Solid electrolyte interface analysis method, device and equipment and storage medium

Abstract

The invention belongs to the technical field of image analysis, and discloses a solid electrolyte interface analysis method, device and equipment and a storage medium. The method comprises the following steps: acquiring a solid electrolyte interface image; inputting the solid electrolyte interface image into a preset instance segmentation model for instance segmentation to obtain a plurality of lattice areas, and classifying the lattice areas to obtain component information corresponding to the lattice areas; and completing solid electrolyte interface analysis according to the component information corresponding to each lattice area. It needs to be explained that according to the scheme, the lattice area and the non-lattice area are distinguished by segmenting the solid electrolyte interface image instance, then the components of the lattice area can be rapidly obtained by classifying all the subareas in the lattice area, the components of the lattice area in the solid electrolyte interface are directly determined through the end-to-end model, and the images are processed in batches; the method is applied to a video analysis process, and the analysis efficiency is improved.

Description

Technical Field

[0001] The present invention relates to the field of image analysis technology, and in particular to a solid electrolyte interface analysis method, device, equipment and storage medium. Background Art

[0002] SEI membrane (solid electrolyte interface) is an important component of the battery. The analysis of its composition is of great significance to battery research. SEI membrane is a thin layer structure formed between the electrode material and the electrolyte. It is formed spontaneously during the first charge and discharge process. When the battery is charged for the first time, the potential on the surface of the negative electrode drops to a certain level, causing the solvent molecules or lithium salts in the electrolyte to undergo a reduction reaction. These reaction products are deposited on the surface of the negative electrode. These decomposition products gradually accumulate to form a solid film layer covering the negative electrode active material. This film layer has a certain conductivity, allowing lithium ions to pass through but preventing electrons from passing through, thus preventing further direct electrochemical reactions. SEI membrane effectively isolates the electrode material and the electrolyte. The properties of SEI membrane will affect the transmission efficiency of lithium ions, and then affect the battery's charge and discharge rate, capacity retention rate and other performance indicators. Therefore, its formation and evolution process is crucial in battery material research.

[0003] Currently, electron microscopy is usually used to photograph SEI for characterization and analysis, but there are great problems with efficiency and accuracy.

[0004] The above contents are only used to assist in understanding the technical solution of the present invention and do not constitute an admission that the above contents are prior art. Summary of the invention

[0005] The main purpose of the present invention is to provide a solid electrolyte interface analysis method, device, equipment and storage medium, aiming to solve the technical problem of low SEI film analysis efficiency in the prior art.

[0006] To achieve the above object, the present invention provides a solid electrolyte interface analysis method, the method comprising the following steps: Acquire a solid electrolyte interface image; input the solid electrolyte interface image into a preset instance segmentation model for instance segmentation to obtain multiple lattice regions, and classify each of the lattice regions to obtain composition information corresponding to each of the lattice regions; complete solid electrolyte interface analysis according to the composition information corresponding to each of the lattice regions.

[0007] It should be noted that compared with electron microscopy analysis, this scheme distinguishes lattice regions and non-lattice regions by segmenting the solid electrolyte interface image instance, and then classifies the various partitions in the lattice region to quickly obtain its composition. Due to the low efficiency of traditional electron microscopy analysis, only static analysis can be performed. The present scheme can directly determine the composition of the lattice region in the solid electrolyte interface through an end-to-end model, batch process images, and be applied in the video analysis process. After batch processing of each frame of the image, the evolution process of the continuous lattice region image can be obtained, thereby improving the analysis efficiency.

[0008] In some embodiments, the preset instance segmentation model is trained in the following manner, including: obtaining image samples of the solid electrolyte interface, the image samples including interface images and lattice region annotation information corresponding to each of the interface images; inputting the interface images into the initial instance segmentation model to obtain a prediction result of at least one lattice region; determining a loss value based on the prediction result of the lattice region and the corresponding lattice region annotation information, and adjusting the initial instance segmentation model based on the loss value until the prediction result converges to obtain a preset instance segmentation model.

[0009] It should be noted that a preset instance segmentation model can be obtained through a certain training process. The prediction result of at least one lattice area is obtained by inputting the interface image into the initial instance segmentation model, and the preset instance segmentation model is obtained by continuously adjusting the model, thereby improving the analysis efficiency.

[0010] In some embodiments, before obtaining the image sample of the solid electrolyte interface, it also includes: obtaining the original solid electrolyte interface image; inputting the original solid electrolyte interface image into a preset semantic segmentation model to determine the initial lattice region in the original solid electrolyte interface image; performing segmentation and boundary adjustment according to the initial lattice region to obtain multiple target lattice regions; inputting each of the target lattice regions into a preset classification model to determine the corresponding annotation information; and determining the image sample of the solid electrolyte interface according to the original solid electrolyte interface image and the annotation information corresponding to the target lattice region.

