A method, apparatus, equipment and medium for detecting the tail section of a sintering machine.
By training the sample set using a recurrent neural network module, the problem of low efficiency in observing the tail section of the sintering machine under high temperature and high dust conditions was solved, and the optimal tail section was automatically detected, which improved production efficiency and saved costs.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- ZHONGYE-CHANGTIAN INT ENG CO LTD
- Filing Date
- 2022-05-12
- Publication Date
- 2026-05-19
AI Technical Summary
Observing the tail section of the sintering machine in a high-temperature and high-dust environment is inefficient and makes it difficult to accurately detect the optimal tail section, which affects production efficiency and cost.
A recurrent neural network module is used to train the sample set. Through segmentation, feature extraction, dense connection and fusion vector processing, the optimal tail section is detected by a classifier.
It enables automatic detection of the optimal tail section in non-high temperature environments, improving production efficiency and saving costs.
Smart Images
Figure CN117115067B_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of sintering machine tail section detection, and in particular to a method, apparatus, equipment and medium for sintering machine tail section detection. Background Technology
[0002] The sintering machine is a key piece of equipment in the sintering process of smelting ore. The sintering machine trolleys are linked end-to-end to form a trolley chain. Raw materials, pre-loaded at the machine head, are loaded onto the trolleys, and then ignited on the surface of the loaded fuel. As the trolleys move forward, the ignited raw materials gradually burn downwards from the surface. At the machine tail, the sinter is unloaded onto the annular cooling trolley at the bottom of the sintering machine, and then the sintering trolley is reversed and runs back to the machine head position. The tail section is the cross-sectional image of the sinter when the trolley is overturned and unloaded. The tail section is a crucial state image connecting the two complex physicochemical processes of sintering and cooling, containing rich implicit information about the thermal state and the quality and yield characteristics of the sinter. An optimal tail section image can help the operator observe the current sintering conditions to a certain extent. For example, the brightness of the red-hot layer and porosity can indicate the fuel consumption; the presence of raw material and smoke can indicate moisture content; and the height of the red-hot layer can indicate whether the material is fully burned, etc.
[0003] Currently, in actual industrial production, the observation point for the tail section is located at the tail of the sintering machine. This location is typically in a high-temperature, high-dust environment, and the optimal tail section is often only revealed at the moment the trolley flips. This requires the inspector to be constantly at the tail of the sintering machine to observe the state of the tail section, which is time-consuming, labor-intensive, and inefficient. Summary of the Invention
[0004] In view of this, the purpose of this application is to provide a method, apparatus, equipment, and medium for detecting the tail section of a sintering machine, which can accurately detect the optimal tail section, improve production efficiency, and save production costs. The specific solution is as follows:
[0005] In a first aspect, this application discloses a method for detecting the cross-section of a sintering machine tail section, including:
[0006] Obtain a training sample set; wherein the training sample set includes optimal sintering machine tail section samples, non-optimal sintering machine tail section samples, and label information;
[0007] The preset initial model is trained using the training sample set to obtain the sintering machine tail section detection model; wherein, the sintering machine tail section detection model includes a recurrent neural network module;
[0008] Once the tail section of the sintering machine to be inspected is obtained, the optimal cross-section inspection result corresponding to the tail section of the sintering machine to be inspected is output using the sintering machine tail section inspection model.
[0009] Optionally, the step of using the sintering machine tail section detection model to output the optimal section detection result corresponding to the sintering machine tail section to be detected includes:
[0010] The tail section of the sintering machine to be tested is segmented to obtain multiple sets of image data sequences;
[0011] The multiple sets of image data sequences are respectively input into multiple recurrent neural network modules in the sintering machine tail section detection model to obtain multiple feature maps;
[0012] The multiple feature maps are densely connected to obtain multiple feature vectors;
[0013] The multiple feature vectors are fused to obtain a fused vector;
[0014] The fused vector is input into the classifier to obtain the optimal cross-section detection result corresponding to the cross-section of the sintering machine tail to be detected.
