Iron ore elemental classification method, apparatus, device, and storage medium
By preprocessing the spectral data of iron ore elements and inputting it into the trained classification model, the problems of high cost and time consumption in traditional methods are solved, and fast and accurate classification of iron ore elements is achieved.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- GUANGDONG WATSON INFORMATION TECHNOLOGY CO LTD
- Filing Date
- 2024-11-28
- Publication Date
- 2026-06-26
AI Technical Summary
Traditional methods for classifying iron ore elements are costly, time-consuming, and difficult to implement for rapid on-site testing.
By acquiring spectral data of iron ore elements, performing data preprocessing, and inputting the data into an iron ore element classification model, the model is classified using feature extraction and training.
This reduces the cost of elemental classification in iron ore and enables rapid and accurate detection.
Smart Images
Figure CN119740141B_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of data analysis technology, and in particular to a method, apparatus, equipment and storage medium for classifying iron ore elements. Background Technology
[0002] Traditional methods for classifying iron ore elements involve complex sample collection and pretreatment processes, and rely on specialized laboratory equipment and technicians for manual qualitative and quantitative analysis. These steps are not only costly and time-consuming, but also difficult to implement for rapid on-site testing. Summary of the Invention
[0003] The main purpose of this application is to provide a method, apparatus, equipment and storage medium for classifying iron ore elements, in order to solve the problems of high cost and time consumption in classifying iron ore elements.
[0004] To achieve the above objectives, this application proposes a method for classifying iron ore elements, the method comprising:
[0005] Obtain the first spectral data corresponding to the elements in iron ore;
[0006] Perform data preprocessing on the first spectral data;
[0007] The preprocessed first spectral data is input into the iron ore element classification model to obtain the iron ore element classification result output by the iron ore element classification model. The iron ore element classification model is obtained by feature extraction and model training based on several sets of second spectral data.
[0008] In one embodiment, the training process of the iron ore element classification model includes:
[0009] Acquire several sets of second spectral data;
[0010] The second spectral data are normalized and preprocessed to obtain several sets of third spectral data;
[0011] The third spectral data are input into the initial iron ore element classification model for iterative training to obtain the iron ore element classification model.
[0012] In one embodiment, the initial iron ore element classification model includes a feature extraction layer and a classification layer;
[0013] The step of inputting each of the third spectral data into the initial iron ore element classification model for iterative training to obtain the iron ore element classification model includes:
[0014] The first data feature is extracted from each of the third spectral data through the feature extraction layer;
[0015] The first data feature is input into the classification layer for classification, and the first predicted value of each element category is obtained from the output of the classification layer.
[0016] Based on the first predicted value, the iron ore element classification model is obtained.
[0017] In one embodiment, extracting the first data features from each of the third spectral data through the feature extraction layer includes:
[0018] Each of the third spectral data is input into the first ordinary non-stiffening module and the first special non-stiffening module in the feature extraction layer for convolution operation to obtain the second data features;
[0019] The second data feature is input into the second ordinary stator module and the second special stator module in the feature extraction layer for convolution operation to obtain the first data feature.
[0020] In one embodiment, the step of inputting the first data feature into the classification layer for classification to obtain the first predicted value of each element category output by the classification layer includes:
[0021] The first data feature is input into the fully connected layer in the classification layer for classification to obtain the second predicted value;
[0022] The second predicted value is then subjected to category mapping to obtain the probability distribution value of each element category;
[0023] The probability distribution values are integrated to generate the first predicted value.
[0024] In one embodiment, obtaining the iron ore element classification model based on the first predicted value includes:
[0025] The error value is obtained by calculating the first predicted value and the preset sample label value;
[0026] The error value is compared with the preset expected range;
[0027] If the error value does not meet the preset expected range, the initial iron ore element classification model is iteratively trained based on the error value to obtain the iron ore element classification model.
[0028] In one embodiment, after comparing the error value with a preset expected range, the method further includes:
[0029] If the error value meets the preset expected range, then the initial iron ore element classification model is used as the iron ore element classification model.
[0030] Obtain the prediction accuracy and the actual number of samples corresponding to the iron ore element classification model;
[0031] Based on the prediction accuracy and the actual number of samples, a confusion matrix is generated to optimize the iron ore element classification model.
[0032] Furthermore, to achieve the above objectives, this application also proposes an iron ore element classification device, which includes:
[0033] The acquisition module is used to acquire the first spectral data corresponding to the elements in iron ore;
[0034] The preprocessing module is used to preprocess the first spectral data;
[0035] The output module is used to input the preprocessed first spectral data into the iron ore element classification model to obtain the iron ore element classification result output by the iron ore element classification model. The iron ore element classification model is obtained by feature extraction and model training based on several sets of second spectral data.
[0036] In addition, to achieve the above objectives, this application also proposes an iron ore element classification device, the device comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, the computer program being configured to implement the steps of the iron ore element classification method as described above.
[0037] In addition, to achieve the above objectives, this application also proposes a storage medium, which is a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, it implements the steps of the iron ore element classification method described above.
[0038] In addition, to achieve the above objectives, this application also provides a computer program product, which includes a computer program that, when executed by a processor, implements the steps of the iron ore element classification method described above.
[0039] This application provides a method, apparatus, device, and storage medium for classifying iron ore elements. The method acquires first spectral data corresponding to iron ore elements, preprocesses the first spectral data, and inputs the preprocessed first spectral data into an iron ore element classification model to obtain the iron ore element classification result output by the model. The iron ore element classification model is obtained by feature extraction and model training based on several sets of second spectral data, thereby reducing the cost of iron ore element classification and achieving rapid detection. Attached Figure Description
[0040] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.
[0041] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0042] Figure 1 A flowchart illustrating an embodiment of the iron ore element classification method of this application;
[0043] Figure 2 This is a flowchart illustrating Embodiment 2 of the iron ore element classification method of this application.
