Carbon solidified material element distribution prediction method and device, electronic equipment and medium

Abstract

The present application relates to a kind of carbon solidification material element distribution prediction method, device, electronic equipment and medium, belong to image processing technical field, wherein, the method includes: obtaining the first backscattering image of the first area of carbon solidification material and at least one first energy scattering spectrum image, data preprocessing is carried out to first backscattering image and first energy scattering spectrum image, respectively obtain second backscattering image and second energy scattering spectrum image;Second backscattering image and second energy scattering spectrum image are both carried out image segmentation and data enhancement, respectively obtain and third backscattering image and third energy scattering spectrum image are input into improved U-Net network model and constitute data set and are trained, obtain trained improved U-Net network model;Target backscattering image is input into the improved U-Net network model of complete training, and target material image element distribution prediction graph is obtained.The present application can quickly predict the distribution of multiple elements in carbon solidification material.

Description

Technical Field

[0001] This invention relates to the field of image processing technology, and in particular to a method, apparatus, electronic device, and medium for predicting the elemental distribution of carbon-cured materials. Background Technology

[0002] In the research and development and performance optimization of carbon-cured materials, the distribution of mineralized products and unreacted phases directly affects the carbonization process and final mechanical properties of the material. Studying the microstructure of carbon-cured materials and the distribution of elements such as calcium, silicon, and aluminum in the corresponding regions is an effective means of analyzing the phase distribution law of the material.

[0003] Traditional characterization techniques such as X-ray diffraction (XRD) and X-ray fluorescence (XRF) can provide information on phase composition, but they cannot reveal differences in elemental distribution in micron-scale regions. Scanning electron microscopy-energy dispersive spectroscopy (SEM-EDS) is currently the main method for analyzing the elemental distribution in micro-regions of materials, but it also has significant drawbacks: traditional SEM-EDS area scanning methods take 30 minutes to 4 hours per sample in carbon-cured materials, while the development of carbon-cured materials requires batch screening of samples with different ratios, making this method time-consuming and costly; existing deep learning models are mostly designed for single-element prediction.

[0004] In summary, existing technologies lack a method for predicting the elemental distribution of carbon-cured materials, thus failing to rapidly predict the distribution of multiple elements in carbon-cured materials. Summary of the Invention

[0005] In view of this, it is necessary to provide a method, apparatus, electronic device and medium for predicting the elemental distribution of carbon-cured materials, so as to achieve the purpose of rapidly predicting the distribution of multiple elements in carbon-cured materials.

[0006] To achieve the above objectives, in a first aspect, the present invention provides a method for predicting the elemental distribution of carbon-cured materials, comprising:

[0007] Acquire a first backscattered image and at least one first energy scattering spectrum image of a first region of the carbon-cured material, wherein the number of first energy scattering spectrum images is the same as the number of chemical elements pre-detected in the carbon-cured material;

[0008] The first backscattered image and the first energy scattering spectrum image are preprocessed to obtain the second backscattered image and the second energy scattering spectrum image, respectively.

[0009] Image segmentation and data augmentation were performed on both the second backscattered image and the second energy scattering spectrum image to obtain the third backscattered image and the third energy scattering spectrum image, respectively.

[0010] The dataset consisting of the third backscattered image and the third energy scattering spectrum image is input into the improved U-Net network model for training, resulting in a fully trained improved U-Net network model. The improved U-Net network model includes an encoder and a decoder. The encoder's convolutional layer uses a 3×3 convolutional kernel, and the downsampling layer uses a 2×2 window for max pooling. A cross-element channel attention mechanism is added between the convolutional layer and the downsampling layer. A stoichiometric constraint is introduced into the upsampling layer of the decoder, and cross-element skip connections are added to the decoder.

[0011] The target backscattered image is input into a fully trained improved U-Net network model to obtain a predicted image of the elemental distribution of the target carbon-cured material.

[0012] In one possible implementation, the chemical elements pre-detected for the carbon-cured material include:

[0013] One or more of the elements aluminum, silicon, and calcium.

[0014] In one possible implementation, the data preprocessing of the first backscattered image and the first energy scattering spectrum image to obtain the second backscattered image and the second energy scattering spectrum image respectively includes:

[0015] The first backscattered image is normalized and interference is removed to obtain the second backscattered image;

[0016] The first energy scattering spectrum image is normalized and interference is removed to obtain the second energy scattering spectrum image.