[0015] Optionally, the sintering machine tail section detection model further includes multiple densely connected processing modules. The number of these densely connected processing modules is the same as the number of the recurrent neural network modules, and they are arranged sequentially. Correspondingly, the dense connection of the multiple feature maps to obtain multiple feature vectors includes:
[0016] The feature vector corresponding to the dense connection processing module is obtained by adding pixels of the feature map input to the dense connection processing module through any of the dense connection processing modules.
[0017] The feature map input to any of the dense connection processing modules includes feature maps output by recurrent neural network modules that have the same order as the dense connection processing module and those that have the same order as the preceding dense connection processing modules.
[0018] Optionally, fusing the multiple feature vectors to obtain a fused vector includes:
[0019] The multiple feature vectors are each subjected to fully connected processing to obtain multiple fully connected processed vectors;
[0020] The multiple fully connected vectors are then subjected to full-connection processing to obtain a fused vector.
[0021] Optionally, the step of performing full-connection processing on the plurality of fully connected vectors to obtain a fused vector includes:
[0022] Multiple weight parameters and multiple fully connected processed vectors are multiplied together to obtain multiple multiplied vectors.
[0023] The vector obtained by multiplying the multiple weights is fully connected to obtain the fused vector.
[0024] Optionally, the step of segmenting the sintering machine tail section to be detected to obtain multiple sets of image data sequences includes:
[0025] Determine the segmentation direction corresponding to the tail section of the sintering machine to be tested;
[0026] The cross-section of the sintering machine tail section to be detected is segmented based on the segmentation direction to obtain multiple sets of image data sequences.
[0027] Optionally, the step of inputting the multiple sets of image data sequences into multiple recurrent neural network modules in the sintering machine tail section detection model to obtain multiple feature maps includes:
[0028] If any set of image data sequences includes multiple image data sequences, then the set of image data sequences is processed into a one-dimensional image data sequence;
[0029] Multiple one-dimensional image data sequences are respectively input into multiple recurrent neural network modules in the sintering machine tail section detection model to obtain multiple feature maps.
[0030] Secondly, this application discloses a sintering machine tail section detection device, comprising:
[0031] The sample acquisition module is used to acquire a training sample set; wherein, the training sample set includes optimal sintering machine tail section samples, non-optimal sintering machine tail section samples, and label information.
[0032] The model training module is used to train a preset initial model using the training sample set to obtain a sintering machine tail section detection model; wherein, the sintering machine tail section detection model includes a recurrent neural network module.
[0033] The cross-section detection module is used to output the optimal cross-section detection result corresponding to the cross-section of the sintering machine tail section when the cross-section of the sintering machine tail section to be detected is obtained, using the cross-section detection model of the sintering machine tail section to be detected.
[0034] Thirdly, this application discloses an electronic device, including a processor and a memory; wherein,
[0035] The memory is used to store computer programs;
[0036] The processor is used to execute the computer program to implement the aforementioned sintering machine tail section detection method.
[0037] Fourthly, this application discloses a computer-readable storage medium for storing a computer program, wherein the computer program, when executed by a processor, implements the aforementioned sintering machine tail section detection method.
[0038] As can be seen, this application first obtains a training sample set, which includes optimal sintering machine tail section samples, non-optimal sintering machine tail section samples, and label information. Then, the training sample set is used to train a preset initial model to obtain a sintering machine tail section detection model. This model includes a recurrent neural network module. Finally, when a sintering machine tail section to be detected is obtained, the model outputs the optimal section detection result corresponding to that section. In other words, this application first uses a training sample set including optimal and non-optimal sintering machine tail section samples to train the model, obtaining a sintering machine tail section detection model. Then, the trained model is used to perform optimal section detection. The model includes a recurrent neural network module, which can extract deep hidden information from the sintering machine tail section. This allows for accurate detection of the optimal tail section, improving production efficiency and saving production costs. Attached Figure Description
[0039] 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, the drawings described below are only embodiments of this application. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.