[0044] Figure 3 A schematic diagram of the execution logic of the feature extraction layer in one embodiment of the iron ore element classification method provided in this application;
[0045] Figure 4 A simplified flowchart illustrating the iron ore element classification method of this application;
[0046] Figure 5 This is a schematic diagram of the module structure of the iron ore element classification device according to an embodiment of this application;
[0047] Figure 6 This is a schematic diagram of the equipment structure of the hardware operating environment involved in the iron ore element classification method in the embodiments of this application.
[0048] The purpose, features, and advantages of this application will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation
[0049] It should be understood that the specific embodiments described herein are merely illustrative of the technical solutions of this application and are not intended to limit this application.
[0050] To better understand the technical solution of this application, a detailed description will be provided below in conjunction with the accompanying drawings and specific implementation methods.
[0051] It should be noted that the executing entity in this embodiment can be a computing service device with data processing, network communication, and program execution functions, such as a tablet computer, personal computer, or mobile phone, or an electronic device, big data service platform, or iron ore element classification system capable of realizing the above functions. The following description uses an iron ore element classification system as an example to illustrate this embodiment and the subsequent embodiments.
[0052] Based on this, the embodiments of this application provide a method for classifying iron ore elements, referring to... Figure 1 , Figure 1 This is a flowchart illustrating an embodiment of the iron ore element classification method of this application.
[0053] In this embodiment, the iron ore element classification method includes steps S11 to S13:
[0054] Step S11: Obtain the first spectral data corresponding to the elements in the iron ore;
[0055] It should be noted that the iron ore elements refer to the chemical elements contained in iron ore, such as iron (Fe), silicon (Si), aluminum (Al), calcium (Ca), magnesium (Mg), etc. The content and proportion of various elements have an important impact on the quality and use of iron ore. Therefore, it is necessary to analyze the specific elemental data of the iron ore in order to accurately classify the iron ore and put it into the corresponding application field.
[0056] It should be further noted that the first spectral data refers to the near-infrared spectral data obtained directly from the iron ore sample, which reflects the specific absorption and emission characteristics of the elements in the iron ore and is used to identify and classify the elemental composition of the iron ore.
[0057] Step S12: Perform data preprocessing on the first spectral data;
[0058] Specifically, the first spectral data is preprocessed, including normalization, standardization, baseline correction, and endpoint alignment, to improve the accuracy of the model, enhance its performance and accuracy, and reduce computational complexity.
[0059] Step S13: Input the preprocessed first spectral data into the iron ore element classification model to obtain the iron ore element classification result output by the iron ore element classification model. The iron ore element classification model is obtained by feature extraction and model training based on several sets of second spectral data.
[0060] It should be noted that the iron ore element classification model is used to train spectral data obtained from iron ore samples to identify and classify elements in iron ore, thereby extracting features from the input spectral data and mapping them to specific element categories.
[0061] It should be further noted that the iron ore element classification result refers to the prediction result output by the model, that is, the elemental composition information obtained after classifying the iron ore sample to which the input first spectral data belongs, including the iron ore element classification type, content ratio, and quality rating (the quality of iron ore is rated according to the element content and ratio), etc., without any restrictions.
[0062] Furthermore, the second spectral data is a set of spectral data used to train the iron ore element classification model. It is usually derived from the known elemental composition of different iron ore samples and is used to train the model to identify the characteristics and patterns of different elements.
[0063] Specifically, the preprocessed first spectral data is input into the iron ore element classification model to obtain the iron ore element classification result output by the iron ore element classification model, thereby realizing the rapid classification of iron ore on site.
[0064] This embodiment obtains the first spectral data corresponding to the elements in iron ore, then preprocesses the first spectral data, and inputs the preprocessed first spectral data into the iron ore element classification model to obtain the iron ore element classification result output by the iron ore element classification model. The iron ore element classification model is obtained by feature extraction and model training based on several sets of second spectral data, thereby improving the efficiency and accuracy of detection, reducing the cost of manual detection, and enhancing the practicality and environmental friendliness of the model. In addition, through model training, the large model learns the common features of different iron ore samples, enhancing the model's generalization ability on diverse data, thereby improving the accuracy of classification.
[0065] In one feasible implementation, the training process of the iron ore element classification model includes:
[0066] Step S21: Obtain several sets of second spectral data;
[0067] Specifically, several sets of second spectral data are acquired. In one embodiment, each set of second spectral data consists of seven types of near-infrared spectral data, each with 117 spectra. By using multiple sets of second spectral data for training, the model learns the common features among different samples, thereby improving its generalization ability on unknown data.
[0068] Step S22: Normalize and preprocess each of the second spectral data to obtain several sets of third spectral data;
[0069] It should be noted that the third spectral data refers to the spectral data after normalization and preprocessing steps, which can be directly used to train machine learning models to improve the accuracy and efficiency of the model in classifying iron ore elements.
[0070] Specifically, in order to eliminate numerical differences between different samples, each second spectral data is normalized and preprocessed, such as normalization, smoothing, baseline correction, and missing value processing, thereby scaling each second spectral data to a specific range, so that the data values between different spectra can be compared and analyzed on the same scale, thereby improving the accuracy of the model.
[0071] Alternatively, the third spectral data can be divided into a training set and a validation set in an 8:2 ratio, that is, 80% of the data is used to train the model so that it can learn and recognize the features of the spectrum, while the remaining 20% of the data is used to validate the performance of the model to ensure that the model has good generalization ability.
[0072] Step S23: Input each of the third spectral data into the initial iron ore element classification model for iterative training to obtain the iron ore element classification model.
