set; A data partitioning module is used to divide the labeled sample set into a training set and a test set according to a preset ratio; A model training module is used to define a hyperparameter space and create a multi-classification cross entropy loss function; based on the TPE algorithm and the cross-validation algorithm, the pre-set LightGBM model is optimized for hyperparameters according to the training set to obtain the optimal hyperparameter combination, wherein the TPE algorithm is a tree Parzen estimator; the LightGBM model is optimized and trained according to the training set and the optimal hyperparameter combination; The model classification module is used to perform a test experiment on the trained LightGBM model according to the test set to obtain the reduction period sampling point judgment model.

6. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the steps of the method for determining the sampling point of the anode furnace refining reduction period as claimed in any one of claims 1 to 4 are implemented.

7. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the steps of the method for determining the sampling point of the anode furnace refining reduction period as claimed in any one of claims 1 to 4 are implemented.