[0011] It is understandable that a model with superior performance can be obtained through reasonable training. This solution uses two models to assist in labeling to achieve rapid model labeling while ensuring the efficiency of generating training samples.

[0012] In some embodiments, a lattice region to be analyzed is obtained; the crystal plane spacing corresponding to the lattice region to be analyzed is determined through the lattice region to be analyzed; composition information of the lattice region to be analyzed is determined by matching the crystal plane spacing; an initial classification model is trained based on the lattice region to be analyzed and the composition information of the lattice region to be analyzed to obtain a preset classification model.

[0013] It should be noted that how to accurately define the classification rules of components determines the accuracy of the prediction of the preset classification model. Therefore, this embodiment predicts the type by determining the crystal plane spacing corresponding to the lattice area to be analyzed. This is because the lattice spacing calculation is relatively convenient and has high computational efficiency. It is suitable for batch operations and has a good characterization effect on different components, thereby improving the efficiency and convenience of sample label generation.

[0014] In some embodiments, the lattice region to be analyzed is analyzed to obtain a spectrum image corresponding to the lattice region to be analyzed; multiple target spot features in the spectrum image are identified; multiple spot spacings are determined based on the spot features; and the crystal plane spacings corresponding to the lattice region to be analyzed are determined based on the multiple spot spacings.

[0015] It can be understood that the calculation of the lattice spacing is through the identification of some target spot features in the lattice area. The spacing between spots represents the distance of the lattice. The actual lattice spacing can be estimated through certain conversions, and the determination of the lattice spacing is completed efficiently and quickly.

[0016] In some embodiments, the crystal plane angle corresponding to the lattice spacing is determined according to the lattice region to be analyzed; the corresponding crystal plane index is matched according to the lattice spacing and the corresponding crystal plane angle; verification is performed according to the crystal plane index, and the composition information of the lattice region to be analyzed is determined according to the verification result.

[0017] It should be noted that since the accuracy of matching composition information with lattice spacing may be average, this embodiment proposes to determine the crystal plane angle, determine the composition information based on the crystal plane angle and the lattice spacing, and verify it based on the crystal plane index corresponding to the composition information, thereby improving the accuracy of composition determination.

[0018] In some embodiments, classifying each of the lattice regions to obtain the composition information corresponding to each of the lattice regions includes: determining a feature vector of each of the lattice regions, and inputting the feature vector into a classifier to determine a category score corresponding to the feature vector; determining a composition type of the lattice region and probability information corresponding to the composition type according to the category score; and determining the composition information corresponding to the lattice region according to the composition type and the probability information corresponding to the composition type.

[0019] It should be noted that since the composition information can be used in the video analysis process, in addition to the specific composition content, this scheme also provides composition probability information. The advantage of proposing probability information is that by continuously playing images of different frames, the change in probability can be seen, which provides a reference for the formation, disappearance and change of the lattice area, thereby improving the analysis effect.

[0020] In a second aspect, in order to achieve the above-mentioned object, the present invention further provides a solid electrolyte interface analysis device, the solid electrolyte interface analysis device comprising: In the third aspect, in order to achieve the above-mentioned purpose, the present invention also provides a solid electrolyte interface analysis device, which includes: a memory, a processor, and a solid electrolyte interface analysis program stored in the memory and executable on the processor, and the solid electrolyte interface analysis program is configured to implement the steps of the solid electrolyte interface analysis method described above.

[0021] In a fourth aspect, in order to achieve the above-mentioned purpose, the present invention also provides a storage medium, on which a solid electrolyte interface analysis program is stored, and when the solid electrolyte interface analysis program is executed by a processor, the steps of the solid electrolyte interface analysis method described above are implemented. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 It is a schematic diagram of the process of the first embodiment of the solid electrolyte interface analysis method of the present invention; Figure 2 It is a schematic flow chart of a second embodiment of the solid electrolyte interface analysis method of the present invention; Figure 3 A schematic diagram of model prediction of an embodiment of a solid electrolyte interface analysis method of the present invention; Figure 4 A schematic diagram showing the comparison between the original image and the frequency domain image of an embodiment of the solid electrolyte interface analysis method of the present invention; Figure 5 A schematic diagram of spot spacing of an embodiment of a solid electrolyte interface analysis method of the present invention; Figure 6 A schematic diagram of a lattice composition database of an embodiment of a solid electrolyte interface analysis method of the present invention; Figure 7 A schematic diagram for verifying an embodiment of a solid electrolyte interface analysis method of the present invention; Figure 8 It is a structural block diagram of the first embodiment of the solid electrolyte interface analysis device of the present invention; The realization of the purpose, functional features and advantages of the present invention will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0023] The following embodiments of the technical solution of the present application are described in detail in conjunction with the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solution of the present application, and are therefore only used as examples, and cannot be used to limit the scope of protection of the present application.