[0073] It should be noted that the initial iron ore element classification model refers to the iron ore element classification model used at the beginning of the training process, which is usually untrained or only preliminarily trained with a small amount of data. In one embodiment, the initial iron ore element classification model is a one-dimensional staggered convolutional network model, wherein the feature extraction module contains 10 staggered network modules, the classification layer consists of 2 fully connected layers, and then by selecting appropriate convolutional kernel size, optimizing activation function and loss function, etc., the spectral features are extracted to the maximum extent and the classification ability is enhanced.
[0074] Specifically, the first data features are extracted from each of the third spectral data by the feature extraction layer, and then the first data features are input to the classification layer for classification to obtain the first predicted value of each element category output by the classification layer. Based on the first predicted value, the iron ore element classification model is obtained.
[0075] This embodiment acquires several sets of second spectral data, then normalizes and preprocesses each set of second spectral data to obtain several sets of third spectral data. These third spectral data are then input into an initial iron ore element classification model for iterative training to obtain the iron ore element classification model. Normalization and other processing accelerate the convergence speed of gradient descent, speeding up model training and enhancing model robustness, thereby improving prediction accuracy. This also helps the model learn more generalized features and reduces overfitting.
[0076] Based on this, the embodiments of this application provide a method for classifying iron ore elements, referring to... Figure 2 , Figure 2 This is a flowchart illustrating Embodiment 2 of the iron ore element classification method of this application.
[0077] In one feasible implementation, the initial iron ore element classification model includes a feature extraction layer and a classification layer; the step of inputting each of the third spectral data into the initial iron ore element classification model for iterative training to obtain the iron ore element classification model includes:
[0078] Step S31: Extract the first data features from each of the third spectral data through the feature extraction layer;
[0079] It should be noted that the feature extraction layer refers to the part of the model responsible for extracting useful features from the input data. It can be implemented through convolutional layers, pooling layers or other types of network layers to identify patterns and features in the input data (such as images, spectral data, etc.).
[0080] It should be further clarified that the classification layer refers to the part of the model responsible for mapping the extracted features to the final output category. In typical classification tasks, the classification layer can consist of a fully connected layer (also known as a dense layer) and an output layer, where the output layer typically uses a softmax function or other activation function to generate the category probability distribution. The first data feature refers to the features extracted from each third spectral data.
[0081] Specifically, the third spectral data are input into the first ordinary non-stress module and the first special non-stress module in the feature extraction layer for convolution operation to obtain the second data feature. Then, the second data feature is input into the second ordinary non-stress module and the second special non-stress module in the feature extraction layer for convolution operation to obtain the first data feature.
[0082] Step S32: Input the first data feature into the classification layer for classification, and obtain the first predicted value of each element category output by the classification layer;
[0083] It should be noted that the element category refers to the different elements or categories that the model needs to identify and classify. These can be different chemical elements or mineral types in iron ore, and are not limited here; they can be set according to the actual situation. The first predicted value refers to the preliminary prediction result given by the model based on the input data and extracted features. In one embodiment, the first predicted value is the category probability or score, representing the model's confidence that each input sample belongs to each category.
[0084] Specifically, the first data feature is input into the fully connected layer in the classification layer for classification to obtain a second predicted value. Then, the second predicted value is mapped to a category to obtain the probability distribution value of each element category. The probability distribution values are then integrated to generate the first predicted value.
[0085] Step S33: Based on the first predicted value, obtain the iron ore element classification model.
[0086] Specifically, the first predicted value and the preset sample label value are calculated to obtain an error value. Then, the error value is compared with a preset expected range. If the error value does not meet the preset expected range, the initial iron ore element classification model is iteratively trained based on the error value to obtain the iron ore element classification model.
[0087] In this embodiment, the first data features are extracted from each of the third spectral data by the feature extraction layer, and then the first data features are input into the classification layer for classification to obtain the first predicted value of each element category output by the classification layer. Based on the first predicted value, the iron ore element classification model is obtained. By extracting key features from the spectral data, the classification accuracy is improved, and the model can more accurately predict the element categories in iron ore, thereby improving the model's generalization ability and robustness. As a result, the model gradually learns the patterns and rules in the data, improving the accuracy of predicting the element categories of iron ore.
[0088] In one feasible implementation, the step of extracting the first data features from each of the third spectral data through the feature extraction layer includes:
[0089] Step S41: Input each of the third spectral data into the first ordinary non-stiffening module and the first special non-stiffening module in the feature extraction layer for convolution operation to obtain the second data features;
[0090] Step S42: Input the second data feature into the second ordinary stator module and the second special stator module in the feature extraction layer for convolution operation to obtain the first data feature.
[0091] It should be noted that the first and second ordinary staggered modules refer to an ordinary staggered module in the feature extraction layer, which is used to perform standard convolution operations. This ordinary staggered module typically contains one or more convolutional layers, as well as a Batch Normalization layer and an activation function (such as ReLU) to extract primary features from the input data (such as spectral data).
[0092] It should be further clarified that the first special staggered module and the second special staggered module refer to modules in the feature extraction layer designed to perform special convolution operations. They have convolution kernel sizes, strides, or structures different from ordinary staggered modules, thereby enabling these modules to capture specific features or patterns that ordinary staggered modules may not be able to capture. The second data feature refers to the intermediate feature representation obtained after processing by the first ordinary staggered module and the first special staggered module.
[0093] Specifically, in one embodiment, the third spectral data is input into the feature extraction layer, which can be referred to... Figure 3 , Figure 3 This is a schematic diagram of the execution logic of the feature extraction layer in one embodiment of the iron ore element classification method provided in this application. The feature extraction layer consists of 8 ordinary non-standard modules and 2 special non-standard modules. The execution logic of each module is as follows: the third spectral data is input into ordinary non-standard module 1 and special non-standard module 1, and output through ordinary non-standard modules 2, 3, 4, 5, 6, 7, and 8. The input of ordinary non-standard module 4 is the output of ordinary non-standard module 3 plus the output of special non-standard module 1. The input of ordinary non-standard module 8 is the output of ordinary non-standard module 7 plus the output of special non-standard module 2. The input of special non-standard module 2 is the output of ordinary non-standard module 4.