[0040] Figure 1 This application provides a flowchart of a method for detecting the tail section of a sintering machine;
[0041] Figure 2 A flowchart of a specific method for detecting the tail section of a sintering machine provided in this application;
[0042] Figure 3 A specific LSTM module schematic diagram is provided for this application;
[0043] Figure 4 This application provides a schematic diagram of the structure of a sintering machine tail section detection device;
[0044] Figure 5 This application provides a structural diagram of an electronic device. Detailed Implementation
[0045] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.
[0046] Currently, in actual industrial production, the observation point for the tail section is located at the tail of the sintering machine, which is typically in a high-temperature, high-dust environment. The optimal tail section is often only revealed in the instant the trolley flips over. This requires the inspector to be constantly at the tail of the sintering machine to observe the tail section, which is time-consuming, labor-intensive, and inefficient. Therefore, this application provides a tail section detection solution for sintering machines that can accurately detect the optimal tail section, improve production efficiency, and save production costs.
[0047] See Figure 1 As shown in the figure, this application discloses a method for detecting the tail section of a sintering machine, including:
[0048] Step S11: Obtain the training sample set; wherein the training sample set includes the optimal sintering machine tail section sample, the non-optimal sintering machine tail section sample, and label information.
[0049] In specific implementations, the optimal sintering machine tail section sample and the non-optimal sintering machine tail section sample can be images of the following types: visible light color image, visible light black and white image, infrared thermal image, etc.
[0050] Step S12: Train the preset initial model using the training sample set to obtain the sintering machine tail section detection model; wherein the sintering machine tail section detection model includes a recurrent neural network module.
[0051] In one implementation, the recurrent neural network module can be an LSTM (Long Short-Term Memory) module.
[0052] Step S13: When the tail section of the sintering machine to be tested is obtained, the optimal cross-section detection result corresponding to the tail section of the sintering machine to be tested is output using the tail section detection model of the sintering machine.
[0053] In a specific implementation, a visible light camera or an infrared camera installed at the tail of the sintering machine can be used to continuously collect images of the tail section to obtain the tail section of the sintering machine to be tested. This allows for the automatic and real-time determination of the optimal tail section, facilitating researchers to analyze the sintering conditions.
[0054] In a specific implementation, step S13 includes the following steps:
[0055] Step 00: Divide the cross-section of the sintering machine tail section to be tested to obtain multiple sets of image data sequences.
[0056] In a specific implementation, the segmentation direction corresponding to the tail section of the sintering machine to be detected can be determined; the tail section of the sintering machine to be detected is segmented based on the segmentation direction to obtain multiple sets of image data sequences.
[0057] The segmentation direction can be either horizontal or vertical. In this specific implementation, the segmentation direction is determined to be the same as the one used when segmenting samples during training. It should be noted that since the formation and feature distribution of the tail section are mostly horizontal, the segmentation direction can be determined to be horizontal.
[0058] Furthermore, the image is segmented based on the parameters of the sintering machine tail section to be inspected. For example, if the sintering machine tail section to be inspected is a 1000×600 visible light color image, in one embodiment, it is horizontally segmented into 600 sets of image data sequences, each set including one image data sequence. In another embodiment, it is horizontally segmented into 300 sets of image data sequences, each set including two image data sequences; that is, one image data sequence includes one row of original pixels from the sintering machine tail section to be inspected.
[0059] Step 01: Input the multiple sets of image data sequences into the multiple recurrent neural network modules in the sintering machine tail section detection model to obtain multiple feature maps.
[0060] If any set of image data sequences includes multiple image data sequences, then the set of image data sequences is processed into a one-dimensional image data sequence; the multiple one-dimensional image data sequences are respectively input into the multiple recurrent neural network modules in the sintering machine tail section detection model to obtain multiple feature maps.
[0061] That is, in this application, each set of image data sequences corresponds to a recurrent neural network module, which ultimately corresponds to a feature map.