[0024] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by technicians in the technical field to which this application belongs; the terms used herein are only for the purpose of describing specific embodiments and are not intended to limit this application; the terms "including" and "having" in the specification and claims of this application and the above-mentioned figure descriptions and any variations thereof are intended to cover non-exclusive inclusions.

[0025] In the description of the embodiments of the present application, the technical terms "first", "second", etc. are only used to distinguish different objects, and cannot be understood as indicating or implying relative importance or implicitly indicating the number, specific order or primary and secondary relationship of the indicated technical features. In the description of the embodiments of the present application, the meaning of "multiple" is more than two, unless otherwise clearly and specifically defined.

[0026] Reference to "embodiments" herein means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present application. The appearance of the phrase in various locations in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0027] In the description of the embodiments of the present application, the term "and / or" is only a description of the association relationship of the associated objects, indicating that there may be three relationships. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone. In addition, the character " / " in this article generally indicates that the associated objects before and after are in an "or" relationship.

[0028] In the description of the embodiments of the present application, the term "multiple" refers to more than two (including two). Similarly, "multiple groups" refers to more than two groups (including two groups), and "multiple pieces" refers to more than two pieces (including two pieces).

[0029] In the description of the embodiments of the present application, the technical terms "center", "longitudinal", "lateral", "length", "width", "thickness", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside", "clockwise", "counterclockwise", "axial", "radial", "circumferential", etc. indicate orientations or positional relationships based on the orientations or positional relationships shown in the accompanying drawings, which are only for the convenience of describing the embodiments of the present application and simplifying the description, and do not indicate or imply that the referred device or element must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be understood as a limitation on the embodiments of the present application.

[0030] In the description of the embodiments of the present application, unless otherwise clearly specified and limited, technical terms such as "installed", "connected", "connected", "fixed" and the like should be understood in a broad sense. For example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection or an indirect connection through an intermediate medium, and it can be the internal connection of two elements or the interaction relationship between two elements. For those of ordinary skill in the art, the specific meanings of the above terms in the embodiments of the present application can be understood according to the specific circumstances.

[0031] The contents disclosed in the embodiments of the present application are mainly applied to the composition analysis process of the SEI film. The SEI film, which is called the Solid Electrolyte Interface, is a key protective layer formed in lithium batteries. It is usually formed spontaneously on the surface of the negative electrode when the battery is charged for the first time. The main components of the SEI film are various organic and inorganic compounds produced by the reduction reaction of the solvent and lithium salt in the electrolyte on the surface of the negative electrode material. Since the composition of the SEI film directly affects the performance of the electrolyte in the battery, specifically, during the first charge and discharge process of the liquid lithium-ion battery, the electrode material reacts with the electrolyte at the solid-liquid interface to form a passivation layer covering the surface of the electrode material. This passivation layer is an interface layer with the characteristics of a solid electrolyte. It is an electronic insulator but an excellent conductor of Li+. Li+ can be freely embedded and removed through the passivation layer. When the battery is charged for the first time, the potential on the surface of the negative electrode drops to a certain extent, causing the solvent molecules or lithium salts in the electrolyte to undergo a reduction reaction. These reaction products are deposited on the surface of the negative electrode. These decomposition products gradually accumulate to form a solid film layer covering the negative electrode active material. This film layer has a certain conductivity, allowing lithium ions to pass through but blocking electrons from passing through, thus preventing further direct electrochemical reactions. Therefore, this passivation film is called a "solid electrolyte interface" (SEI film for short). In the SEI film image taken by an electron microscope, the lattice areas can be seen by the naked eye. These lattice areas are material areas of certain components. How to identify these lattice areas is crucial to the analysis of SEI films. The analysis of the lattice areas in the SEI film not only affects the basic electrochemical performance of the battery, but also affects the safety, reliability and service life of the battery, and plays an important guiding role in optimizing electrolytes and related materials.

[0032] The lattice region in the SEI film refers to the part with an ordered structure in the SEI film. These regions may be composed of inorganic compounds, which are usually more stable than the amorphous part. The ordered lattice structure is often more stable than the amorphous structure, so the lattice region helps to improve the stability of the entire SEI film.