[0094] Furthermore, the ordinary staggered modules 1, 3, 5, and 7 consist of two parts: The first part's input data first undergoes a first convolutional operation with a 3×3 kernel and a stride of 3, followed by a Batch Normalization layer to normalize the output, thus accelerating model convergence and reducing overfitting. Then, it passes through a ReLU activation function to effectively alleviate neuron death issues, followed by dropout to further reduce overfitting, and then undergoes a second convolutional operation with a 3×3 kernel and a stride of 1, followed by a second Batch Normalization to ensure output stability. The second part's input data undergoes a convolutional operation with a 1×1 kernel and a stride of 3 to adjust the number of input channels to match the output. This convolutional layer is followed by a Batch Normalization layer. The data is input into these two parts, and their outputs are summed before applying the ReLU activation function.
[0095] Furthermore, the ordinary staggered modules 2, 4, 6, and 8 are as follows: the data first undergoes a first convolution operation with a kernel size of 3×3 and a stride of 1, then goes through a Batch Normalization layer, then through a ReLU activation function, then through a dropout method, then through a second convolution operation with a kernel size of 3×3 and a stride of 1, then undergoes a second Batch Normalization, and then through a ReLU activation function.
[0096] Furthermore, in Special Staggered Module 1: The first part of the input first undergoes a first convolutional operation with a kernel size of 3×3, a stride of 9, 1 input channel, and 32 output channels, followed by a Batch Normalization layer, and then a ReLU activation function. It then undergoes dropout, followed by a second convolutional operation with a kernel size of 3×3, a stride of 1, 32 input channels, and 32 output channels, followed by a second Batch Normalization, and then a ReLU activation function. The second part of the data undergoes a convolutional operation with a kernel size of 1×1, a stride of 3, 1 input channel, and 32 output channels. The data is then processed in two parts, and the two outputs are summed before a ReLU activation function is applied.
[0097] Additionally, in Special Staggered Module 2: The first part of the input undergoes a first convolutional operation with a kernel size of 3×3, a stride of 9, 32 input channels, and 128 output channels. The input then passes through a Batch Normalization layer, followed by a ReLU activation function. Next, it undergoes dropout, followed by a second convolutional operation with a kernel size of 3×3, a stride of 1, and 128 input and output channels, and then a second Batch Normalization. The second part of the input undergoes a convolutional operation with a kernel size of 1×1, a stride of 3, 32 input channels, and 128 output channels. The data is then processed in two parts, and the two output parts are summed before applying a ReLU activation function.
[0098] In this embodiment, the third spectral data is input into the first ordinary stator module and the first special stator module in the feature extraction layer for convolution to obtain the second data feature. Then, the second data feature is input into the second ordinary stator module and the second special stator module in the feature extraction layer for convolution to obtain the first data feature. By using the features of the ordinary stator module and the special stator module with different convolution kernel sizes or strides, the model can capture features at different scales, thereby improving the model's understanding of data and the accuracy of classification, realizing multi-scale feature capture. At the same time, through the stacked convolution of multiple modules, the model can learn deeper feature representations, accelerate the model training process, and improve the convergence speed.
[0099] In one feasible implementation, the step of inputting the first data feature into the classification layer for classification to obtain the first predicted value of each element category output by the classification layer includes:
[0100] Step S51: Input the first data features into the fully connected layer in the classification layer for classification to obtain the second predicted value;
[0101] It should be noted that a fully connected layer (FCL) is a layer in a neural network where each neuron is connected to all neurons in the previous layer. In convolutional neural networks (CNNs), fully connected layers are typically located at the end of the network and are used to map learned high-level features to the final output category. Furthermore, in a fully connected layer, each input feature is considered independently, and the contribution of each input feature to each output node is weighted.
[0102] It should be further noted that the second predicted value refers to the original prediction result directly output by the fully connected layer.
[0103] Specifically, the first data features are input into the classification layer, which uses two fully connected layers and incorporates the dropout method to reduce overfitting, thereby obtaining the second predicted value.
[0104] Step S52: Map the second predicted value to a category to obtain the probability distribution value of each element category;
[0105] It should be noted that the probability distribution values refer to the normalized predicted values, which represent the model's probability estimate of the input data belonging to each category.
[0106] Specifically, the second predicted value is mapped to the category space, for example, the output category is 7 categories. Then, the output of the fully connected layer is normalized using the Softmax function to obtain the probability distribution value of each category. This probability distribution value is directly used as the confidence level. The normalization is usually achieved by the softmax function, which converts the second predicted value (logits) into a probability value so that the sum of the probabilities of all categories is 1.
[0107] Step S53: Integrate the probability distribution values to generate the first predicted value.
[0108] This embodiment classifies the first data features by inputting them into a fully connected layer in the classification layer to obtain a second predicted value. The second predicted value is then mapped to a category to obtain the probability distribution value of each element category. These probability distribution values are then integrated to generate the first predicted value. Furthermore, the fully connected layer learns the complex relationships between features, improving classification accuracy and ensuring that the output values are on the same scale. This contributes to the model's stability and convergence, helping decision-makers understand the model's confidence level for each category and thus make better decisions.
[0109] In one feasible implementation, obtaining the iron ore element classification model based on the first predicted value includes:
[0110] Step S61: Calculate the error value by comparing the first predicted value with the preset sample label value;
[0111] It should be noted that the preset sample label value refers to the true category or value corresponding to each sample during the training and testing of the machine learning model, so that the model's prediction results can be compared with the label value to evaluate its performance. The error value refers to the difference or gap between the model's predicted value and the true label value.