[0062] Step 02: Perform dense connections on the multiple feature maps to obtain multiple feature vectors.
[0063] The sintering machine tail section detection model further includes multiple densely connected processing modules. The number of these densely connected processing modules is the same as that of the recurrent neural network modules, and they are arranged sequentially. Accordingly, the specific steps for densely connecting the multiple feature maps to obtain multiple feature vectors are as follows: the feature map input to any of the densely connected processing modules is pixel-wise summed to obtain the feature vector corresponding to that densely connected processing module. The feature map input to any of the densely connected processing modules includes feature maps output by recurrent neural network modules that are in the same order as the current densely connected processing module and those that are in the same order as the preceding densely connected processing modules.
[0064] That is, the number of image data sequence groups, the number of dense connection processing modules, and the number of recurrent neural network modules are the same. Furthermore, the image data sequence groups, dense connection processing modules, and recurrent neural network modules can be labeled with sequence numbers, for example, from 1 to 600. The preceding dense connection processing modules include all dense connection processing modules with sequence numbers lower than that module.
[0065] It should be noted that dense connections can enrich the variety and quantity of features and prevent the loss of important data. Existing DenseNet networks use channel concatenation operations, which makes the feature maps larger as the network deepens. This application uses pixel addition for dense connections, which does not change the size of the feature maps as the network deepens, thus maintaining speed.
[0066] Step 03: Fuse the multiple feature vectors to obtain a fused vector.
[0067] In a specific implementation, the multiple feature vectors can be processed by full connection to obtain multiple fully connected vectors; the multiple fully connected vectors can then be processed by full connection to obtain a fused vector.
[0068] Furthermore, a weight multiplication operation is performed on multiple weight parameters and multiple fully connected processed vectors to obtain multiple weighted multiplied vectors; these multiple weighted multiplied vectors are then subjected to fully connected processing to obtain a fused vector. The weight parameters are the parameters obtained during training.
[0069] Step 04: Input the fused vector into the classifier to obtain the optimal cross-section detection result corresponding to the cross-section of the sintering machine tail to be detected.
[0070] The classifier can be a softmax classifier, and the optimal cross-section detection result is the optimal tail section and the non-optimal tail section.
[0071] It should be noted that the solution provided in this application can support images of any size. Specifically, an initial model can be constructed and trained based on the corresponding segmentation results. The training process can be referenced from the detection process and will not be elaborated here. Below, the solution provided in this application is further illustrated using a 1000×600 visible light color image as an example. For example, see... Figure 2 As shown, Figure 2 This is a flowchart of a specific sintering machine tail section detection method disclosed in an embodiment of this application.
[0072] First, the image is segmented, dividing the two-dimensional image data into one-dimensional array data to facilitate feature extraction by the subsequent LSTM module. The input image is a 1000×600 visible light color image, meaning its two-dimensional data structure should be:
[0073] [(R,G,B) 0,0 (R,G,B) 0,1 ,……,……,……,(R,G,B) 0,998 (R,G,B) 0,999 ]
[0074] [(R,G,B) 1,0 (R,G,B) 1,1 ,……,……,……,(R,G,B) 1,998 (R,G,B) 1,999 ]
[0075] ...
[0076] ...
[0077] [(R,G,B) 598,0 (R,G,B) 598,1 ,……,……,(R,G,B) 598,998 (R,G,B) 598,999 ]
[0078] [(R,G,B) 599,0 (R,G,B) 599,1 ,……,……,(R,G,B) 599,998 (R,G,B) 599,999 ]
[0079] In this context, R, G, and B represent the color components of each pixel in a color image.
[0080] Since the formation and characteristic distribution of the tail section are mostly transverse, transverse segmentation was selected for sequence segmentation. The segmentation result is as follows:
[0081] [A0] = [(R,G,B)] 0,0 (R,G,B)0,1 ,…,…,…,(R,G,B) 0,998 (R,G,B) 0,999 ]
[0082] [A1] = [(R,G,B)] 1,0 (R,G,B) 1,1 ,…,…,…,(R,G,B) 1,998 (R,G,B) 1,999 ]
[0083] ...