[0017] In one possible implementation, the normalization and interference removal of the first energy scattering spectrum image to obtain the second energy scattering spectrum image includes:

[0018] Based on the grayscale value of the first backscattered image, the surface epoxy resin distribution area of ​​the carbon-cured material is obtained;

[0019] Using the surface epoxy resin distribution area as a mask, the corresponding position of the first energy scattering spectrum image is masked to obtain a temporary energy scattering spectrum image;

[0020] The temporary energy scattering spectrum image is normalized and interference is removed to obtain the second energy scattering spectrum image.

[0021] In one possible implementation, before performing image segmentation and data augmentation on both the second backscattered image and the second energy scattering spectrum image to obtain the third backscattered image and the third energy scattering spectrum image respectively, the method further includes:

[0022] Image enhancement was performed on the silicon-rich region in the third energy scattering spectrum image.

[0023] In one possible implementation, the expression for the cross-element channel attention mechanism is:

[0024]

[0025] In the formula, This indicates a cross-element channel attention mechanism. Characteristic diagram of aluminum element, This represents a characteristic map of silicon. A diagram showing the characteristics of calcium elements. Represents a 1×1 convolution. This indicates global average pooling for aluminum, and this indicates global max pooling for silicon. This represents the input feature map.

[0026] One possible implementation also includes:

[0027] Calculate the first stoichiometry of the target region in the target material image element distribution prediction map;

[0028] Calculate the second stoichiometry of the target region in the energy scattering spectrum image of the target material;

[0029] Based on the first stoichiometry and the second stoichiometry, an error distribution diagram is obtained.

[0030] Secondly, the present invention also provides a device for predicting the elemental distribution of carbon-cured materials, comprising:

[0031] A dataset construction module is used to acquire a first backscattered image and at least one first energy scattering spectrum image of a first region of the carbon-cured material, wherein the number of the first energy scattering spectrum images is the same as the number of chemical elements pre-detected in the carbon-cured material;

[0032] The dataset preprocessing module is used to preprocess the first backscattered image and the first energy scattering spectrum image to obtain the second backscattered image and the second energy scattering spectrum image, respectively.

[0033] The image segmentation and data enhancement module is used to perform image segmentation and data enhancement on both the second backscattered image and the second energy scattering spectrum image to obtain the third backscattered image and the third energy scattering spectrum image, respectively.

[0034] The model training module is used to input the dataset consisting of the third backscattered image and the third energy scattering spectrum image into the improved U-Net network model for training, and obtain a fully trained improved U-Net network model. The improved U-Net network model includes an encoder and a decoder. The encoder's convolutional layer uses a 3×3 convolutional kernel, the downsampling layer uses a 2×2 window for max pooling, and a cross-element channel attention mechanism is added between the convolutional layer and the downsampling layer. The decoder's upsampling layer introduces stoichiometry constraints, and cross-element skip connections are added to the decoder.

[0035] The element distribution prediction map acquisition module is used to input the target backscattered image into a fully trained improved U-Net network model to obtain the element distribution prediction map of the target carbon-cured material image.

[0036] Thirdly, the present invention also provides an electronic device, including a memory and a processor, wherein,

[0037] The memory is used to store programs;

[0038] The processor, coupled to the memory, is used to execute the program stored in the memory to implement the steps in the carbon-cured material element distribution prediction method described in any of the above implementations.

[0039] Fourthly, the present invention also provides a computer-readable storage medium for storing a computer-readable program or instructions, which, when executed by a processor, can implement the steps in the carbon-cured material element distribution prediction method described in any of the above implementations.

[0040] The beneficial effects of this invention are as follows: This invention provides a method for predicting the elemental distribution of carbon-cured materials. First, it acquires a first backscattered image and at least one first energy scattering spectrum image of a first region of the carbon-cured material. The number of first energy scattering spectrum images is the same as the number of chemical elements to be pre-detected in the carbon-cured material, establishing a mapping relationship between the backscattered spectrum images and energy scattering spectrum images of the same region of the carbon-cured material. Data preprocessing is performed on the first backscattered image and the first energy scattering spectrum image to obtain a second backscattered image and a second energy scattering spectrum image, respectively. Image segmentation and data enhancement are then performed on both the second backscattered image and the second energy scattering spectrum image to obtain a third backscattered image and a third energy scattering spectrum image, respectively. The third backscattered image and the third energy scattering spectrum image are then combined to form a... The dataset is input into an improved U-Net network model for training, resulting in a fully trained improved U-Net network model. The encoder of the improved U-Net network model uses 3×3 convolutional kernels, and the downsampling layer uses a 2×2 window for max pooling. A cross-element channel attention mechanism is added between the convolutional and downsampling layers. The decoder of the improved U-Net network model introduces a stoichiometric constraint upsampling layer and adds cross-element skip connections. This improvement on the original U-Net network model enhances prediction efficiency. The target backscattered image is input into the fully trained improved U-Net network model to obtain a predicted elemental distribution map of the target carbon-cured material image. This invention uses backscattered images as a dataset to train the improved U-Net network model, thus obtaining the elemental distribution of carbon-cured materials directly from the backscattered images. This significantly reduces the time compared to obtaining elemental distribution prediction maps through SEM-EDS area scanning, and the improved U-Net network model can predict the distribution maps of multiple elements. Attached Figure Description