[0112] Specifically, the error value is obtained by calculating the first predicted value and the preset sample label value according to the error value calculation formula. In one embodiment, the error value calculation formula is as follows:
[0113]
[0114] Where H(p, q) is the cross-entropy loss between two probability distributions p and q, where p and q represent the model prediction and the actual label value, respectively, and p(x i ) is the true distribution (i.e., the true probability of each class; typically in classification problems, the probability of the correct class is 1, and the probability of other classes is 0), q(x) i ) is the prediction distribution, which is the probability of each class predicted by the model, and n is the total number of classes.
[0115] Step S62: Compare the error value with the preset expected range;
[0116] It should be noted that the preset expected range refers to an acceptable range or threshold set for error values during model training, used to determine whether the model's performance has reached the expected standard. Specifically, if the error value meets the preset expected range, it indicates that the model is good enough, and training can be stopped or further adjustments made; if the error value does not meet the preset expected range, it indicates that the model needs further training or parameter adjustments.
[0117] Step S63: If the error value does not meet the preset expected range, then based on the error value, the initial iron ore element classification model is iteratively trained to obtain the iron ore element classification model.
[0118] Specifically, if the error value does not conform to the preset expected range, the initial iron ore element classification model is iteratively trained based on the error value to obtain the iron ore element classification model. In one embodiment, when the error value does not conform to the preset expected range, the error value is fed back to the model, and the weights are updated according to the error value. This iterative training process continues until the error value conforms to the preset expected range. The initial iron ore element classification model with the error value conforming to the preset expected range is then used as the iron ore element classification model, thereby enabling the model to gradually learn the patterns and rules in the data and improve the prediction accuracy of solid waste elements. The weight update uses a chain rule, that is, the partial derivative or deviation matrix of the weights is calculated, and the partial derivative value is added to the original weights to obtain a new weight matrix. Furthermore, the optimizer uses an adaptive time estimation method, which can estimate the step size of each parameter based on its historical gradient, thereby adaptively adjusting the learning rate to update the model parameters.
[0119] In addition, after determining the elemental classification model for iron ore, the performance of the model can be evaluated using validation set data. Evaluation metrics include accuracy, recall, and F1 score.
[0120] This embodiment calculates an error value by comparing the first predicted value with the preset sample label value. The error value is then compared with a preset expected range. If the error value does not conform to the preset expected range, the initial iron ore element classification model is iteratively trained based on the error value to obtain the iron ore element classification model. The model is then dynamically adjusted according to the actual error value to adapt to specific characteristics or changes in the data. This continuous iterative training gradually reduces the error between the predicted value and the true label, thereby improving classification accuracy. Simultaneously, it avoids overfitting the model to the training data, ensuring the model has good generalization ability.
[0121] In one feasible implementation, after comparing the error value with a preset expected range, the method further includes:
[0122] Step S71: If the error value meets the preset expected range, then the initial iron ore element classification model is used as the iron ore element classification model.
[0123] Step S72: Obtain the prediction accuracy and actual number of samples corresponding to the iron ore element classification model;
[0124] It should be noted that the prediction accuracy refers to the proportion of samples correctly predicted by the model out of the total number of samples. The actual number of samples refers to the number of samples actually used in model training.
[0125] Step S73: Based on the prediction accuracy and the actual number of samples, generate a confusion matrix to optimize the iron ore element classification model.
[0126] It should be noted that the confusion matrix is used to show the relationship between the model's prediction results and the actual labels. It displays the correct and incorrect predictions for each category in matrix form. Each cell of the confusion matrix represents a combination of the actual category and the predicted category, such as true positive (TP), false positive (FP), true negative (TN), and false negative (FN). This helps to identify which categories the model performs well on, which categories need improvement, and whether the model has specific biases.
[0127] Specifically, a confusion matrix is created by combining the accuracy and the number of actual samples on the validation set. In one embodiment, the X-axis of the confusion matrix represents the actual class, the Y-axis represents the predicted class, and the diagonal represents the correctly predicted class. This allows the model's performance to be visualized by plotting the confusion matrix, showing the model's prediction accuracy for each iron ore classification.
[0128] This embodiment uses the initial iron ore element classification model as the iron ore element classification model if the error value meets the preset expected range. It then obtains the prediction accuracy and actual sample number corresponding to the iron ore element classification model, and generates a confusion matrix based on the prediction accuracy and actual sample number to optimize the iron ore element classification model. This ensures that the model's performance meets the expected standard. The confusion matrix visually displays the model's performance in each category, including the number of correct and incorrect classifications, helping to identify which categories the model performs poorly in and providing direction for model optimization. Furthermore, if the model is found to perform poorly in certain categories, it can guide the collection of more data in those categories to improve the model.
[0129] For example, to help understand the implementation process of the iron ore element classification method, please refer to... Figure 4 , Figure 4 A simplified flowchart illustrating the iron ore element classification method provided in this application.
[0130] It should be noted that the examples in the figure are only for understanding this application and do not constitute a limitation on the iron ore element classification method of this application. Any simple transformations based on this technical concept are within the protection scope of this application.
[0131] It should be understood that the sequence number of each step in the above embodiments does not imply 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 embodiments of the present invention.
[0132] This application also provides an iron ore element classification device, please refer to... Figure 5 The iron ore element classification device includes:
[0133] The acquisition module 51 is used to acquire the first spectral data corresponding to the elements in iron ore;
[0134] Preprocessing module 52 is used to preprocess the first spectral data;
[0135] The output module 53 is used to input the preprocessed first spectral data into the iron ore element classification model to obtain the iron ore element classification result output by the iron ore element classification model, wherein the iron ore element classification model is obtained by feature extraction and model training based on several sets of second spectral data.
[0136] The iron ore element sorting device is also used for:
[0137] Acquire several sets of second spectral data;
[0138] The second spectral data are normalized and preprocessed to obtain several sets of third spectral data;
[0139] The third spectral data are input into the initial iron ore element classification model for iterative training to obtain the iron ore element classification model.