[0084] ...
[0085] [A 598 ] = [(R,G,B) 598,0 (R,G,B) 598,1 ,…,…,(R,G,B) 598,998 (R,G,B) 598,999 ]
[0086] [A 599 ] = [(R,G,B) 599,0 (R,G,B) 599,1 ,…,…,(R,G,B) 599,998 (R,G,B) 599,999 ]
[0087] A total of 600 sequences were collected, each consisting of an image data sequence, which includes a row of original pixels from the tail section of the sintering machine to be inspected.
[0088] Then, feature extraction is performed using LSTM modules. The aforementioned 600 sets of sequences are input into 600 LSTM modules in sequence. For example, see... Figure 3 As shown, Figure 3 This is a schematic diagram of a specific LSTM module provided in an embodiment of this application. Wherein, x t h is the input for the data in the current state. t-1 c t-1 These represent the input received from the previous node and the state value of the previous node, respectively. t This is the output for the current node state, while h t c t The current node's output and state value are passed to the next node. The first step in LSTM is to decide what information to discard from the input. This decision is made through a forget gate. This gate reads h... t-1 and x t Output a value between 0 and 1 to each value in c. t-1The numbers in the table represent their importance. 1 means "completely retain," and 0 means "completely discard." The second step is to decide how much new information to add to the processing state. This involves two steps: first, deciding which information needs to be updated and the alternative content to be updated, i.e., c. t Then, add the corresponding pixels from these two parts together to update the processed state. Next, add c... t-1 Updated to c t Finally, the output value needs to be determined. First, a sigmoid layer is run to determine which part of the current state will be output. Next, the current state is processed using the tanh function (resulting in a value between -1 and 1) and multiplied by the output of the sigmoid gate before being output. For example, in the LSTM(599) module: x t =x 598 =[A 598 ]、h t-1 =h 597 c t-1 =c 597 y t =y 598 h t =h 598 ct=c 598 .
[0089] Because the tail section image contains a lot of detailed information and the sintering process is a very complex chemical process, in order to ensure that more information is input into the decision-making stage, the features extracted by 600 LSTMs are densely connected. The input of the dense connection processing module 1 is the output of LSTM(1), and the input of the dense connection processing module 2 is the output of LSTM(1) and LSTM(2). The dense connection processing module (2) adds the input feature maps pixel by pixel, and so on. The dense connection processing module 600 adds the features output by LSTM(1) to LSTM(600) pixel by pixel. In this way, 600 vectors are obtained. Through dense connection, the types and number of features can be enriched, and important data can be prevented from being lost when forgetting the gate in the previous step. Furthermore, each feature vector is fully connected and then multiplied by weights. The weight parameters are the parameters after training, representing the importance of the vector. As shown in the figure, the 600 vectors are input into FC (fully connected processing module) 1, FC2, ..., FC600 in sequence, and fully connected are performed respectively. Then, they are multiplied by weight parameters W1, W2, ..., W600 respectively. 600 The vectors are multiplied together, and then fully connected again. The vectors are then input into a softmax classifier for classification, with the categories being optimal tail section and non-optimal tail section, thus obtaining the optimal section detection result.
[0090] As can be seen, this embodiment first obtains a training sample set, which includes optimal sintering machine tail section samples, non-optimal sintering machine tail section samples, and label information. Then, the training sample set is used to train a preset initial model to obtain a sintering machine tail section detection model. This model includes a recurrent neural network module. Finally, when a sintering machine tail section to be detected is obtained, the model outputs the optimal section detection result corresponding to that section. In other words, this embodiment first uses a training sample set including optimal and non-optimal sintering machine tail section samples to train the model, obtaining a sintering machine tail section detection model. Then, the trained model is used to perform optimal section detection. The model includes a recurrent neural network module, which can extract deep hidden information from the sintering machine tail section. This allows for accurate detection of the optimal tail section, improving production efficiency and saving production costs.
