WO2023125111A1

WO2023125111A1 - Material identification method and apparatus based on artificial intelligence atomic force microscope

Info

Publication number: WO2023125111A1
Application number: PCT/CN2022/140068
Authority: WO
Inventors: 黄博远; 谭州瑜; 朱庆丰; 李江宇
Original assignee: 中国科学院深圳先进技术研究院
Priority date: 2021-12-27
Filing date: 2022-12-19
Publication date: 2023-07-06
Also published as: CN116402736A

Abstract

Disclosed in embodiments of the present invention are a material identification method and apparatus based on an artificial intelligence atomic force microscope. The method comprises: obtaining a microscope scanning image; inputting the microscope scanning image into a pre-trained material identification model to obtain a classification result output by an image classification model, wherein the material identification model is obtained by training based on simulation sample images; and determining a target material corresponding to the microscope scanning image according to the classification result. The embodiments of the present invention provide a material identification model which has a simple structure and a wide applicability, thereby improving material identification efficiency and accuracy.

Description

Material identification method and device based on artificial intelligence atomic force microscope

technical field

Embodiments of the present invention relate to the field of image technology, and in particular, to a material identification method and device based on an artificial intelligence atomic force microscope.

Background technique

Atomic Force Microscope (AFM) is a powerful tool for probing, elucidating and manipulating materials and structures at the nanoscale.

In the process of realizing the present invention, the inventor found that there are at least the following technical problems in the prior art: the operation of the current atomic force microscope AFM relies heavily on the subjective experience of the operator, and the user often ignores important but subtle information during operation. Data processing finds out when it's too late.

technical problem

Embodiments of the present invention provide a material identification method and device based on an artificial intelligence atomic force microscope, so as to avoid the influence of subjective operations on the topography image obtained by the atomic force microscope and obtain effective topography images.

technical solution

In the first aspect, an embodiment of the present invention provides a material identification method based on an artificial intelligence atomic force microscope, including:

Obtain a microscope scan image;

Inputting the microscope scanning image into the pre-trained material recognition model to obtain the classification result output by the image classification model, wherein the material recognition model is trained based on the simulation sample image;

The target material corresponding to the microscope scanning image is determined according to the classification result.

Optionally, further, the material recognition model includes a feature extraction module and a classification module, input the microscope scanning image into the pre-trained material recognition model, and obtain the classification result output by the material recognition model, including:

Inputting the microscope scanning image into the feature extraction module to obtain the structural features output by the feature extraction module;

The structural features are input into the classification module, and the classification results output by the classification module are obtained.

Optionally, further, the target material corresponding to the microscope scanning image is determined according to the classification result, including:

Verification of classification results based on target materials;

When the classification result is verified to pass the verification, the material corresponding to the classification result is taken as the target material.

Optionally, further, verification of classification results based on target materials, including:

When the classification result is ferroelectric, adjust the scanning area to obtain domain wall information of the scanning object;

Based on the domain wall information, the switching spectroscopy piezoelectric response force microscopy experiment is performed. When the loop parameters corresponding to the scanning object are generated, the classification result of the ferroelectric category is verified to pass.

When the classification result is a non-ferroelectric category, adjust the scanning area to obtain the grain boundary information of the scanning object;

Perform first harmonic piezoelectric response and second harmonic piezoelectric response based on grain boundary information;

When the second harmonic piezoelectric response dominates the first harmonic piezoelectric response, it is determined that the classification results of the non-ferroelectric category pass the verification.

Optionally, further, the training of the material recognition model includes:

Determine the sample image category according to the classification category corresponding to the material recognition model;

performing simulation based on the category of the sample image to obtain a simulation sample image;

A model training sample is constructed based on the sample image category and the simulation sample image, and the pre-built material recognition model is trained by using the model training sample to obtain a trained material recognition model.

Optionally, further, the material identification model is constructed based on a support vector machine model.

In the second aspect, the embodiment of the present invention also provides a material identification device based on an artificial intelligence atomic force microscope, including:

A scan image acquisition module, configured to acquire a microscope scan image;

The scanning image classification module is used to input the microscope scanning image into the pre-trained material recognition model, and obtain the classification result output by the image classification model, wherein the material recognition model is obtained based on the simulation sample image training;

The target material determination module is configured to determine the target material corresponding to the microscope scanning image according to the classification result.