[0033] Therefore, this scheme proposes to automatically segment the representation image of the SEI film by training the instance segmentation model, and calculate the area of ​​the instance segmentation region and other information by digital image processing methods to achieve automatic and efficient analysis. This scheme proposes a training method for the SEI component instance segmentation model, which can effectively train an end-to-end machine learning model for analyzing SEI components and assist in the study of SEI films.

[0034] According to some embodiments of the present application, Figure 1 A solid electrolyte interface analysis method is provided, the method comprising the following steps: Acquire a solid electrolyte interface image; input the solid electrolyte interface image into a preset instance segmentation model for instance segmentation to obtain multiple lattice regions, and classify each of the lattice regions to obtain composition information corresponding to each of the lattice regions; complete solid electrolyte interface analysis according to the composition information corresponding to each of the lattice regions.

[0035] It can be understood that the solid electrolyte interface image is the image content of the solid electrolyte interface, which can be obtained through an electron microscope. In this embodiment, what is collected is generally a video image. The video image is decomposed into several frames of images and input into a preset instance segmentation model to obtain a video image containing lattice region composition information, so as to help technicians complete the analysis of the solid electrolyte interface. Specifically, the composition information may include the composition of the lattice region, the probability of the composition, or several possible components, etc.

[0036] In addition, in order to conduct a detailed analysis of the solid electrolyte interface (SEI), high-resolution images of the SEI can be first obtained through microscopic imaging technology. Subsequently, these images are input into a pre-trained instance segmentation model that can identify and segment multiple lattice regions in the image. By further classifying each segmented lattice region, the specific component information corresponding to each region is determined. Finally, based on this component information, a comprehensive analysis of the SEI film is completed, including its structural characteristics, component distribution, and possible functional characteristics, which helps to deeply understand the mechanism of action of the SEI film in the battery and also provides a scientific basis for optimizing battery performance. Finally, the frames are reassembled into a video, and the evolution process of the lattice region can be clearly observed, further improving the analysis effect.

[0037] Among them, the specific prediction process of the model can be: input the solid electrolyte interface image into a preset instance segmentation model, perform image feature extraction to segment the image into patches of fixed size, and linearly map these patches into vectors, and then generate feature representations rich in semantic information through a multi-layer Transformer encoder. Next, it is necessary to process multi-target recognition and classification. Through an encoder-decoder structure, the encoder is responsible for extracting global context information from the image and introducing a set of learnable queries. Each query represents a potential target instance. These queries are input to the decoder and are gradually refined by the decoder to finally generate the corresponding target box, which is equivalent to an instance of a lattice area. The classifier is then used to predict the category label corresponding to each query to obtain the component information, such as a fully connected layer plus a softmax activation function.

[0038] It should be noted that compared with electron microscopy analysis, this scheme distinguishes lattice regions and non-lattice regions by segmenting the solid electrolyte interface image instance, and then classifies the various partitions in the lattice region to quickly obtain its composition. Due to the low efficiency of traditional electron microscopy analysis, only static analysis can be performed. The present scheme can directly determine the composition of the lattice region in the solid electrolyte interface through an end-to-end model, batch process images, and be applied in the video analysis process. After batch processing of each frame of the image, the evolution process of the continuous lattice region image can be obtained, thereby improving the analysis efficiency.

[0039] In some embodiments, Figure 2 As shown, the preset instance segmentation model is trained in the following manner, including: obtaining image samples of the solid electrolyte interface, the image samples including the interface image and the lattice region annotation information corresponding to each of the interface images; inputting the interface image into the initial instance segmentation model to obtain a prediction result of at least one lattice region; determining a loss value based on the prediction result of the lattice region and the corresponding lattice region annotation information, and adjusting the initial instance segmentation model based on the loss value until the prediction result converges to obtain the preset instance segmentation model.

[0040] It can be understood that the specific training process of the preset instance segmentation model is as follows: First, it is necessary to obtain high-quality image samples containing solid electrolyte interfaces. These images can be obtained by electron microscopy to collect SEI films in experiments, SEI film formation processes in generation, or simulations, showing the microstructure of SEI. Each image is equipped with detailed annotation information, which clearly indicates the location and boundaries of each lattice region in the image. Then, these images with annotation information are input into a pre-designed but not yet fully trained instance segmentation model. The model attempts to perform instance segmentation on the input image to determine one or more lattice regions, and then predicts each lattice region to obtain a high-dimensional vector representing the type of lattice component, and the output result is a preliminary prediction of the lattice region. Next, the prediction results of the model are compared with the real lattice region information manually annotated. By calculating the difference between the two, a loss value can be obtained, which quantifies the accuracy of the model prediction. According to this loss value, the model parameters are adjusted using an optimization algorithm (such as gradient descent) so that the model can more accurately identify the lattice region in the next prediction. The above process is not completed in one go, but the model is gradually improved through multiple iterations. After each iteration, the loss value is recalculated based on the new prediction results, and the model is further adjusted accordingly. This process continues until the model's prediction results no longer change significantly, that is, convergence, or the preset performance indicators are reached. After multiple rounds of training and optimization, when the model's performance is stable and meets the expected requirements, a preset instance segmentation model is obtained. This end-to-end model can directly, automatically and efficiently identify and annotate lattice regions from new solid electrolyte interface images without human intervention, providing strong support for materials science research.