[0091] See Figure 4 As shown in the figure, this application discloses a sintering machine tail section detection device, including:
[0092] The sample acquisition module 11 is used to acquire a training sample set; wherein, the training sample set includes optimal sintering machine tail section samples, non-optimal sintering machine tail section samples, and label information.
[0093] The model training module 12 is used to train a preset initial model using the training sample set to obtain a sintering machine tail section detection model; wherein, the sintering machine tail section detection model includes a recurrent neural network module.
[0094] The cross-section detection module 13 is used to output the optimal cross-section detection result corresponding to the cross-section of the sintering machine tail section when the cross-section of the sintering machine tail section to be detected is obtained using the cross-section detection model of the sintering machine tail section to be detected.
[0095] As can be seen, this embodiment first obtains a training sample set, which includes optimal sintering machine tail section samples, non-optimal sintering machine tail section samples, and label information. Then, the training sample set is used to train a preset initial model to obtain a sintering machine tail section detection model. This model includes a recurrent neural network module. Finally, when a sintering machine tail section to be detected is obtained, the model outputs the optimal section detection result corresponding to that section. In other words, this embodiment first uses a training sample set including optimal and non-optimal sintering machine tail section samples to train the model, obtaining a sintering machine tail section detection model. Then, the trained model is used to perform optimal section detection. The model includes a recurrent neural network module, which can extract deep hidden information from the sintering machine tail section. This allows for accurate detection of the optimal tail section, improving production efficiency and saving production costs.
[0096] Specifically, the cross-section detection module 13 includes:
[0097] The cross-section segmentation submodule is used to segment the cross-section of the tail section of the sintering machine to be detected, and obtain multiple sets of image data sequences;
[0098] The feature extraction submodule is used to input the multiple sets of image data sequences into multiple recurrent neural network modules in the sintering machine tail section detection model to obtain multiple feature maps.
[0099] The dense connection submodule is used to perform dense connections on the multiple feature maps to obtain multiple feature vectors;
[0100] The feature fusion submodule is used to fuse the multiple feature vectors to obtain a fused vector;
[0101] The detection result acquisition submodule is used to input the fused vector into the classifier to obtain the optimal cross-section detection result corresponding to the cross-section of the sintering machine tail to be detected.
[0102] Furthermore, the sintering machine tail section detection model also includes multiple densely connected processing modules. The number of these densely connected processing modules is the same as the number of the recurrent neural network modules, and they are arranged sequentially. Accordingly, each densely connected sub-module is specifically used for:
[0103] The feature vector corresponding to the dense connection processing module is obtained by adding pixels of the feature map input to the dense connection processing module through any of the dense connection processing modules.
[0104] The feature map input to any of the dense connection processing modules includes feature maps output by recurrent neural network modules that have the same order as the dense connection processing module and those that have the same order as the preceding dense connection processing modules.
[0105] Furthermore, the feature fusion submodule specifically includes:
[0106] The first fully connected processing unit is used to perform fully connected processing on the multiple feature vectors respectively to obtain multiple fully connected processed vectors;
[0107] The second fully connected processing unit is used to perform fully connected processing on the multiple fully connected processed vectors to obtain a fused vector.
[0108] Furthermore, the second fully connected processing unit is specifically used to perform weight multiplication operations on multiple weight parameters and multiple fully connected processed vectors respectively to obtain multiple weight multiplied vectors; and to perform fully connected processing on the multiple weight multiplied vectors to obtain a fused vector.
[0109] In a specific implementation, the cross-section segmentation submodule specifically includes:
[0110] The segmentation direction determination unit is used to determine the segmentation direction corresponding to the tail section of the sintering machine to be detected;
[0111] The section segmentation unit is used to segment the cross section of the sintering machine tail section to be detected based on the segmentation direction to obtain multiple sets of image data sequences.