[0041] Figure 1 A flowchart illustrating an embodiment of the method for predicting the elemental distribution of carbon-cured materials provided by the present invention;

[0042] Figure 2 This is a schematic diagram of a sample of an embodiment of the elemental distribution prediction method for carbon-cured materials provided by the present invention;

[0043] Figure 3 This is a comparison chart of predicted and actual values ​​for an embodiment of the elemental distribution prediction method for carbon-cured materials provided by the present invention;

[0044] Figure 4 A backscattered image of a backscattered image in one embodiment of a method for predicting the elemental distribution of carbon-cured materials provided by the present invention;

[0045] Figure 5This is a schematic flowchart of an embodiment of the elemental distribution prediction device for carbon-cured materials provided by the present invention;

[0046] Figure 6 A schematic diagram of an embodiment of the electronic device provided by the present invention. Detailed Implementation

[0047] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0048] In the description of the embodiments of the present invention, unless otherwise stated, "multiple" means two or more. "And / or" describes the relationship between related objects, indicating that there can be three relationships. For example, A and / or B can represent three situations: A exists alone, A and B exist simultaneously, and B exists alone.

[0049] The terms "first," "second," etc., used in the embodiments of this invention are for descriptive purposes only and should not be construed as indicating or implying their relative importance or implicitly specifying the number of technical features indicated. Therefore, a technical feature defined with "first" or "second" may explicitly or implicitly include at least one of that feature.

[0050] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of the invention. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0051] Before demonstrating the embodiments, the following terms will be explained.

[0052] Backscattered electrons (BSE) play a crucial role in scanning electron microscopy (SEM) analysis. They not only provide researchers with detailed information on the surface composition and structure of samples but are also important tools for revealing the microscopic world of materials. The generation of backscattered electrons stems from the interaction between incident electrons and the atomic nuclei of the sample; this interaction typically manifests as elastic or inelastic scattering. It is this unique interaction that has led to the widespread application of BSE images in various fields, including materials science, earth science, biology, and medical research.

[0053] EDS images are images obtained through energy dispersive spectroscopy (EDS). EDS is a method used to analyze the types and amounts of elements in micro-regions of materials, and it is usually used in conjunction with scanning electron microscopy (SEM).

[0054] This invention provides a method, apparatus, electronic device, and medium for predicting the elemental distribution of carbon-cured materials, which are described below.

[0055] Figure 1 This is a schematic flowchart of an embodiment of the elemental distribution prediction method for carbon-cured materials provided by the present invention, as shown below. Figure 1 As shown, the method for predicting the elemental distribution of carbon-cured materials includes:

[0056] S101. Obtain a first backscattered image and at least one first energy scattering spectrum image of a first region of the carbon-cured material, wherein the number of first energy scattering spectrum images is the same as the number of chemical elements pre-detected in the carbon-cured material.

[0057] It should be noted that the backscattered image is a grayscale image, such as... Figure 2 As shown in (a), the first energy scattering spectrum image is a predicted elemental distribution map. Figure 2 (b) Figure 2 (c) and Figure 2 As shown in (d), the backscattered image corresponds to the first energy scattering spectrum image; both are images of the same region of the carbon-cured material. The EDS spectrometer has three modes: point scan, line scan, and area scan. In this embodiment, the area scan mode is used.

[0058] S102. Perform data preprocessing on the first backscattered image and the first energy scattering spectrum image to obtain the second backscattered image and the second energy scattering spectrum image, respectively.

[0059] S103. Perform image segmentation and data augmentation on both the second backscattered image and the second energy scattering spectrum image to obtain the third backscattered image and the third energy scattering spectrum image, respectively.