In a third aspect, the embodiment of the present invention also provides a computer device, which includes:

one or more processors;

storage means for storing one or more programs;

When one or more programs are executed by one or more processors, the one or more processors implement the material identification method based on artificial intelligence atomic force microscope provided in any embodiment of the present invention.

In the fourth aspect, the embodiment of the present invention also provides a computer-readable storage medium, on which a computer program is stored, and when the program is executed by a processor, the material based on the artificial intelligence atomic force microscope as provided in any embodiment of the present invention is realized. recognition methods.

Beneficial effect

In the embodiment of the present invention, the microscope scanning image of the current scanning area is obtained; the microscope scanning image is input into the pre-trained material recognition model to obtain the classification result output by the material recognition model; and the target topography image is determined according to the classification result. The validity of the microscope scanning image is judged by the pre-trained material recognition model to classify the topography image, and the confidence level of the target topography image is determined according to the effectiveness of the microscope scanning image, which avoids the influence of subjective operations on the wrong judgment of the topography image and improves In order to obtain the accuracy of the real topography image.

Description of drawings

Fig. 1 is a flowchart of a material identification method based on an artificial intelligence atomic force microscope provided in Embodiment 1 of the present invention;

Fig. 2 is a schematic structural diagram of a material identification device based on an artificial intelligence atomic force microscope provided in Embodiment 3 of the present invention;

FIG. 3 is a schematic structural diagram of a computer device provided by Embodiment 4 of the present invention.

Embodiments of the present invention

The present invention will be further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described here are only used to explain the present invention, but not to limit the present invention. In addition, it should be noted that, for the convenience of description, only some structures related to the present invention are shown in the drawings but not all structures.

Embodiment one

FIG. 1 is a flowchart of a material identification method based on an artificial intelligence atomic force microscope provided in Embodiment 1 of the present invention. This embodiment is applicable to the situation of material identification of the imaging object of the atomic force microscope, especially suitable to the situation of material identification through the artificial intelligence atomic force microscope AFM. The method can be performed by a material identification device based on an artificial intelligence atomic force microscope, and the material identification device based on an artificial intelligence atomic force microscope can be implemented in the form of software and/or hardware, for example, the material identification device based on an artificial intelligence atomic force microscope can be configured In computer equipment, such as atomic force microscopes. As shown in Figure 1, the method includes:

S110. Obtain a microscope scanning image.

Aiming at the technical problems in the prior art that artificial intelligence-driven AFM has a complex structure and is only applicable to an ideal laboratory environment. The embodiment of the present invention proposes a material identification method with simple structure and wide application. In general, by pre-training the material recognition model, the scanned image of the atomic force microscope is input into the pre-trained material recognition model as a microscope scan image, and the structural features of the microscope scan image are identified and classified through the material recognition model, and finally according to the material The classification result output by the recognition model determines the material category of the scanned object corresponding to the microscope scanned image.

In this embodiment, the microscope scanning image may be the scanning image when the atomic force microscope starts to scan the scanning object, and may also be the scanning image during the scanning process. Wherein, the scanning area corresponding to the microscope scanning image can be set as a fixed value, or can be set by the user based on experience. After the atomic force microscope scans the scanning area, a microscope scanning image is generated. Wherein, the method of generating the microscope scanning image according to the scanning data of the scanning area may refer to the imaging method of the atomic force microscope in the prior art, and no limitation is made here.

S120. Input the microscope scanning image into the pre-trained material recognition model, and obtain the classification result output by the image classification model, wherein the material recognition model is trained based on the simulation sample image.

In this embodiment, in order to avoid the influence of subjective experience on the topography image, the material category of the scanned object is judged by a machine learning model. That is to say, the microscope scanning images are classified through the pre-trained machine learning model-material recognition model, and the material category corresponding to the microscope scanning images is determined according to the classification category.

Optionally, the material identification model can be constructed based on a support vector machine model. Support Vector Machine (SVM) is a generalized linear classifier that performs binary classification on data in a supervised learning manner. Its decision boundary is the maximum margin hyperplane for solving learning samples. In pattern recognition problems such as text classification. The classifier has high adaptability to the image data of this specific scene of AFM, and can realize feature recognition of AFM data images well.

