STM image recognition method and device and electronic equipment

Abstract

The embodiment of the invention provides an STM image recognition method. The STM image recognition method comprises the following steps: acquiring an STM image; identifying circular molecular features of the STM image; converting the STM image containing the circular molecular features to obtain a space image and a frequency domain image corresponding to the STM image; and inputting the space image and the frequency domain image into a convolutional neural network to learn features of the STM image. According to the STM image recognition method provided by the embodiment of the invention, the convolutional neural network is input by using two channels of the frequency domain and the space domain, and a relatively high iteration speed can be achieved only by using a small number of training samples, so that an efficient molecular recognition mode of the STM image with low training cost and high accuracy is realized. The embodiment of the invention further provides an STM image recognition device and electronic equipment.

Description

Technical Field

[0001] The embodiments of the present application relate to the field of image processing technology. Specifically, the embodiments of the present application relate to a method, an apparatus, and an electronic device for STM image recognition. Background Art

[0002] Currently, a Scanning Tunneling Microscope (STM) is a high-precision surface imaging instrument for exploring the surface topography of a sample with atomic-level spatial resolution. Utilizing the quantum tunneling effect in quantum mechanics, it characterizes the surface topography of the sample based on the change in tunneling current. STM technology is widely applied in fields such as surface science, materials science, chemistry, biology, and physics. The information contained in STM images is relatively rich, such as defects and a series of regular changes generated by molecular surface self-assembly.

[0003] Existing STM image recognition methods mainly include methods based on Convolutional Neural Networks (CNNs) and methods based on manually extracting features combined with machine learning. Existing image recognition methods require a large amount of training or require manual search for effective classification and recognition features, and it is difficult to obtain an efficient and practical recognition model in a short time. Summary of the Invention

[0004] In view of the problems existing in the above-mentioned prior art, the embodiments of the present application provide a method, an apparatus, and an electronic device for STM image recognition. By applying a convolutional neural network with dual-channel input in the frequency domain and the spatial domain, only a small number of training samples are required, and it can have a relatively fast iteration speed to achieve an efficient molecular recognition mode for STM images with low training cost and high accuracy.

[0005] In a first aspect, the embodiments of the present application provide a method for STM image recognition, including the following steps:

[0006] Obtain an STM image;

[0007] Identify the circular molecular features of the STM image;

[0008] Convert the STM image containing the circular molecular features to obtain a spatial image and a frequency domain image corresponding to the STM image; and

[0009] Input the spatial image and the frequency domain image into a convolutional neural network to learn the features of the STM image.

[0010] Further, the convolutional neural network includes an ascending part, a top part, and a descending part. The ascending part includes a spatial attention module, a feature expansion convolutional block, and a first pooling layer. The top part includes a channel attention module. The descending part includes a feature condensation convolutional block, a second pooling layer, and a fully connected layer. The step of inputting the spatial image and the frequency domain image into the convolutional neural network to learn the features of the STM image includes:

[0011] Taking the spatial image and the frequency domain image as inputs simultaneously, calibrating the important image regions of the STM image through the spatial attention module in the ascending part to learn the image features of the STM image, expanding the learning of the image features of the STM image through the feature expansion convolutional block, and reducing the image size of the STM image through the first pooling layer to reduce the computational amount;

[0012] Learning from the extracted image features of the STM image through the channel attention module in the top part; and

[0013] Summarizing the features of the STM image through the feature condensation convolutional block in the descending part, reducing the image size of the STM image through the second pooling layer to reduce the computational amount, and judging the pattern of the molecular image of the STM image through the fully connected layer.

[0014] Further, the convolutional neural network further includes a skip connection part, which connects the ascending part and the descending part in the regions of the first pooling layer and the second pooling layer to learn the spatial features of the STM image.

[0015] Further, the step of converting the STM image containing the circular molecular features to obtain the corresponding spatial image and frequency domain image includes:

[0016] Performing a fast Fourier transform on the STM image containing the circular molecular features to obtain the corresponding spatial image and frequency domain image.

[0017] Further, the step of identifying the circular molecular features of the STM image includes:

[0018] Obtaining a two-dimensional matrix of the feedback current from the STM and performing normalization processing to form the STM image;

[0019] Preprocessing the STM image to reduce the noise of the STM image;

[0020] Separating the foreground and background of the STM image; and

[0021] Detect circular molecules in the STM image.

[0022] Further, separating the foreground and background of the STM image includes:

[0023] Performing clustering processing on the STM image through a Gaussian mixture model to separate the foreground and background of the STM image.