[0060] S104. Input the dataset consisting of the third backscattered image and the third energy scattering spectrum image into the improved U-Net network model for training to obtain a fully trained improved U-Net network model. The improved U-Net network model includes an encoder and a decoder. The encoder's convolutional layer uses a 3×3 convolutional kernel, and the downsampling layer uses a 2×2 window for max pooling. A cross-element channel attention mechanism is added between the convolutional layer and the downsampling layer. A stoichiometric constraint is introduced into the upsampling layer of the decoder, and a cross-element skip connection is added to the decoder.

[0061] S105. Input the target backscattered image into the fully trained improved U-Net network model to obtain the predicted image of the element distribution of the target carbon-cured material.

[0062] Compared with existing technologies, this embodiment provides a method for predicting the elemental distribution of carbon-cured materials. First, it acquires a first backscattered image and at least one first energy scattering spectrum image of a first region of the carbon-cured material. The number of first energy scattering spectrum images is the same as the number of chemical elements to be pre-detected in the carbon-cured material, establishing a mapping relationship between the backscattered spectrum images and energy scattering spectrum images of the same region of the carbon-cured material. Data preprocessing is then performed on the first backscattered image and the first energy scattering spectrum image to obtain a second backscattered image and a second energy scattering spectrum image, respectively. Image segmentation and data enhancement are then performed on both the second backscattered image and the second energy scattering spectrum image to obtain a third backscattered image and a third energy scattering spectrum image, respectively. The data composed of the third backscattered image and the third energy scattering spectrum image is then... The dataset is input into an improved U-Net network model for training, resulting in a fully trained improved U-Net network model. The encoder of the improved U-Net network model uses 3×3 convolutional kernels, and the downsampling layer uses a 2×2 window for max pooling. A cross-element channel attention mechanism is added between the convolutional and downsampling layers. The decoder of the improved U-Net network model introduces a stoichiometric constraint upsampling layer and adds cross-element skip connections. This improvement on the original U-Net network model enhances prediction efficiency. The target backscattered image is input into the fully trained improved U-Net network model to obtain a predicted element distribution map of the target carbon-cured material image. This invention uses backscattered images as a dataset to train the improved U-Net network model, thus obtaining the element distribution of carbon-cured materials directly from the backscattered images. This significantly reduces the time compared to obtaining element distribution prediction maps through SEM-EDS area scanning, and the improved U-Net network model can predict the distribution maps of multiple elements.

[0063] In some embodiments of the present invention, the chemical elements pre-detected for the carbon-cured material include:

[0064] One or more of the elements aluminum, silicon, and calcium.

[0065] In a specific embodiment of the present invention, step S101 involves acquiring a first backscattered image of a first region of the carbon-cured material and at least one first energy scattering spectrum image. The number of first energy scattering spectrum images is the same as the number of chemical elements pre-detected in the carbon-cured material. Specifically, this includes:

[0066] Data acquisition was performed using an environmental scanning electron microscope and an energy dispersive spectroscopy (EDS) spectrometer, with a resolution of at least 300 dpi and a pixel size of 0.085 μm. The acquired images were saved in 16-bit uncompressed TIFF format.

[0067] Collect more than 5 sets of standard samples, such as Figure 2 As shown, each group contains one BSE image and three EDS area scan element distribution prediction maps, corresponding to the element distribution of Al, Si, and Ca in the BSE image capture area, respectively.

[0068] The element distribution prediction map corresponds to the RGB channels: Al: green channel, Si: blue-green dual channel (G and B channel values ​​are equal), and Ca: red channel. The colors of the element distribution prediction map can be adjusted according to the test results of the image acquisition equipment and the actual experimental needs.

[0069] In some embodiments of the present invention, the step of performing data preprocessing on the first backscattered image and the first energy scattering spectrum image to obtain the second backscattered image and the second energy scattering spectrum image, respectively, includes:

[0070] The first backscattered image is normalized and interference is removed to obtain the second backscattered image;

[0071] The first energy scattering spectrum image is normalized and interference is removed to obtain the second energy scattering spectrum image.

[0072] In some embodiments of the present invention, the step of normalizing and removing interference from the first energy scattering spectrum image to obtain the second energy scattering spectrum image includes:

[0073] Based on the grayscale value of the first backscattered image, the surface epoxy resin distribution area of ​​the carbon-cured material is obtained;

[0074] Using the surface epoxy resin distribution area as a mask, the corresponding position of the first energy scattering spectrum image is masked to obtain a temporary energy scattering spectrum image;

[0075] The temporary energy scattering spectrum image is normalized and interference is removed to obtain the second energy scattering spectrum image.