Electromechanical coupling is ubiquitous in natural materials, synthetic devices, and biological systems, such as ferroelectric materials, lithium-ion batteries, and voltage ion channels, providing a wide range of applications for information processing, energy conversion, and biological processes. Despite their vast differences in their microscopic mechanisms, these electromechanical couplings often manifest as distinct piezoelectric responses in ds-SPM, and it is extremely challenging to manually discern their dominant microscopic origins. Taking the classification of ferroelectric and electrochemical materials as an example, 180° domains often exist in ferroelectric materials and usually exhibit greatly reduced piezoelectric response. On the other hand, the amplitude and phase behavior of non-ferroelectric solids, such as electrochemical materials, is usually not well defined, where the phase contrast is less than 180°. Based on the above differences, an embodiment of the present invention proposes a physics-based classifier developed using a support vector machine (SVM) algorithm, that is, a method for a material recognition model. A support vector machine model is able to extract ferroelectric domain walls pixel-by-pixel from input PFM imaging, thus helping to distinguish ferroelectric materials from electrochemical materials, where a different algorithm is introduced to extract grain boundaries from AFM topography imaging . In contrast, although the currently popular convolutional neural network (CNN) has achieved remarkable success in the field of image recognition, it can only classify the entire image and cannot accurately delineate the domain walls or grain boundaries of interest. Fully convolutional networks derived from CNNs are able to identify lattice atoms in raw scanning transmission electron microscopy (STEM) data. Accurately labeled training data pixels at the wall or grain boundary level. And SVM-based AI algorithms require only a small dataset and can be trained in less than 10 seconds on an ordinary personal computer, making them widely applicable. More importantly, this SVM-based algorithm is much more efficient than CNN in terms of classification and immediate control.

In this embodiment, when training the material recognition model, a method for training the material recognition model based on image generation model training samples obtained through simulation is provided in consideration of the complexity of labeling work. Optionally, the training of the material recognition model includes: determining the sample image category according to the classification category corresponding to the material recognition model; performing simulation based on the sample image category to obtain a simulation sample image; constructing a model training sample based on the sample image category and the simulation sample image, using The model training sample trains the pre-built material recognition model to obtain the trained material recognition model.

Still take the classification of ferroelectric materials and electrochemical materials as an example. Considering that the amplitude and phase histograms often exist at several specific angles, it is often inefficient to use real images for labeling training. When training the SVM classifier, in order to reduce the workload of labeling, the characteristics of the amplitude and phase are simulated through simulation, and white noise is added on the basis of it, thus avoiding the tedious work of labeling, and successfully training the model based on the simulation shape. The model of the map. Specifically, random binary codes of domain walls can be generated according to the material category to be classified, and then the real imaging can be simulated from the perspective of morphology, and the simulated sample image can be obtained after adding white noise. After obtaining the simulation sample image, construct a model training sample based on the simulation sample image and its corresponding classification category, use the model training sample to train the material recognition model, and obtain the trained material recognition model. For the training method of the material recognition model, reference may be made to the model training method in the prior art, which will not be repeated here.

In one embodiment, the material recognition model includes a feature extraction module and a classification module, inputting the microscope scanning image into the pre-trained material recognition model, and obtaining the classification result output by the material recognition model, including: inputting the microscope scanning image into the feature extraction module In , the structural features output by the feature extraction module are obtained; the structural features are input into the classification module, and the classification results output by the classification module are obtained.

Overall, the classification of microscope scanning images includes two parts: feature extraction and classification. Among them, the feature extraction is realized by the feature extraction module in the material recognition model, and the classification is realized by the classification module in the material recognition model. Combined with the features of the AFM data images, the material categories corresponding to the microscope scanning images are classified according to the structural features. Exemplarily, whether there is a ferroelectric domain wall can be judged by identifying the length of the longest line on the binary mask in the structural feature, otherwise it can be judged as a grain boundary.

S130. Determine the target material corresponding to the microscope scanning image according to the classification result.

In an embodiment, the material corresponding to the classification result may be directly used as the target material corresponding to the microscope scanning image. Taking the identification of ferroelectric and non-ferroelectric materials as an example, the classification results output by the material identification model can be either ferroelectric or non-ferroelectric.

In another embodiment, in order to further confirm the accuracy of material identification, after classification by the material identification model, the key points in the structural features can also be tracked, zoomed in and scanned for further verification. Based on this, the target material corresponding to the microscope scanning image is determined according to the classification result, including: verifying the classification result based on the target material; when the verification of the classification result is verified as passing, the material corresponding to the classification result is used as the target material. It can be understood that different material categories have different characteristics. Based on this, the classification result can be further verified by verifying whether the scanned object has the material characteristics of the material corresponding to the classification category.