[0041] It should be noted that a preset instance segmentation model can be obtained through a certain training process. The prediction result of at least one lattice area is obtained by inputting the interface image into the initial instance segmentation model, and the preset instance segmentation model is obtained by continuously adjusting the model, thereby improving the analysis efficiency.

[0042] In some embodiments, before obtaining the image sample of the solid electrolyte interface, it also includes: obtaining the original solid electrolyte interface image; inputting the original solid electrolyte interface image into a preset semantic segmentation model to determine the initial lattice region in the original solid electrolyte interface image; performing segmentation and boundary adjustment according to the initial lattice region to obtain multiple target lattice regions; inputting each of the target lattice regions into a preset classification model to determine the corresponding annotation information; and determining the image sample of the solid electrolyte interface according to the original solid electrolyte interface image and the annotation information corresponding to the target lattice region.

[0043] It should be noted that there is a problem in the actual training process, that is, how to obtain a large number of refined samples. This embodiment proposes to complete sample annotation by training two lightweight models. For example, a lattice semantic segmentation model of SEI image is trained to segment the lattice region. For SEI electron microscope images, we manually annotated a batch of lattice region image data using the naked eye judgment method, and then generated a Figure 1 A two-dimensional matrix of the same size is used, the corresponding lattice area is set to 1, and the non-lattice area is set to 0. The annotated image and the processed 0 1 matrix are used as the input and output of the model. The large image model is used for fine-tuning training, and finally a semantic segmentation model for the lattice is trained. A classification model for different lattice images is trained to classify lattice images of different components. The above-predicted lattice area image is cut into images of 256 or 512 sizes, and then these images are Fourier transformed, and the intercept and other information of the lattice are calculated in the reciprocal space, so as to infer the corresponding components of the lattice. In this way, the data of the corresponding components of the lattice image are annotated, and the corresponding components are one-hot encoded. The image classification model is trained using the cut lattice image and the corresponding one-hot encoding, and finally a classification model for the lattice image is obtained. Combining the above two models, the predicted lattice area image is cut into small pieces of images, and the cut images are classified and predicted, and then the results are mapped back to the original image and the annotation is fine-tuned. Use the trained semantic segmentation model to perform lattice segmentation on the SEI image, then cut the segmented lattice region image and input it into the classification model for classification prediction. Finally, map each predicted classification into a different color mask and map it back to the original image. Finally, get an instance segmentation mask of different component lattices, and then perform human correction. Use the fine-tuned annotated data to train an instance segmentation model for component regions. Use the instance segmentation masks of different component lattices as the input and output of the model to train an end-to-end component region instance segmentation model.

[0044] It is understandable that a model with superior performance can be obtained through reasonable training. This solution uses two models to assist in labeling to achieve rapid model labeling while ensuring the efficiency of generating training samples.

[0045] In some embodiments, a lattice region to be analyzed is obtained; the crystal plane spacing corresponding to the lattice region to be analyzed is determined through the lattice region to be analyzed; composition information of the lattice region to be analyzed is determined by matching the crystal plane spacing; an initial classification model is trained based on the lattice region to be analyzed and the composition information of the lattice region to be analyzed to obtain a preset classification model.

[0046] It can be understood that the interplanar spacing refers to the vertical distance between two adjacent parallel crystal planes in the crystal structure. By inferring the composition based on information such as the lattice spacing, the composition information can be obtained more efficiently. The specific matching process can match the lattice spacing with the lattice database corresponding to each component to determine the possible components in the lattice area. The inferred composition information is used as the label information corresponding to the lattice area, and the initial classification model is trained with the lattice area to be analyzed and the composition information of the lattice area to be analyzed to obtain a mature and usable preset classification model as the basis for batch labeling of image samples of solid electrolyte interfaces.