[0076] In a specific embodiment of the present invention, step S102 involves data preprocessing of the first backscattered image and the first energy scattering spectrum image to obtain the second backscattered image and the second energy scattering spectrum image, specifically including:

[0077] The first energy scattering spectrum image is processed by resin region masking. Pixels with gray values ​​≤5±2 are identified as resin regions and set to zero. The maximum-minimum method is used to linearly map all image data to the [0,1] interval. The mean-variance standardization is performed on each sample, and outliers are removed using the 3σ principle to obtain the second backscattering image and the second energy scattering spectrum image.

[0078] In some embodiments of the present invention, before performing image segmentation and data enhancement on both the second backscattered image and the second energy scattering spectrum image to obtain the third backscattered image and the third energy scattering spectrum image respectively, the method further includes:

[0079] Image enhancement was performed on the silicon-rich region in the third energy scattering spectrum image.

[0080] In a specific embodiment of the present invention, in order to solve the problem of memory overflow caused by directly inputting large-size backscattered images (≥2048×2048) into the network, a multi-scale dynamic block segmentation strategy is proposed. At the same time, overlapping region design and boundary processing are used to ensure that the segmented image blocks can be seamlessly stitched together.

[0081] Window parameter settings: Based on GPU memory capacity and element distribution characteristics, set the window size to 86×128 pixels; to prevent artifacts at the stitching boundaries, set the vertical overlap to 5 pixels (overlap rate 5.8%).

[0082] Image boundary processing employs a combination of mirror filling and zero-value filling. Zero-value filling is used to fill to the standard size, and incomplete areas are specially marked to facilitate the identification of invalid areas during subsequent stitching.

[0083] The data augmentation part uses geometric transformation parameters to rotate the angle from -15° to 15°; it uses bicubic interpolation for interpolation; the horizontal flip probability is 50%±5%, the vertical flip probability is 50%±5%, allowing simultaneous horizontal and vertical flips; the transformation matrix is ​​shared to ensure spatial consistency; and stoichiometry is introduced to maintain the augmentation.

[0084] In addition, the enhancement of Si-rich regions is further enhanced, specifically by locally cropping regions with Si concentration > 0.38 by ±10 pixels.

[0085] In some embodiments of the present invention, the expression for the cross-element channel attention mechanism is:

[0086]

[0087] In the formula, This indicates a cross-element channel attention mechanism. Characteristic diagram of aluminum element, This represents a characteristic map of silicon. A diagram showing the characteristics of calcium elements. Represents a 1×1 convolution. This indicates global average pooling for aluminum, and this indicates global max pooling for silicon. This represents the input feature map.

[0088] In a specific embodiment of the present invention, the specific process of training the improved U-Net network model in step S104 is as follows:

[0089] In the encoder section, the convolutional layers use a 3×3 kernel with a stride of 1 (except for the downsampling layers), and edge regions are padded to maintain size; the downsampling method is 2×2 max pooling with a stride of 2. A cross-element channel attention mechanism is designed, prioritizing feature extraction from high-calcium regions and coupling with gradient backpropagation in silicon and aluminum channels. The expression for the attention mechanism is:

[0090] In the formula, This indicates a cross-element channel attention mechanism. Characteristic diagram of aluminum element, This represents a characteristic map of silicon. A diagram showing the characteristics of calcium elements. Represents a 1×1 convolution. This indicates global average pooling for aluminum, and this indicates global max pooling for silicon. This represents the input feature map.

[0091] In the decoder section, a stoichiometric constraint upsampling layer is added; cross-element skip connections are added: the expression for the skip connections is:

[0092]

[0093] In the formula, Indicates the characteristics of upsampled aluminum. Indicates horizontally offset silicon features. Indicates the calcium characteristics of the downsampled sample. This represents a 3×3 convolution.

[0094] The training hyperparameters are set as follows: AdamW optimizer (β1=0.9, β2=0.99), initial learning rate: η=5e-4, element-aware learning rate decay is implemented.

[0095]

[0096] In some embodiments of the present invention, it further includes:

[0097] Calculate the first stoichiometry of the target region in the target material image element distribution prediction map;

[0098] Calculate the second stoichiometry of the target region in the energy scattering spectrum image of the target material;

[0099] Based on the first stoichiometry and the second stoichiometry, an error distribution diagram is obtained.

[0100] It is understandable that once the image element distribution prediction map of the target carbon-cured material is obtained, the relative content data of the corresponding element can be obtained through each pixel in the image element distribution prediction map of the target carbon-cured material.