[0112] The feature extraction submodule is specifically used to process any set of image data sequences into a one-dimensional image data sequence if any set of image data sequences includes multiple image data sequences; and to input the multiple one-dimensional image data sequences into the multiple recurrent neural network modules in the sintering machine tail section detection model to obtain multiple feature maps.
[0113] See Figure 5 As shown in the figure, this application discloses an electronic device 20, including a processor 21 and a memory 22; wherein, the memory 22 is used to store a computer program; the processor 21 is used to execute the computer program, the sintering machine tail section detection method disclosed in the foregoing embodiment.
[0114] The specific process of the above-mentioned sintering machine tail section detection method can be found in the corresponding content disclosed in the foregoing embodiments, and will not be repeated here.
[0115] Furthermore, the memory 22, as a carrier for resource storage, can be a read-only memory, random access memory, disk, or optical disk, and the storage method can be temporary storage or permanent storage.
[0116] In addition, the electronic device 20 also includes a power supply 23, a communication interface 24, an input / output interface 25, and a communication bus 26; wherein, the power supply 23 is used to provide operating voltage for the various hardware devices on the electronic device 20; the communication interface 24 can create a data transmission channel between the electronic device 20 and external devices, and the communication protocol it follows can be any communication protocol applicable to the technical solution of this application, and is not specifically limited here; the input / output interface 25 is used to acquire external input data or output data to the outside world, and its specific interface type can be selected according to specific application needs, and is not specifically limited here.
[0117] Furthermore, this application also discloses a computer-readable storage medium for storing a computer program, wherein the computer program, when executed by a processor, implements the sintering machine tail section detection method disclosed in the foregoing embodiments.
[0118] The specific process of the above-mentioned sintering machine tail section detection method can be found in the corresponding content disclosed in the foregoing embodiments, and will not be repeated here.
[0119] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the apparatus disclosed in the embodiments, since it corresponds to the method disclosed in the embodiments, the description is relatively simple; relevant parts can be referred to in the method section.
[0120] The steps of the methods or algorithms described in conjunction with the embodiments disclosed herein can be implemented directly by hardware, a software module executed by a processor, or a combination of both. The software module can be located in random access memory (RAM), main memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disk, removable disk, CD-ROM, or any other form of storage medium known in the art.
[0121] The above provides a detailed description of the sintering machine tail section detection method, apparatus, equipment, and medium provided in this application. Specific examples have been used to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of this application. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of this application. Therefore, the content of this specification should not be construed as a limitation of this application.
Claims
1. A method for detecting the cross-section of a sintering machine tail section, characterized in that, include: Obtain a training sample set; wherein the training sample set includes optimal sintering machine tail section samples, non-optimal sintering machine tail section samples, and label information; The preset initial model is trained using the training sample set to obtain the sintering machine tail section detection model; wherein, the sintering machine tail section detection model includes a recurrent neural network module; Once the tail section of the sintering machine to be inspected is obtained, the optimal cross-section inspection result corresponding to the tail section of the sintering machine to be inspected is output using the sintering machine tail section inspection model. The step of using the sintering machine tail section detection model to output the optimal section detection result corresponding to the sintering machine tail section to be detected includes: segmenting the sintering machine tail section to be detected to obtain multiple sets of image data sequences; inputting the multiple sets of image data sequences into multiple recurrent neural network modules in the sintering machine tail section detection model to obtain multiple feature maps; densely connecting the multiple feature maps to obtain multiple feature vectors; fusing the multiple feature vectors to obtain a fusion vector; and inputting the fusion vector into a classifier to obtain the optimal section detection result corresponding to the sintering machine tail section to be detected. The sintering machine tail section detection model further includes multiple densely connected processing modules. The number of these densely connected processing modules is the same as the number of the recurrent neural network modules, and they are arranged sequentially. Accordingly, the process of densely connecting the multiple feature maps to obtain multiple feature vectors includes: adding pixels of the feature map input to any of the densely connected processing modules to obtain the feature vector corresponding to that densely connected processing module; wherein, the feature map input to any of the densely connected processing modules includes feature maps output by recurrent neural network modules that have the same order as the densely connected processing module and those that have the same order as the preceding densely connected processing modules.