[0140] The iron ore element sorting device is also used for:
[0141] The first data feature is extracted from each of the third spectral data through the feature extraction layer;
[0142] The first data feature is input into the classification layer for classification, and the first predicted value of each element category is obtained from the output of the classification layer.
[0143] Based on the first predicted value, the iron ore element classification model is obtained.
[0144] The iron ore element sorting device is also used for:
[0145] Each of the third spectral data is input into the first ordinary non-stiffening module and the first special non-stiffening module in the feature extraction layer for convolution operation to obtain the second data features;
[0146] The second data feature is input into the second ordinary stator module and the second special stator module in the feature extraction layer for convolution operation to obtain the first data feature.
[0147] The iron ore element sorting device is also used for:
[0148] The first data feature is input into the fully connected layer in the classification layer for classification to obtain the second predicted value;
[0149] The second predicted value is then subjected to category mapping to obtain the probability distribution value of each element category;
[0150] The probability distribution values are integrated to generate the first predicted value.
[0151] The iron ore element sorting device is also used for:
[0152] The error value is obtained by calculating the first predicted value and the preset sample label value;
[0153] The error value is compared with the preset expected range;
[0154] If the error value does not meet the preset expected range, the initial iron ore element classification model is iteratively trained based on the error value to obtain the iron ore element classification model.
[0155] The iron ore element sorting device is also used for:
[0156] If the error value meets the preset expected range, then the initial iron ore element classification model is used as the iron ore element classification model.
[0157] Obtain the prediction accuracy and the actual number of samples corresponding to the iron ore element classification model;
[0158] Based on the prediction accuracy and the actual number of samples, a confusion matrix is generated to optimize the iron ore element classification model.
[0159] The iron ore element classification device provided in this application, employing the iron ore element classification method in the above embodiments, can solve the technical problems mentioned in the background art. Compared with the prior art, the beneficial effects of the iron ore element classification device provided in this application are the same as those of the iron ore element classification method provided in the above embodiments, and other technical features in the iron ore element classification device are the same as those disclosed in the methods of the above embodiments, and will not be repeated here.
[0160] This application provides an iron ore element classification device, which includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, which are executed by the at least one processor to enable the at least one processor to perform the iron ore element classification method in the first embodiment described above.
[0161] The following is for reference. Figure 6The diagram illustrates a structural schematic suitable for implementing an iron ore element classification device according to embodiments of this application. The iron ore element classification device in these embodiments may include, but is not limited to, mobile terminals such as mobile phones, laptops, digital radio receivers, PDAs (Personal Digital Assistants), PADs (Portable Application Description), PMPs (Portable Media Players), and in-vehicle terminals (e.g., in-vehicle navigation terminals), as well as fixed terminals such as digital TVs and desktop computers. Figure 6 The iron ore element sorting device shown is merely an example and should not impose any limitations on the functionality and scope of use of the embodiments of this application.
[0162] like Figure 6 As shown, the iron ore element sorting device may include a processing unit 1001 (e.g., a central processing unit, a graphics processing unit, etc.), which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 1002 or a program loaded from a storage device 1003 into a random access memory (RAM) 1004. The RAM 1004 also stores various programs and data required for the operation of the iron ore element sorting device. The processing unit 1001, ROM 1002, and RAM 1004 are interconnected via a bus 1005. An input / output (I / O) interface 1006 is also connected to the bus. Typically, the following systems can be connected to I / O interface 1006: input devices 1007 including, for example, touchscreens, touchpads, keyboards, mice, image sensors, microphones, accelerometers, gyroscopes, etc.; output devices 1008 including, for example, liquid crystal displays (LCDs), speakers, vibrators, etc.; storage devices 1003 including, for example, magnetic tapes, hard disks, etc.; and communication devices 1009. Communication device 1009 allows the iron ore element sorting equipment to communicate wirelessly or wiredly with other devices to exchange data. Although iron ore element sorting equipment with various systems is shown in the figure, it should be understood that it is not required to implement or possess all the systems shown. More or fewer systems can be implemented alternatively.
[0163] Specifically, according to the embodiments disclosed in this application, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments disclosed in this application include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via a communication device, or installed from storage device 1003, or installed from ROM 1002. When the computer program is executed by processing device 1001, it performs the functions defined in the methods of the embodiments disclosed in this application.
[0164] The iron ore element classification device provided in this application, employing the iron ore element classification method in the above embodiments, can solve the technical problems mentioned in the background art. Compared with the prior art, the beneficial effects of the iron ore element classification device provided in this application are the same as those of the iron ore element classification method provided in the above embodiments, and other technical features of this iron ore element classification device are the same as those disclosed in the previous embodiment method, and will not be repeated here.
[0165] It should be understood that the various parts disclosed in this application can be implemented using hardware, software, firmware, or a combination thereof. In the description of the above embodiments, specific features, structures, materials, or characteristics can be combined in any suitable manner in one or more embodiments or examples.
[0166] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.
[0167] This application provides a computer-readable storage medium having computer-readable program instructions (i.e., a computer program) stored thereon, the computer-readable program instructions being used to execute the iron ore element classification method in the above embodiments.
[0168] The computer-readable storage medium provided in this application may be, for example, a USB flash drive, but is not limited to, electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices, or any combination thereof. More specific examples of computer-readable storage media may include, but are not limited to: electrical connections having one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof. In this embodiment, the computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, system, or device. The program code contained on the computer-readable storage medium may be transmitted using any suitable medium, including but not limited to: wires, optical cables, RF (Radio Frequency), etc., or any suitable combination thereof.
[0169] The aforementioned computer-readable storage medium may be included in the iron ore element sorting device; or it may exist independently and not assembled into the iron ore element sorting device.