[0101] In a specific embodiment of the present invention, the multi-scale prediction results are fused and stitched together, and the prediction results are subjected to adaptive histogram equalization. The channel mapping of the prediction distribution image output is consistent with the mapping of each element in the image acquisition step (i.e., in this embodiment, A1: green channel; Si: blue-green dual channel; Ca: red channel).

[0102] The regional stoichiometric ratio error was calculated for elemental distribution verification, and an error distribution map was generated.

[0103]

[0104]

[0105] In the formula, Indicates the predicted value. This represents the actual value obtained through EDS.

[0106] The following specific example illustrates the technical effects produced by this application.

[0107] Taking calcium aluminosilicate-based carbon curing material as an example, the BSE image acquisition parameters are: accelerating voltage 20kV and working distance 10mm. The model still maintains Train Loss<0.020 and Test Loss<0.025 with a low sample size (<10 groups), which is better than the conventional model (requiring 1000+ samples).

[0108] The distribution of Al, Si, and Ca elements was predicted using the model described in the above embodiment. The prediction results were compared with the test results as follows:

[0109] pass Figure 3 It can be seen that the prediction method of the present invention has a good match between the prediction results and the actual test results, and the contrast of the predicted image is relatively high, showing the main enriched areas of elements. Furthermore, the morphology of some element-rich areas in the predicted element distribution image of the present invention is similar to that of the BSE image (e.g., ...). Figure 4 The high degree of matching between the phase morphology and the BSE image (as shown) proves that the prediction method based on this application can effectively extract the element morphology information from the BSE image and correct the prediction of element distribution to a certain extent.

[0110] To better implement the elemental distribution prediction method for carbon-cured materials in this embodiment of the invention, based on the elemental distribution prediction method for carbon-cured materials, correspondingly, as follows: Figure 5As shown, this embodiment of the invention also provides a carbon-cured material element distribution prediction device. A carbon-cured material element distribution prediction device 500 includes:

[0111] The dataset construction module 501 is used to acquire a first backscattered image and at least one first energy scattering spectrum image of a first region of the carbon-cured material, wherein the number of the first energy scattering spectrum images is the same as the number of chemical elements pre-detected in the carbon-cured material;

[0112] The dataset preprocessing module 502 is used to preprocess the first backscattered image and the first energy scattering spectrum image to obtain the second backscattered image and the second energy scattering spectrum image, respectively.

[0113] The image segmentation and data enhancement module 503 is used to perform image segmentation and data enhancement on both the second backscattered image and the second energy scattering spectrum image to obtain the third backscattered image and the third energy scattering spectrum image, respectively.

[0114] The model training module 504 is used to input the dataset consisting of the third backscattered image and the third energy scattering spectrum image into the improved U-Net network model for training, so as to obtain a fully trained improved U-Net network model. The improved U-Net network model includes an encoder and a decoder. The encoder's convolutional layer uses a 3×3 convolutional kernel, the downsampling layer uses a 2×2 window for max pooling, and a cross-element channel attention mechanism is added between the convolutional layer and the downsampling layer. The decoder's upsampling layer introduces stoichiometry constraints, and cross-element skip connections are added to the decoder.

[0115] The element distribution prediction map acquisition module 505 is used to input the target backscattered image into a fully trained improved U-Net network model to obtain the element distribution prediction map of the target carbon-cured material image.

[0116] The carbon-cured material element distribution prediction device 500 provided in the above embodiments can realize the technical solution described in the above embodiments of the carbon-cured material element distribution prediction method. The specific implementation principle of each module or unit can be found in the corresponding content in the above embodiments of the carbon-cured material element distribution prediction method, which will not be repeated here.

[0117] like Figure 6 As shown, the present invention also provides an electronic device 600. The electronic device 600 includes a processor 601, a memory 602, and a display 603. Figure 6 Only some components of the electronic device 600 are shown, but it should be understood that it is not required to implement all the components shown, and more or fewer components may be implemented instead.

[0118] In some embodiments, processor 601 may be a central processing unit (CPU), microprocessor, or other data processing chip, used to run program code stored in memory 602 or process data, such as a method for predicting the elemental distribution of carbon-cured materials in this invention.

[0119] In some embodiments, processor 601 may be a single server or a group of servers. The server group may be centralized or distributed. In some embodiments, processor 601 may be local or remote. In some embodiments, processor 601 may be implemented on a cloud platform. In some embodiments, the cloud platform may include a private cloud, public cloud, hybrid cloud, community cloud, distributed cloud, internal cloud, multi-cloud, or any combination thereof.