[0047] It should be noted that how to accurately define the classification rules of components determines the accuracy of the prediction of the preset classification model. Therefore, this embodiment predicts the type by determining the crystal plane spacing corresponding to the lattice area to be analyzed. This is because the lattice spacing calculation is relatively convenient and has high computational efficiency. It is suitable for batch operations and has a good characterization effect on different components, thereby improving the efficiency and convenience of sample label generation.

[0048] In some embodiments, the lattice region to be analyzed is analyzed to obtain a spectrum image corresponding to the lattice region to be analyzed; multiple target spot features in the spectrum image are identified; multiple spot spacings are determined based on the spot features; and the crystal plane spacings corresponding to the lattice region to be analyzed are determined based on the multiple spot spacings.

[0049] It should be noted that the prediction process of the interplanar spacing can be, for example, to obtain the spectrum corresponding to the original image through Fourier transformation, such as Figure 4 The spectrum image from the left "lattice area to be analyzed" to the right. It can be seen that the spectrum image on the right contains spots, among which multiple target spot features are Figure 4 The spot image in the middle right image can determine the spot spacing between spots by measuring and calculating the spacing of the spot images. By statistically analyzing multiple spot spacings, the crystal plane spacing corresponding to the lattice area to be analyzed can be predicted.

[0050] Specifically, Figure 5 As shown, corresponding to the 7 spots in the spectrum image, four of them are selected, and a set of parallelograms corresponding to the spots in the spectrum image ( Figure 5 ) calibrate the lengths of l1, l2, l3 and measure the corresponding three angles, and calculate the interplanar spacings d1, d2, d3 according to the formula; taking d1 as an example, the formula is as follows:

[0051] Where l0 is the line segment in the figure, and the scale is the actual size of the line segment in the figure, i.e. 5 / nm. By conversion, the actual length of l1, l2, l3, i.e. the spot spacing, can be obtained, and the crystal plane spacing d1, d2, d3 can be calculated. It can be understood that the calculation of the lattice spacing is through the identification of some target spot features in the lattice area. The spacing between spots represents the distance of the lattice. The actual lattice spacing can be estimated through certain conversions, and the determination of the lattice spacing is completed efficiently and quickly.

[0052] In some embodiments, the crystal plane angle corresponding to the lattice spacing is determined according to the lattice region to be analyzed; the corresponding crystal plane index is matched according to the lattice spacing and the corresponding crystal plane angle; verification is performed according to the crystal plane index, and the composition information of the lattice region to be analyzed is determined according to the verification result.

[0053] Specifically, although it is highly efficient to determine the composition information of the lattice region to be analyzed only by matching the crystal plane spacing, the accuracy may be average. Therefore, this embodiment further proposes to calculate the crystal plane angle, where the crystal plane angle refers to the angle between two different crystal plane normals in the crystal, and match the composition information through the lattice spacing and the corresponding crystal plane angle, such as Figure 6 As shown, the closest composition is obtained by querying the data based on the lattice spacing and the corresponding crystal plane angle.

[0054] Furthermore, in order to filter out invalid information, verification can also be performed based on the crystal face index, such as Figure 6 As shown in the figure, the crystal plane index can also be obtained by querying the lattice database corresponding to each component. The crystal plane index usually refers to the Miller indices, which is a standard representation used in crystallography to identify different planes inside a crystal. Figure 6 The database data is compared and three matching lattice indices R1, R2, and R3 are selected. If R1, R2, and R3 conform to the vector addition law, the component corresponding to the image can be marked as valid component information. Figure 7 shown.

[0055] It should be noted that since the accuracy of matching composition information with lattice spacing may be average, this embodiment proposes to determine the crystal plane angle, determine the composition information based on the crystal plane angle and the lattice spacing, and verify it based on the crystal plane index corresponding to the composition information, thereby improving the accuracy of composition determination.

[0056] In some embodiments, classifying each of the lattice regions to obtain the composition information corresponding to each of the lattice regions includes: determining a feature vector of each of the lattice regions, and inputting the feature vector into a classifier to determine a category score corresponding to the feature vector; determining a composition type of the lattice region and probability information corresponding to the composition type according to the category score; and determining the composition information corresponding to the lattice region according to the composition type and the probability information corresponding to the composition type.