Optionally, verify the classification result based on the target material, including: when the classification result is ferroelectric, adjust the scanning area to obtain the domain wall information of the scanned object; perform switching spectroscopy piezoelectric response force microscopy experiments based on the domain wall information, When the loop parameters corresponding to the scanning object are generated, it is determined that the classification result of the ferroelectric category has passed the verification. When a domain wall is detected, that is, when the classification result is ferroelectric, the program will automatically trigger the "ferroelectric program". The program will adjust the scanning area, move the scanning tip to the identified domain wall, and zoom in on the scan. The SS-PFM experiment is carried out on a point line of the ferroelectric material, such as hysteresis loop and butterfly loop. When the loop parameters corresponding to the scanned object are generated, it means that the scanned object has the characteristics of ferroelectric material If the material characteristics are not met, it is judged that the verification of the classification result is passed; otherwise, it is judged that the verification of the classification result is not passed.

Optionally, verify the classification result based on the target material, including: when the classification result is non-ferroelectric, adjust the scanning area to obtain the grain boundary information of the scanned object; perform the first harmonic piezoelectric response and the second order based on the grain boundary information Harmonic piezoelectric response; when the second harmonic piezoelectric response dominates the first harmonic piezoelectric response, it is determined that the classification result of the non-ferroelectric category has passed the verification. If the program does not detect the 180° domain wall from the amplitude and phase imaging, that is, when the classification result is a non-ferroelectric category, trigger the "non-ferroelectric program" to identify the grain boundary superimposed on the morphology, and zoom in on the scan, and then The first and second harmonic piezoelectric responses are performed on the grain boundaries. If the second harmonic piezoelectric response dominates the first harmonic piezoelectric response, indicating that the scanned object has the material properties of non-ferroelectric materials, it is determined that the classification result verification is passed; otherwise, it is determined that the classification result verification fails.

When the verification of the classification result fails, the microscope scanning image can be reacquired by adjusting the scanning area for material identification, and the material identification model can also be retrained to perform material identification based on the retrained material identification model.

In the embodiment of the present invention, by acquiring the microscope scanning image; inputting the microscope scanning image into the pre-trained material recognition model, the classification result output by the image classification model is obtained, wherein the material recognition model is obtained based on the simulation sample image training; the microscope is determined according to the classification result The scanned image corresponds to the target material. A material identification model with simple structure and wide applicability is provided, and the efficiency and accuracy of material identification are improved.

Embodiment two

This embodiment provides a preferred embodiment on the basis of the above solutions.

The material recognition method provided by the embodiment of the present invention classifies the scanned images for AI-AFM targeted learning, and conducts feature recognition training according to its amplitude and phase response, so that AI-AFM can distinguish its different structural features by itself , and track and zoom in on the key points for further verification.

Specifically, the material identification method provided by the embodiment of the present invention is executed by an artificial intelligence atomic force microscope AI-AFM, and the AI-AFM provides scanning data to a machine learning algorithm in real time. The algorithm is pre-trained using data for material classification and feature identification, and based on artificial intelligence to classify specific material identification, it will dynamically identify other features related to the underlying system, such as domain walls or grain boundaries. Through the control algorithm, the detector will return the key features identified in real time, and further experiments will be carried out in suitable areas on the fly. The key point is that through an efficient machine learning algorithm, high-fidelity pixel-by-pixel recognition can be performed instead of relying on all scanned data, making it possible to identify and verify the microscopic physical mechanism of the microscopic image during the scanning process.

In this embodiment, the machine learning algorithm adopts the SVM classifier, and the artificial intelligence algorithm based on the SVM only needs a small data set and can be trained in less than 10 seconds on an ordinary personal computer. SVMs can be easily trained with a set of labeled examples, each of which consists of a fixed number of features (x1,x2,...,xn) and a label y stating whether it belongs to one of two classes (y=1 or 0). Specifically, a training data set is first prepared for the SVM model, in which the amplitude and phase changes on the morphological interface are used as indicators to classify whether the interface is a ferroelectric domain wall, and then the pixels with pixel labels (domain wall or not) are 14 features are input into the SVM model. Since each image contains 256×256 pixels, almost the same amount of training data is generated (except image borders). Experiments have proved that only 5 pairs of ds-SPM images are enough to train the SVM model, which is much more efficient than CNN which must use the entire image as a training example.