[0120] In some embodiments, memory 602 may be an internal storage unit of electronic device 600, such as a hard disk or memory of electronic device 600. In other embodiments, memory 602 may also be an external storage device of electronic device 600, such as a plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, etc. equipped on electronic device 600.

[0121] Furthermore, the memory 602 may include both internal storage units of the electronic device 600 and external storage devices. The memory 602 is used to store application software and various types of data installed on the electronic device 600.

[0122] In some embodiments, display 603 may be an LED display, a liquid crystal display, a touch-sensitive liquid crystal display, or an OLED (Organic Light-Emitting Diode) touchscreen. Display 603 is used to display information from electronic device 600 and to display a visual user interface. Components 601-603 of electronic device 600 communicate with each other via a system bus.

[0123] In one embodiment, when processor 601 executes a carbon-cured material element distribution prediction program stored in memory 602, the following steps can be implemented:

[0124] Acquire a first backscattered image and at least one first energy scattering spectrum image of a first region of the carbon-cured material, wherein the number of first energy scattering spectrum images is the same as the number of chemical elements pre-detected in the carbon-cured material;

[0125] The first backscattered image and the first energy scattering spectrum image are preprocessed to obtain the second backscattered image and the second energy scattering spectrum image, respectively.

[0126] Image segmentation and data augmentation were performed on both the second backscattered image and the second energy scattering spectrum image to obtain the third backscattered image and the third energy scattering spectrum image, respectively.

[0127] The dataset consisting of the third backscattered image and the third energy scattering spectrum image is input into the improved U-Net network model for training, resulting in a fully trained improved U-Net network model. The improved U-Net network model includes an encoder and a decoder. The encoder's convolutional layer uses a 3×3 convolutional kernel, and the downsampling layer uses a 2×2 window for max pooling. A cross-element channel attention mechanism is added between the convolutional layer and the downsampling layer. A stoichiometric constraint is introduced into the upsampling layer of the decoder, and cross-element skip connections are added to the decoder.

[0128] The target backscattered image is input into a fully trained improved U-Net network model to obtain a predicted image of the target material element distribution.

[0129] It should be understood that when the processor 601 executes a carbon-cured material element distribution prediction program in the memory 602, in addition to the functions mentioned above, it can also perform other functions, as detailed in the description of the corresponding method embodiments above.

[0130] Furthermore, the embodiments of the present invention do not specifically limit the type of electronic device 600 mentioned. Electronic device 600 can be a mobile phone, tablet computer, personal digital assistant (PDA), wearable device, laptop computer, or other portable electronic device. Exemplary embodiments of portable electronic devices include, but are not limited to, portable electronic devices running iOS, Android, Microsoft, or other operating systems. The aforementioned portable electronic device can also be other portable electronic devices, such as a laptop computer with a touch-sensitive surface (e.g., a touch panel). It should also be understood that in some other embodiments of the present invention, electronic device 600 may not be a portable electronic device, but rather a desktop computer with a touch-sensitive surface (e.g., a touch panel).

[0131] Those skilled in the art will understand that all or part of the processes of the methods described in the above embodiments can be implemented by a computer program instructing related hardware, and the program can be stored in a computer-readable storage medium. The computer-readable storage medium may be a disk, optical disk, read-only memory, or random access memory, etc.

[0132] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for predicting the elemental distribution of carbon-cured materials, characterized in that, include: Acquire a first backscattered image and at least one first energy scattering spectrum image of a first region of the carbon-cured material, wherein the number of first energy scattering spectrum images is the same as the number of chemical elements pre-detected in the carbon-cured material; The first backscattered image and the first energy scattering spectrum image are preprocessed to obtain the second backscattered image and the second energy scattering spectrum image, respectively. Image segmentation and data augmentation were performed on both the second backscattered image and the second energy scattering spectrum image to obtain the third backscattered image and the third energy scattering spectrum image, respectively. The dataset consisting of the third backscattered image and the third energy scattering spectrum image is input into the improved U-Net network model for training, resulting in a fully trained improved U-Net network model. The improved U-Net network model includes an encoder and a decoder. The encoder's convolutional layer uses a 3×3 convolutional kernel, and the downsampling layer uses a 2×2 window for max pooling. A cross-element channel attention mechanism is added between the convolutional layer and the downsampling layer. A stoichiometric constraint is introduced into the upsampling layer of the decoder, and cross-element skip connections are added to the decoder. The expression for the cross-element channel attention mechanism is: In the formula, This indicates a cross-element channel attention mechanism. Characteristic diagram of aluminum element, This represents a characteristic map of silicon. A diagram showing the characteristics of calcium elements. Represents a 1×1 convolution. This indicates global average pooling for aluminum, and this indicates global max pooling for silicon. Represents the input feature map; The target backscattered image is input into a fully trained improved U-Net network model to obtain a predicted image of the elemental distribution of the target carbon-cured material.