[0170] The aforementioned computer-readable storage medium carries one or more programs, which, when executed by the iron ore element classification device, cause the iron ore element classification device to:
[0171] Obtain the first spectral data corresponding to the elements in iron ore;
[0172] Perform data preprocessing on the first spectral data;
[0173] The preprocessed first spectral data is input into the iron ore element classification model to obtain the iron ore element classification result output by the iron ore element classification model. The iron ore element classification model is obtained by feature extraction and model training based on several sets of second spectral data.
[0174] Computer program code for performing the operations of this application can be written in one or more programming languages or a combination thereof, including object-oriented programming languages such as Java, Smalltalk, and C++, and conventional procedural programming languages such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including a Local Area Network (LAN) or a Wide Area Network (WAN)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).
[0175] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this application. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.
[0176] The modules described in the embodiments of this application can be implemented in software or hardware. The names of the modules do not necessarily limit the functionality of the unit itself.
[0177] The readable storage medium provided in this application is a computer-readable storage medium that stores computer-readable program instructions (i.e., a computer program) for executing the above-described iron ore element classification method, and is capable of solving the technical problems described in the background art. Compared with the prior art, the beneficial effects of the computer-readable storage medium provided in this application are the same as the beneficial effects of the iron ore element classification method provided in the above embodiments, and will not be repeated here.
[0178] This application provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the iron ore element classification method described above.
[0179] The computer program product provided in this application can solve the technical problems described in the background section. Compared with the prior art, the beneficial effects of the computer program product provided in the embodiments of this application are the same as the beneficial effects of the iron ore element classification method provided in the above embodiments, and will not be repeated here.
[0180] The above description is only a part of the embodiments of this application and does not limit the patent scope of this application. All equivalent structural transformations made under the technical concept of this application and using the contents of the specification and drawings of this application, or direct / indirect applications in other related technical fields, are included in the patent protection scope of this application.
Claims
1. A method for classifying iron ore elements, characterized in that, include: Obtain the first spectral data corresponding to the elements in iron ore; The first spectral data is preprocessed, including normalization, standardization, baseline correction, and endpoint alignment. The preprocessed first spectral data is input into the iron ore element classification model to obtain the iron ore element classification result output by the iron ore element classification model. The iron ore element classification model is obtained by feature extraction and model training based on several sets of second spectral data. The training process of the iron ore element classification model includes: Acquire several sets of second spectral data; normalize and preprocess each set of second spectral data to obtain several sets of third spectral data; input each set of third spectral data into an initial iron ore element classification model for iterative training to obtain the iron ore element classification model; the initial iron ore element classification model is a one-dimensional stochastic convolutional network model, which includes a feature extraction layer and a classification layer. The feature extraction layer consists of 8 ordinary stochastic modules and 2 special stochastic modules. The execution logic of each module is as follows: each set of third spectral data is input into ordinary stochastic module 1 and special stochastic module 1, and outputs through ordinary stochastic module 2, ordinary stochastic module 3, ordinary stochastic module 4, ordinary stochastic module 5, ordinary stochastic module 6, ordinary stochastic module 7, and ordinary stochastic module 8. The input of ordinary stochastic module 4 is the output of ordinary stochastic module 3 plus the output of special stochastic module 1, the input of ordinary stochastic module 8 is the output of ordinary stochastic module 7 plus the output of special stochastic module 2, and the input of special stochastic module 2 is the output of ordinary stochastic module 4. The ordinary stochastic module 1, ordinary stochastic module 3, ordinary stochastic module 5, and ordinary stochastic module 7 consist of two parts: The first part of the input data first passes through a convolutional layer with a kernel size of 3×3 and a stride of 3, then through a BatchNormalization layer, and then through a ReLU activation function; after dropping out to reduce overfitting, it passes through a second convolutional layer with a kernel size of 3×3 and a stride of 1, and then undergoes a second Batch Normalization; The second part of the input data passes through a convolutional layer with a kernel size of 1×1 and a stride of 3, followed by a BatchNormalization layer. The data is input into these two parts, and the outputs of these two parts are added together, and then the ReLU activation function is applied. Ordinary staggered modules 2, 4, 6, and 8 work as follows: The data first goes through a first convolutional operation with a kernel size of 3×3 and a stride of 1, then through a Batch Normalization layer, then through a ReLU activation function, then through the dropout method, then through a second convolutional operation with a kernel size of 3×3 and a stride of 1, then through a second Batch Normalization, and then through a ReLU activation function. Special Staggered Module 1: The first part of the input first goes through a first convolutional layer with a kernel size of 3×3, a stride of 9, 1 input channel, and 32 output channels. Then it goes through a Batch Normalization layer, followed by a ReLU activation function. Then it goes through a dropout method, then through a second convolutional layer with a kernel size of 3×3, a stride of 1, 32 input channels, and 32 output channels. Then it goes through a second Batch Normalization layer, followed by a ReLU activation function. The second part of the data goes through a convolutional layer with a kernel size of 1×1, a stride of 3, 1 input channel, and 32 output channels. The data is processed in two parts, the two outputs are added together, and then a ReLU activation function is applied. Special Staggered Module 2: The first part first undergoes a first convolutional operation with a kernel size of 3×3, a stride of 9, 32 input channels, and 128 output channels. The input then passes through a Batch Normalization layer, followed by a ReLU activation function. After that, it goes through a dropout method, then a second convolutional operation with a kernel size of 3×3, a stride of 1, 128 input channels, and 128 output channels, followed by a second Batch Normalization. The second part of the input passes through a convolutional operation with a kernel size of 1×1, a stride of 3, 32 input channels, and 128 output channels. The data is then processed in two parts, and the two output parts are added together before applying a ReLU activation function. Specifically, the first data features are extracted from each of the third spectral data by the feature extraction layer; the first data features are input to the classification layer for classification to obtain the first predicted value of each element category output by the classification layer; and the iron ore element classification model is obtained based on the first predicted value.