[0024] Further, detecting circular molecules in the STM image includes:

[0025] Performing binarization processing on the STM image;

[0026] Performing gradient detection on the binarized STM image, establishing a circumferential equation for each edge pixel of the STM image and transforming it into the parameter space; and

[0027] Judging the edge pixels, if it is greater than a preset threshold, it is determined as a circular molecule in the STM image.

[0028] In a second aspect, an embodiment of the present application further provides an STM image recognition device, including:

[0029] An image acquisition module for acquiring an STM image;

[0030] A feature recognition module for recognizing circular molecule features of the STM image;

[0031] A frequency-domain image acquisition module for converting the STM image containing the circular molecule features to obtain a spatial image and a frequency-domain image corresponding to the STM image; and

[0032] A feature learning module for inputting the spatial image and the frequency-domain image into a convolutional neural network to learn the features of the STM image.

[0033] In a third aspect, an embodiment of the present application further provides an electronic device, including a memory, a processor, and a computer program stored on the memory and executable on the processor, where the processor is configured to implement the STM image recognition method according to the first aspect described above when executing the program.

[0034] In a fourth aspect, an embodiment of the present application further provides a computer-readable storage medium, on which a computer program is stored, and the computer program is used to implement the STM image recognition method according to the first aspect described above.

[0035] The embodiments of the present application bring the following beneficial effects:

[0036] In the STM image recognition method provided by the embodiments of the present application, first, an STM image is obtained, and the circular molecular features of the STM image are recognized. By converting the STM image containing the circular molecular features, a spatial image and a frequency-domain image corresponding to the STM image are obtained. Finally, the spatial image and the frequency-domain image are input into a convolutional neural network to learn the features of the STM image. The STM image recognition method provided by the embodiments of the present application, by using a convolutional neural network with dual-channel input in the frequency domain and the spatial domain, only requires a small number of training samples and can have a relatively fast iteration speed to achieve an efficient molecular recognition mode for STM images with low training costs and high accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the drawings in the following description are only some embodiments of the present application. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on the structures shown in these drawings.

[0038] Figure 1 It is a schematic flowchart of the STM image recognition method provided by the embodiments of the present application;

[0039] Figure 2 It is a schematic structural diagram of the convolutional neural network used in the STM image recognition method provided by the embodiments of the application;

[0040] Figure 3 It is a schematic diagram of the steps for recognizing circular molecules in an STM image used in the STM image recognition method provided by the embodiments of the application;

[0041] Figure 4 It is a structural block diagram of the STM image recognition device provided by the embodiments of the present application;

[0042] Figure 5 It is a schematic structural diagram of the electronic device provided by the embodiments of the present application.

[0043] The realization, functional features, and advantages of the objectives of the present application will be further described in conjunction with the embodiments and with reference to the drawings. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0044] The following will clearly and completely describe the technical solutions in the embodiments of the present application with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all of the embodiments. Based on the embodiments of the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the scope of protection of the present application.

[0045] In the description, claims and the above-mentioned drawings of the present application, the terms "first" and "second" are only used for descriptive purposes and cannot be construed as indicating or implying relative importance or implicitly specifying the quantity of the indicated technical features. Thus, the features defined with "first" and "second" may explicitly or implicitly include one or more of such features. In the description of the present application, unless otherwise specified, the meaning of "a plurality" is two or more. For those of ordinary skill in the art, the specific meanings of the above terms in the present application can be understood according to specific circumstances.

[0046] Figure 1 is a flowchart of the STM image recognition method according to an embodiment of the present application. As Figure 1 shown, the STM image recognition method according to an embodiment of the present application includes the following steps:

[0047] S101: Obtain an STM image;

[0048] A scanning tunneling microscope is a high-precision surface imaging instrument that explores the surface topography of a sample with atomic-scale spatial resolution. Utilizing the quantum tunneling effect in quantum mechanics, it characterizes the surface topography of the sample based on the change in tunneling current, thereby obtaining an STM image.

[0049] S102: Identify the circular molecular features of the STM image;

[0050] The STM image obtained by an STM electron microscope usually shows a dot-like distribution. Generally, it is necessary to identify the circular molecular features of the STM image to accurately identify each component or various particles in the image by extracting the effective features of the image.

[0051] S103: Convert the STM image containing the circular molecular features to obtain a spatial image and a frequency-domain image corresponding to the STM image; and

[0052] That is, by calculating the input STM image, its frequency-domain image is obtained. The spatial image and the frequency-domain image are used as inputs simultaneously and applied to the input of a convolutional neural network, providing more processing basis for the convolutional neural network and introducing periodic features into the feature extraction process of the image. More feature selections can reduce the requirements for training costs and effectively improve the model accuracy.

[0053] S104: Input the spatial image and the frequency-domain image into a convolutional neural network to learn the features of the STM image.