When constructing model training samples, considering that amplitude and phase histograms often exist at several specific angles, it is often inefficient to use real images for labeled training. In order to improve the construction efficiency of model training samples, the training data set can be generated by simulating the amplitude and phase images. It replaces the small marker data by simulating the microstructure to help identify, which can improve the reliability of training and the success rate of identification. Specifically, the characteristics of the amplitude and phase can be simulated through simulation, and at the same time, white noise can be added on the basis to obtain a simulated sample image, thereby avoiding tedious labeling work.

Optionally, since the domain wall is a continuous line on the graph, the length of the longest line on the binary mask is used to judge whether there is a ferroelectric domain wall. Based on this, the material identification model can judge whether there is a ferroelectric domain wall according to the length of the longest line on the binary mask, otherwise it will be judged as a grain boundary. After the material is initially classified, it is also possible to probe the details of the apparent piezoelectric response at the property and mechanism-critical material interface through dynamic adaptive experiments, and to detect the presence of ferroelectrics in ferroelectrics through the identification and tracking of AFM microstructural features. Domain walls and grain boundaries in electrochemical materials. Specifically, the dynamic adaptive experiment can be as follows: when a domain wall is detected, the program will automatically trigger the "ferroelectric program". The SS-PFM experiment is carried out to generate hysteresis loops and butterfly loops corresponding to ferroelectrics. Conversely, if the program does not detect 180° domain walls from amplitude and phase imaging, a "non-ferroelectric program" is triggered to identify grain boundaries superimposed on topography, zoom in on the scan, and then perform a summation on the grain boundaries. Second harmonic piezoelectric response. If the second harmonic piezoelectric response dominates the first harmonic piezoelectric response, its non-ferroelectric nature can be confirmed.

It should be noted that the SVM classifier provided by the embodiment of the present invention has been used to identify and track the structural features of the amplitude and phase responses under the PFM microstructure, and finally cooperate with subsequent verification methods to achieve the identification and verification of the microscopic physical mechanism. For 7174 ds-SPM images, the normalized confusion matrix in Fig. 5d shows that 97.3% of the 475 images predicted to have 180° domain walls were correctly classified, while 6699 images predicted to have no 180° domain walls, 99.6% of them were also correctly identified. These results confirm that the SVM-based AI algorithm is capable of classifying and characterizing ferroelectric materials with 180° domain walls.

Embodiment three

Fig. 2 is a schematic structural diagram of a material identification device based on an artificial intelligence atomic force microscope provided in Embodiment 3 of the present invention. The material identification device based on the artificial intelligence atomic force microscope can be realized by software and/or hardware, for example, the material identification device based on the artificial intelligence atomic force microscope can be configured in a computer device, such as an atomic force microscope. As shown in Figure 2, the device includes a scanned image acquisition module 210, a scanned image classification module 220 and a target material determination module 230, wherein:

A scanning image acquisition module 210, configured to acquire a microscope scanning image;

The scanned image classification module 220 is used to input the microscope scanned image into the pre-trained material recognition model, and obtain the classification result output by the image classification model, wherein the material recognition model is obtained based on the simulation sample image training;

The target material determination module 230 is configured to determine the target material corresponding to the microscope scanning image according to the classification result.

Optionally, on the basis of the above solution, the material recognition model includes a feature extraction module and a classification module, and the scanned image classification module 220 is specifically used for:

Optionally, on the basis of the above solution, the target material determination module 230 is specifically used to:

Verification of classification results based on target materials;

Optionally, on the basis of the above solution, the device also includes a classification model training module for:

Optionally, on the basis of the above solution, the material identification model is constructed based on the support vector machine model.

The material identification device based on artificial intelligence atomic force microscope provided in the embodiment of the present invention can execute the material identification method based on artificial intelligence atomic force microscope provided in any embodiment of the present invention, and has corresponding functional modules and beneficial effects for executing the method.

Embodiment four

FIG. 3 is a schematic structural diagram of a computer device provided by Embodiment 4 of the present invention. Figure 3 shows a block diagram of an exemplary computer device 312 suitable for use in implementing embodiments of the present invention. The computer device 312 shown in FIG. 3 is only an example, and should not limit the functions and scope of use of this embodiment of the present invention.