2. The iron ore element classification method as described in claim 1, characterized in that, The step of extracting the first data features from each of the third spectral data through the feature extraction layer includes: Each of the third spectral data is input into the first ordinary non-stiffening module and the first special non-stiffening module in the feature extraction layer for convolution operation to obtain the second data features; The second data feature is input into the second ordinary stator module and the second special stator module in the feature extraction layer for convolution operation to obtain the first data feature.
3. The iron ore element classification method as described in claim 1, characterized in that, The step of inputting the first data features into the classification layer for classification, and obtaining the first predicted value of each element category output by the classification layer, includes: The first data feature is input into the fully connected layer in the classification layer for classification to obtain the second predicted value; The second predicted value is then subjected to category mapping to obtain the probability distribution value of each element category; The probability distribution values are integrated to generate the first predicted value.
4. The iron ore element classification method as described in claim 1, characterized in that, The step of obtaining the iron ore element classification model based on the first predicted value includes: The error value is obtained by calculating the first predicted value and the preset sample label value; The error value is compared with the preset expected range; If the error value does not meet the preset expected range, the initial iron ore element classification model is iteratively trained based on the error value to obtain the iron ore element classification model.
5. The iron ore element classification method as described in claim 4, characterized in that, After comparing the error value with the preset expected range, the method further includes: If the error value meets the preset expected range, then the initial iron ore element classification model is used as the iron ore element classification model. Obtain the prediction accuracy and the actual number of samples corresponding to the iron ore element classification model; Based on the prediction accuracy and the actual number of samples, a confusion matrix is generated to optimize the iron ore element classification model.
6. An iron ore element classification device, characterized in that, include: The acquisition module is used to acquire the first spectral data corresponding to the elements in iron ore; The preprocessing module is used to perform data preprocessing on the first spectral data. The data preprocessing includes normalization, standardization, baseline correction, and endpoint alignment. The output module is used to input the preprocessed first spectral data into the iron ore element classification model to obtain the iron ore element classification result output by the iron ore element classification model, wherein the iron ore element classification model is obtained by feature extraction and model training based on several sets of second spectral data. The model training module is used to acquire several sets of second spectral data; normalize and preprocess each set of second spectral data to obtain several sets of third spectral data; input each set of third spectral data into an initial iron ore element classification model for iterative training to obtain the iron ore element classification model; the initial iron ore element classification model is a one-dimensional staggered convolutional network model, which includes a feature extraction layer and a classification layer. The feature extraction layer consists of 8 ordinary staggered modules and 2 special staggered modules. The execution logic of each module is as follows: each set of third spectral data is input into ordinary staggered module 1 and special staggered module 1, and then processed through ordinary staggered module 2 and ordinary staggered module 3. The outputs are generated by modules 3, 4, 5, 6, 7, and 8. The input to module 4 is the sum of the output of module 3 and module 1 (specialized), the input to module 8 is the sum of the output of module 7 and module 2 (specialized), and the input to module 2 is the output of module 4 (specialized). Modules 1, 3, 5, and 7 consist of two parts: the first part of the input data is processed through a 3×3 convolutional kernel with a stride of 3 (first convolutional layer), and then processed through a batch... The data undergoes a normalization layer, followed by a ReLU activation function; then dropout is applied to reduce overfitting, followed by a second convolutional layer with a 3×3 kernel and a stride of 1, and then a second batch normalization. The second part of the input data undergoes a convolutional layer with a 1×1 kernel and a stride of 3, followed by a batch normalization layer. The data is input into both parts, and the outputs of the two parts are summed before applying a ReLU activation function. The normalization modules 2, 4, 6, and 8 are as follows: the data first undergoes a first convolutional layer with a 3×3 kernel and a stride of 1, then a batch normalization layer, followed by a ReLU activation function, then dropout, then a second convolutional layer with a 3×3 kernel and a stride of 1, and then a second batch normalization. Normalization is performed, followed by ReLU activation. Special staggered module 1: The first part of the input first goes through a first convolutional layer with a kernel size of 3×3, a stride of 9, 1 input channel, and 32 output channels, then goes through a Batch Normalization layer, and then through the ReLU activation function.After dropout, the data undergoes a second convolutional layer with a 3×3 kernel, a stride of 1, 32 input channels, and 32 output channels. This is followed by a second batch normalization layer, then ReLU activation. The second part of the data then undergoes a convolutional layer with a 1×1 kernel, a stride of 3, 1 input channel, and 32 output channels. The data is then processed in two parts, and the two outputs are summed before ReLU activation. Special staggered module 2: The first part of the data undergoes a first convolutional layer with a 3×3 kernel, a stride of 9, 32 input channels, and 128 output channels. The input then undergoes a batch normalization layer, followed by ReLU activation. This is then droppedout, followed by a second convolutional layer with a 3×3 kernel, a stride of 1, and 128 input and output channels. This is followed by a second batch normalization layer, then ReLU activation. Normalization; the second part of the input undergoes a convolution operation with a kernel size of 1×1, a stride of 3, 32 input channels, and 128 output channels. The data is then processed in two parts, and the two output parts are summed before applying the ReLU activation function. Specifically, the feature extraction layer extracts the first data features from each of the third spectral data. These first data features are then input to the classification layer for classification, yielding the first predicted value for each element category output by the classification layer. Based on these first predicted values, the iron ore element classification model is obtained.
7. An iron ore element classification device, characterized in that, The iron ore element classification device includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, the computer program being configured to implement the steps of the iron ore element classification method as described in any one of claims 1 to 5.
8. A storage medium, characterized in that, The storage medium is a computer-readable storage medium, and a computer program is stored on the storage medium. When the computer program is executed by a processor, it implements the steps of the iron ore element classification method as described in any one of claims 1 to 5.