2. The method for detecting the tail section of a sintering machine according to claim 1, characterized in that, The process of fusing the multiple feature vectors to obtain a fused vector includes: The multiple feature vectors are each subjected to fully connected processing to obtain multiple fully connected processed vectors; The multiple fully connected vectors are then subjected to full-connection processing to obtain a fused vector.
3. The method for detecting the tail section of a sintering machine according to claim 2, characterized in that, The step of performing full-connection processing on the multiple fully connected vectors to obtain a fused vector includes: Multiple weight parameters and multiple fully connected processed vectors are multiplied together to obtain multiple multiplied vectors. The vector obtained by multiplying the multiple weights is fully connected to obtain the fused vector.
4. The method for detecting the tail section of a sintering machine according to claim 1, characterized in that, The step involves segmenting the cross-section of the sintering machine tail section to be inspected, resulting in multiple sets of image data sequences, including: Determine the segmentation direction corresponding to the tail section of the sintering machine to be tested; The cross-section of the sintering machine tail section to be detected is segmented based on the segmentation direction to obtain multiple sets of image data sequences.
5. The method for detecting the tail section of a sintering machine according to claim 1, characterized in that, The process involves inputting the multiple sets of image data sequences into multiple recurrent neural network modules within the sintering machine tail section detection model to obtain multiple feature maps, including: If any set of image data sequences includes multiple image data sequences, then the set of image data sequences is processed into a one-dimensional image data sequence; Multiple one-dimensional image data sequences are respectively input into multiple recurrent neural network modules in the sintering machine tail section detection model to obtain multiple feature maps.
6. A sintering machine tail section detection device, characterized in that, include: The sample acquisition module is used to acquire a training sample set; wherein, the training sample set includes optimal sintering machine tail section samples, non-optimal sintering machine tail section samples, and label information. The model training module is used to train a preset initial model using the training sample set to obtain a sintering machine tail section detection model; wherein, the sintering machine tail section detection model includes a recurrent neural network module. The cross-section detection module is used to output the optimal cross-section detection result corresponding to the cross-section of the sintering machine tail section when the cross-section of the sintering machine tail section to be detected is obtained using the cross-section detection model of the sintering machine tail section. The cross-section detection module specifically includes: The cross-section segmentation submodule is used to segment the cross-section of the tail section of the sintering machine to be detected, and obtain multiple sets of image data sequences; The feature extraction submodule is used to input the multiple sets of image data sequences into multiple recurrent neural network modules in the sintering machine tail section detection model to obtain multiple feature maps. The dense connection submodule is used to perform dense connections on the multiple feature maps to obtain multiple feature vectors; The feature fusion submodule is used to fuse the multiple feature vectors to obtain a fused vector; The detection result acquisition submodule is used to input the fused vector into the classifier to obtain the optimal cross-section detection result corresponding to the cross-section of the sintering machine tail to be detected. The sintering machine tail section detection model also includes multiple densely connected processing modules. The number of densely connected processing modules is the same as that of the recurrent neural network modules, and they are arranged sequentially. Accordingly, each densely connected sub-module is specifically used to: add pixels of the feature map input to the densely connected processing module through any of the densely connected processing modules to obtain the feature vector corresponding to the densely connected processing module; wherein, the feature map input to any of the densely connected processing modules includes feature maps output by recurrent neural network modules that have the same order as the densely connected processing module and those that have the same order as the preceding densely connected processing modules.
7. An electronic device, characterized in that, Includes processor and memory; among which, The memory is used to store computer programs; The processor is configured to execute the computer program to implement the sintering machine tail section detection method as described in any one of claims 1 to 5.
8. A computer-readable storage medium, characterized in that, Used to store a computer program, wherein the computer program, when executed by a processor, implements the sintering machine tail section detection method as described in any one of claims 1 to 5.