[0057] It should be noted that after obtaining the preset instance segmentation model, for each predicted lattice region, its feature vector needs to be further extracted. These features can be based on image attributes such as shape, texture, color, position, etc., or they can be high-level features automatically learned from the image through a convolutional neural network (CNN). The feature vector is a mathematical representation of the intrinsic characteristics of the lattice region, which can capture the differences between different materials. Once the feature vector of the lattice region is obtained, it can be classified in the classifier part of the model. The task of the classifier is to predict which type of component each lattice region belongs to based on the feature vector. Commonly used classifiers include support vector machines (SVM), random forests, neural networks, etc. In this case, the classifier will output a series of category scores for each lattice region, and each score corresponds to the probability that the lattice region may belong to a specific component type. For example: If there are three possible component types A, B and C, then the classifier will output three scores for each lattice region, representing the probability that it belongs to A, B or C respectively. The purpose of obtaining probability information is that general analysis is based on the analysis process of video images, and the evolution process of different lattice regions can be further analyzed through probability, for example Figure 3 shown.

[0058] It should be noted that since the composition information can be used in the video analysis process, in addition to the specific composition content, this scheme also provides composition probability information. The advantage of proposing probability information is that by continuously playing images of different frames, the change in probability can be seen, which provides a reference for the formation, disappearance and change of the lattice area, thereby improving the analysis effect.

[0059] Second, to achieve the above objectives, Figure 8 As shown, the present invention also provides a solid electrolyte interface analysis device, characterized in that the solid electrolyte interface analysis device includes: an acquisition module 10, used to acquire a solid electrolyte interface image; a processing module 20, used to input the solid electrolyte interface image into a preset instance segmentation model for instance segmentation to obtain a plurality of lattice regions, and classify each of the lattice regions to obtain composition information corresponding to each of the lattice regions; the processing module 20 is used to complete the solid electrolyte interface analysis according to the composition information corresponding to each of the lattice regions.

[0060] In some embodiments, the processing module 20 is used to obtain image samples of the solid electrolyte interface, wherein the image samples include interface images and lattice region annotation information corresponding to each of the interface images; Inputting the interface image into an initial instance segmentation model to obtain a prediction result of at least one lattice region; A loss value is determined according to the prediction result of the lattice region and the corresponding lattice region annotation information, and the initial instance segmentation model is adjusted according to the loss value until the prediction result converges to obtain a preset instance segmentation model.

[0061] In some embodiments, the processing module 20 is used to obtain an original solid electrolyte interface image; Inputting the original solid electrolyte interface image into a preset semantic segmentation model to determine an initial lattice region in the original solid electrolyte interface image; Segmenting and adjusting the boundaries of the initial lattice region to obtain a plurality of target lattice regions; Inputting each of the target lattice regions into a preset classification model to determine corresponding annotation information; An image sample of the solid electrolyte interface is determined according to the original solid electrolyte interface image and the annotation information corresponding to the target lattice region.

[0062] In some embodiments, the processing module 20 is used to obtain a lattice region to be analyzed; Determine the interplanar spacing corresponding to the crystal lattice region to be analyzed by the crystal lattice region to be analyzed; Matching according to the interplanar spacing determines the composition information of the lattice region to be analyzed; The initial classification model is trained according to the lattice region to be analyzed and the composition information of the lattice region to be analyzed to obtain a preset classification model.

[0063] In some embodiments, the processing module 20 is used to analyze the lattice region to be analyzed to obtain a spectrum image corresponding to the lattice region to be analyzed; identifying a plurality of target spot features in the spectrum image; Determining a plurality of spot spacings according to the spot characteristics; The crystal plane spacing corresponding to the lattice region to be analyzed is determined according to the plurality of spot spacings.

[0064] In some embodiments, the processing module 20 is used to determine the crystal plane angle corresponding to the lattice spacing according to the lattice region to be analyzed; Matching corresponding crystal plane indices according to the lattice spacing and the corresponding crystal plane angles; Verification is performed according to the crystal plane index, and composition information of the lattice region to be analyzed is determined according to the verification result.

[0065] In some embodiments, the processing module 20 is used to determine a feature vector of each of the lattice regions, and input the feature vector into a classifier to determine a category score corresponding to the feature vector; Determine the component type of the lattice region and probability information corresponding to the component type according to the category score; The component information corresponding to the lattice region is determined according to the component type and the probability information corresponding to the component type.

[0066] In the third aspect, in order to achieve the above-mentioned purpose, the present invention also provides a solid electrolyte interface analysis device, which includes: a memory, a processor, and a solid electrolyte interface analysis program stored in the memory and executable on the processor, and the solid electrolyte interface analysis program is configured to implement the steps of the solid electrolyte interface analysis method described above.

[0067] In a fourth aspect, in order to achieve the above-mentioned purpose, the present invention also provides a storage medium, on which a solid electrolyte interface analysis program is stored, and when the solid electrolyte interface analysis program is executed by a processor, the steps of the solid electrolyte interface analysis method described above are implemented.