As shown in FIG. 3, computer device 312 takes the form of a general-purpose computing device. Components of computer device 312 may include, but are not limited to: one or more processors 316 , system memory 328 , bus 318 connecting various system components including system memory 328 and processor 316 .

Bus 318 represents one or more of several types of bus structures, including a memory bus or memory controller, a peripheral bus, an accelerated graphics port, processor 316, or a local bus using any of a variety of bus structures. Examples of these architectures include, but are not limited to, the Industry Standard Architecture (ISA) bus, the Micro Channel Architecture (MAC) bus, the Enhanced ISA bus, the Video Electronics Standards Association (VESA) local bus, and the Peripheral Component Interconnect ( PCI) bus.

Computer device 312 typically includes a variety of computer system readable media. These media can be any available media that can be accessed by computing device 312 and include both volatile and nonvolatile media, removable and non-removable media.

System memory 328 may include computer system readable media in the form of volatile memory, such as random access memory (RAM) 330 and/or cache memory 332 . Computer device 312 may further include other removable/non-removable, volatile/nonvolatile computer system storage media. By way of example only, storage device 334 may be used to read from and write to non-removable, non-volatile magnetic media (not shown in FIG. 3, commonly referred to as a "hard drive"). Although not shown in Figure 3, disk drives for reading and writing to removable non-volatile disks (such as "floppy disks") may be provided, as well as for removable non-volatile optical disks (such as CD-ROM, DVD-ROM or other optical media) read and write optical drives. In these cases, each drive may be connected to bus 318 through one or more data media interfaces. Memory 328 may include at least one program product having a set (eg, at least one) of program modules configured to perform the functions of various embodiments of the present invention.

A program/utility 340 having a set (at least one) of program modules 342, such as but not limited to, an operating system, one or more application programs, other program modules, and program data, may be stored, for example, in memory 328 , each or some combination of these examples may include implementations of network environments. Program modules 342 generally perform the functions and/or methodologies of the described embodiments of the invention.

Computer device 312 may also communicate with one or more external devices 314 (e.g., a keyboard, pointing device, display 324, etc.), and with one or more devices that enable a user to interact with computer device 312, and/or with Any device (eg, network card, modem, etc.) that enables the computing device 312 to communicate with one or more other computing devices. Such communication may occur through input/output (I/O) interface 322 . Moreover, the computer device 312 can also communicate with one or more networks (eg, a local area network (LAN), a wide area network (WAN) and/or a public network, such as the Internet) through the network adapter 320 . As shown, network adapter 320 communicates with other modules of computer device 312 via bus 318 . It should be appreciated that although not shown, other hardware and/or software modules may be used in conjunction with computer device 312, including but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives And data backup storage system, etc.

The processor 316 executes various functional applications and data processing by running the program stored in the system memory 328, such as realizing the material identification method based on the artificial intelligence atomic force microscope provided by the embodiment of the present invention, the method includes:

Obtain a microscope scan image;

Of course, those skilled in the art can understand that the processor can also implement the technical solution of the artificial intelligence atomic force microscope-based material identification method provided by any embodiment of the present invention.

Embodiment five

Embodiment 5 of the present invention also provides a computer-readable storage medium on which a computer program is stored. When the program is executed by a processor, the material identification method based on artificial intelligence atomic force microscope provided in the embodiment of the present invention is realized. The method include:

Obtain a microscope scan image;

Of course, the computer-readable storage medium provided by the embodiment of the present invention, the computer program stored thereon is not limited to the above-mentioned method operation, and can also execute the material identification method based on artificial intelligence atomic force microscope provided by any embodiment of the present invention related operations.

The computer storage medium in the embodiments of the present invention may use any combination of one or more computer-readable media. The computer readable medium may be a computer readable signal medium or a computer readable storage medium. A computer readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device, or device, or any combination thereof. More specific examples (non-exhaustive list) of computer-readable storage media include: electrical connections with one or more leads, portable computer disks, hard disks, random access memory (RAM), read only memory (ROM), Erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination of the above. In this document, a computer-readable storage medium may be any tangible medium that contains or stores a program that can be used by or in conjunction with an instruction execution system, apparatus, or device.

A computer readable signal medium may include a data signal carrying computer readable program code in baseband or as part of a carrier wave. Such propagated data signals may take many forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the foregoing. A computer-readable signal medium may also be any computer-readable medium other than a computer-readable storage medium, which can send, propagate, or transmit a program for use by or in conjunction with an instruction execution system, apparatus, or device. .