2. The method for predicting the elemental distribution of carbon-cured materials according to claim 1, characterized in that, The chemical elements pre-detected for the carbon-cured material include: One or more of the elements aluminum, silicon, and calcium.

3. The method for predicting the elemental distribution of carbon-cured materials according to claim 1, characterized in that, The step of preprocessing the first backscattered image and the first energy scattering spectrum image to obtain the second backscattered image and the second energy scattering spectrum image includes: The first backscattered image is normalized and interference is removed to obtain the second backscattered image; The first energy scattering spectrum image is normalized and interference is removed to obtain the second energy scattering spectrum image.

4. The method for predicting the elemental distribution of carbon-cured materials according to claim 1, characterized in that, The process of normalizing and removing interference from the first energy scattering spectrum image to obtain the second energy scattering spectrum image includes: Based on the grayscale value of the first backscattered image, the surface epoxy resin distribution area of ​​the carbon-cured material is obtained; Using the surface epoxy resin distribution area as a mask, the corresponding position of the first energy scattering spectrum image is masked to obtain a temporary energy scattering spectrum image; The temporary energy scattering spectrum image is normalized and interference is removed to obtain the second energy scattering spectrum image.

5. The method for predicting the elemental distribution of carbon-cured materials according to claim 2, characterized in that, Before performing image segmentation and data augmentation on both the second backscattered image and the second energy scattering spectrum image to obtain the third backscattered image and the third energy scattering spectrum image, respectively, the method further includes: Image enhancement was performed on the silicon-rich region in the third energy scattering spectrum image.

6. The method for predicting the elemental distribution of carbon-cured materials according to claim 1, characterized in that, Also includes: Calculate the first stoichiometry of the target region in the target material image element distribution prediction map; Calculate the second stoichiometry of the target region in the energy scattering spectrum image of the target material; Based on the first stoichiometry and the second stoichiometry, an error distribution diagram is obtained.

7. A device for predicting the elemental distribution of carbon-cured materials, characterized in that, include: A dataset construction module is used to acquire a first backscattered image and at least one first energy scattering spectrum image of a first region of the carbon-cured material, wherein the number of the first energy scattering spectrum images is the same as the number of chemical elements pre-detected in the carbon-cured material; The dataset preprocessing module is used to preprocess the first backscattered image and the first energy scattering spectrum image to obtain the second backscattered image and the second energy scattering spectrum image, respectively. The image segmentation and data enhancement module is used to perform image segmentation and data enhancement on both the second backscattered image and the second energy scattering spectrum image to obtain the third backscattered image and the third energy scattering spectrum image, respectively. The model training module is used to input the dataset consisting of the third backscattered image and the third energy scattering spectrum image into the improved U-Net network model for training, and obtain a fully trained improved U-Net network model. The improved U-Net network model includes an encoder and a decoder. The encoder's convolutional layer uses a 3×3 convolutional kernel, the downsampling layer uses a 2×2 window for max pooling, and a cross-element channel attention mechanism is added between the convolutional layer and the downsampling layer. The decoder's upsampling layer introduces stoichiometry constraints, and cross-element skip connections are added to the decoder. The expression for the cross-element channel attention mechanism is: In the formula, This indicates a cross-element channel attention mechanism. Characteristic diagram of aluminum element, This represents a characteristic map of silicon. A diagram showing the characteristics of calcium elements. Represents a 1×1 convolution. This indicates global average pooling for aluminum, and this indicates global max pooling for silicon. Represents the input feature map; The element distribution prediction map acquisition module is used to input the target backscattered image into a fully trained improved U-Net network model to obtain the element distribution prediction map of the target carbon-cured material image.

8. An electronic device, characterized in that, Including memory and processor, among which, The memory is used to store programs; The processor, coupled to the memory, is used to execute the program stored in the memory to implement the steps in the method for measuring the elemental distribution of carbon-cured materials according to any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that, Used to store computer-readable programs or instructions, which, when executed by a processor, can implement the steps in the method for predicting the elemental distribution of carbon-cured materials as described in any one of claims 1 to 6.

Images