[0068] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them; although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein by equivalents; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the embodiments of the present application, and they should all be included in the scope of the claims and specification of the present application. In particular, as long as there is no structural conflict, the various technical features mentioned in the various embodiments can be combined in any way. The present application is not limited to the specific embodiments disclosed herein, but includes all technical solutions that fall within the scope of the claims.

Claims

1. A solid electrolyte interface analysis method, characterized in that: The solid electrolyte interface analysis method comprises: Acquire solid electrolyte interface images; Inputting the solid electrolyte interface image into a preset instance segmentation model for instance segmentation to obtain a plurality of lattice regions, and classifying each of the lattice regions to obtain composition information corresponding to each of the lattice regions; The solid electrolyte interface analysis is completed according to the composition information corresponding to each of the lattice regions.

2. The method according to claim 1, characterized in that The preset instance segmentation model is trained in the following manner, including: Acquire image samples of the solid electrolyte interface, wherein the image samples include interface images and lattice region annotation information corresponding to each of the interface images; Inputting the interface image into an initial instance segmentation model to obtain a prediction result of at least one lattice region; A loss value is determined according to the prediction result of the lattice region and the corresponding lattice region annotation information, and the initial instance segmentation model is adjusted according to the loss value until the prediction result converges to obtain a preset instance segmentation model.

3. The method according to claim 2, characterized in that Before acquiring the image sample of the solid electrolyte interface, the method further includes: Acquire pristine solid electrolyte interface images; Inputting the original solid electrolyte interface image into a preset semantic segmentation model to determine an initial lattice region in the original solid electrolyte interface image; Segmenting and adjusting the boundaries of the initial lattice region to obtain a plurality of target lattice regions; Inputting each of the target lattice regions into a preset classification model to determine corresponding annotation information; An image sample of the solid electrolyte interface is determined according to the original solid electrolyte interface image and the annotation information corresponding to the target lattice region.

4. The method according to claim 3, characterized in that The preset classification model is trained in the following manner, including: Obtaining a lattice region to be analyzed; Determine the interplanar spacing corresponding to the crystal lattice region to be analyzed by the crystal lattice region to be analyzed; Matching according to the interplanar spacing determines the composition information of the lattice region to be analyzed; The initial classification model is trained according to the lattice region to be analyzed and the composition information of the lattice region to be analyzed to obtain a preset classification model.

5. The method according to claim 4, characterized in that The step of determining the interplanar spacing corresponding to the lattice region to be analyzed by using the lattice region to be analyzed includes: Analyzing the lattice region to be analyzed to obtain a spectrum image corresponding to the lattice region to be analyzed; identifying a plurality of target spot features in the spectrum image; Determining a plurality of spot spacings according to the spot characteristics; The crystal plane spacing corresponding to the lattice region to be analyzed is determined according to the plurality of spot spacings.

6. The method according to claim 4, characterized in that The determining the composition information of the lattice region to be analyzed by matching according to the crystal plane spacing includes: Determine the crystal plane angle corresponding to the lattice spacing according to the lattice region to be analyzed; Matching corresponding crystal plane indices according to the lattice spacing and the corresponding crystal plane angles; Verification is performed according to the crystal plane index, and composition information of the lattice region to be analyzed is determined according to the verification result.

7. The method according to claim 1, characterized in that The step of classifying the lattice regions to obtain the component information corresponding to the lattice regions includes: Determining a feature vector of each of the lattice regions, and inputting the feature vector into a classifier to determine a category score corresponding to the feature vector; Determine the component type of the lattice region and probability information corresponding to the component type according to the category score; The component information corresponding to the lattice region is determined according to the component type and the probability information corresponding to the component type.

8. A solid electrolyte interface analysis device, characterized in that: The solid electrolyte interface analysis device comprises: An acquisition module, used for acquiring a solid electrolyte interface image; A processing module, used for inputting the solid electrolyte interface image into a preset instance segmentation model for instance segmentation to obtain a plurality of lattice regions, and classifying each of the lattice regions to obtain composition information corresponding to each of the lattice regions; The processing module is used to complete the solid electrolyte interface analysis according to the composition information corresponding to each of the lattice regions.

9. A solid electrolyte interface analysis device, characterized in that: The device comprises: a memory, a processor, and a solid electrolyte interface analysis program stored in the memory and executable on the processor, wherein the solid electrolyte interface analysis program is configured to implement the steps of the solid electrolyte interface analysis method according to any one of claims 1 to 7.

10. A storage medium, characterized in that: The storage medium stores a solid electrolyte interface analysis program, and when the solid electrolyte interface analysis program is executed by the processor, the steps of the solid electrolyte interface analysis method according to any one of claims 1 to 7 are implemented.

Images