Program code embodied on a computer readable medium may be transmitted using any appropriate medium, including - but not limited to wireless, wireline, optical fiber cable, RF, etc., or any suitable combination of the foregoing.

Computer program code for carrying out the operations of the present invention may be written in one or more programming languages, or combinations thereof, including object-oriented programming languages—such as Java, Smalltalk, C++, and conventional procedural programming languages. Programming language - such as "C" or a similar programming language. The program code may execute entirely on the user's computer, partly on the user's computer, as a stand-alone software package, partly on the user's computer and partly on a remote computer or entirely on the remote computer or server. In cases involving a remote computer, the remote computer can be connected to the user computer through any kind of network, including a local area network (LAN) or a wide area network (WAN), or it can be connected to an external computer (such as through an Internet service provider). Internet connection).

Note that the above are only preferred embodiments of the present invention and applied technical principles. Those skilled in the art will understand that the present invention is not limited to the specific embodiments herein, and various obvious changes, readjustments and substitutions can be made by those skilled in the art without departing from the protection scope of the present invention. Therefore, although the present invention has been described in detail through the above embodiments, the present invention is not limited to the above embodiments, and can also include more other equivalent embodiments without departing from the concept of the present invention, and the present invention The scope is determined by the scope of the appended claims.

Claims

A material identification method based on an artificial intelligence atomic force microscope, characterized in that it includes:

Obtain a microscope scan image;

Inputting the microscope scanning image into a pre-trained material recognition model to obtain a classification result output by the image classification model, wherein the material recognition model is obtained based on simulation sample image training;

The target material corresponding to the microscope scanning image is determined according to the classification result.
The method according to claim 1, wherein the material recognition model includes a feature extraction module and a classification module, and the scanning microscope image is input into a pre-trained material recognition model to obtain the material recognition model The output classification results include:

Inputting the scanned microscope image into the feature extraction module to obtain the structural features output by the feature extraction module;

The structural features are input into the classification module, and the classification result output by the classification module is obtained.
The method according to claim 1, wherein the determining the target material corresponding to the microscope scanning image according to the classification result comprises:

Carrying out classification result verification based on the target material;

When the classification result is verified to pass the verification, the material corresponding to the classification result is used as the target material.
The method according to claim 3, wherein the verification of classification results based on the target material comprises:

When the classification result is ferroelectric, adjusting the scanning area to obtain domain wall information of the scanning object;

A switching spectroscopy piezoelectric response force microscopy experiment is performed based on the domain wall information, and when the loop parameters corresponding to the scanning object are generated, it is determined that the classification result of the ferroelectric category has passed the verification.
The method according to claim 3, wherein the verification of classification results based on the target material comprises:

When the classification result is a non-ferroelectric class, adjusting the scanning area to obtain grain boundary information of the scanning object;

performing a first harmonic piezoelectric response and a second harmonic piezoelectric response based on the grain boundary information;

When the second harmonic piezoelectric response dominates the first harmonic piezoelectric response, it is determined that the verification of the classification result of the non-ferroelectric category is passed.
The method according to claim 1, wherein the training of the material recognition model comprises:

determining the sample image category according to the classification category corresponding to the material recognition model;

performing simulation based on the category of the sample image to obtain a simulation sample image;

Constructing a model training sample based on the sample image category and the simulation sample image, using the model training sample to train a pre-built material recognition model to obtain a trained material recognition model.
The method according to claim 1, wherein the material identification model is constructed based on a support vector machine model.
A material identification device based on an artificial intelligence atomic force microscope, characterized in that it includes:

A scan image acquisition module, configured to acquire a microscope scan image;

A scanned image classification module, configured to input the microscope scanned image into a pre-trained material recognition model, and obtain a classification result output by the image classification model, wherein the material recognition model is trained based on a simulation sample image;

The target material determination module is configured to determine the target material corresponding to the microscope scanning image according to the classification result.
A computer device, characterized in that the device comprises:

one or more processors;

storage means for storing one or more programs;

When the one or more programs are executed by the one or more processors, so that the one or more processors implement the material identification method based on artificial intelligence atomic force microscope as described in any one of claims 1-7 .
A computer-readable storage medium, on which a computer program is stored, characterized in that, when the program is executed by a processor, the material identification method based on an artificial intelligence atomic force microscope as described in any one of claims 1-7 is realized.