[0054] As described above, by applying a convolutional neural network with dual-channel input in the frequency domain and the spatial domain, only a small number of training samples are required, and it can have a fast iteration speed to achieve an efficient molecular recognition mode for STM images with low training cost and high accuracy.

[0055] Therefore, in the STM image recognition method provided by the embodiments of the present application, first, an STM image is obtained, and the circular molecular features of the STM image are recognized. By converting the STM image including the circular molecular features, a spatial image and a frequency-domain image corresponding to the STM image are obtained. Finally, the spatial image and the frequency-domain image are input into a convolutional neural network to learn the features of the STM image. The STM image recognition method provided by the embodiments of the present application, by applying a convolutional neural network with dual-channel input in the frequency domain and the spatial domain, only requires a small number of training samples, and can have a fast iteration speed to achieve an efficient molecular recognition mode for STM images with low training cost and high accuracy.

[0056] Further, referring to Figure 2 , in some embodiments of the present application, the convolutional neural network includes an ascending part, a top part, and a descending part. The ascending part includes a spatial attention module, a feature expansion convolutional block, and a first pooling layer. The top part includes a channel attention module. The descending part includes a feature condensation convolutional block, a second pooling layer, and a fully connected layer. The step of inputting the spatial image and the frequency-domain image into the convolutional neural network to learn the features of the STM image includes:

[0057] Taking the spatial image and the frequency-domain image as inputs simultaneously, calibrating the important image regions of the STM image through the spatial attention module in the ascending part to learn the image features of the STM image, expanding the learning of the image features of the STM image through the feature expansion convolutional block, and reducing the image size of the STM image through the first pooling layer to reduce the calculation amount;

[0058] Learning from the extracted image features of the STM image through the channel attention module in the top part; and

[0059] Summarizing the features of the STM image through the feature condensation convolutional block in the descending part, reducing the image size of the STM image through the second pooling layer to reduce the calculation amount, and judging the mode of the molecular image of the STM image through the fully connected layer.

[0060] Specifically, as Figure 2As shown, the convolutional neural model adopted by the STM image recognition method provided by the embodiments of the present application is shaped like a capital English letter A, so it can be called an A-shaped neural network. When recognizing the STM image, first, a 64*64 image is intercepted from a part of the detected center of the circle, and its frequency-domain image is calculated. The spatial image and the frequency-domain image are used as inputs at the same time. First, the important image regions are calibrated through the spatial attention module, and then the image features are enlarged through the feature expansion convolutional block, and combined with the first pooling layer to reduce the image size to reduce the calculation amount. A channel attention module is set at the top part to learn the important parts of the STM image from the features of 128 channels extracted, and higher weights are assigned. The features of 128 channels are summarized through the feature condensation convolutional block in the descending part, and combined with the second pooling layer to reduce the image size to reduce the calculation amount. Finally, the molecular image pattern is judged through the fully connected layer.

[0061] Therefore, in the STM image recognition method provided by the embodiments of the present application, after the image features of the STM image are input, the feature learning is continuously expanded in the ascending part, important feature channels are selected at the top, the features learned before are continuously summarized in the descending part, and combined with the skip connection technology, the features of the ascending part and the descending part are combined, which can improve the recognition accuracy of the STM image, expand the feature learning space of the STM image at the same time, and can learn the effective features of the STM image in fewer trainings to further reduce the training cost.

[0062] Furthermore, referring again to Figure 2 , in some embodiments of the present application, the convolutional neural network further includes a skip connection part, and the skip connection part connects the ascending part and the descending part in the areas of the first pooling layer and the second pooling layer to learn the spatial features of the STM image.

[0063] Specifically, the skip connection part of the convolutional neural network connects the ascending part and the descending part of the model in the pooling layer area, so as to prevent the model from ignoring the importance of spatial features during the continuous abstract feature learning process, so as to further improve the recognition accuracy of the STM image features.

[0064] Furthermore, in some embodiments of the present application, the conversion of the STM image including the circular molecular features to obtain the spatial image and the frequency-domain image corresponding to the STM image includes:

[0065] Performing a fast Fourier transform on the STM image including the circular molecular features to obtain the spatial image and the frequency-domain image corresponding to the STM image.

[0066] Specifically, the Fast Fourier Transform (FFT) can convert the STM image into a frequency-domain image, thereby revealing the frequency components of the STM image and introducing periodic features into the image feature extraction process. More feature selections can reduce the requirements for the training cost of the convolutional neural network and can effectively improve the model accuracy.

[0067] Further, referring to Figure 3 , in some embodiments of the present application, the recognition of the circular molecular features of the STM image includes:

[0068] Obtaining a two-dimensional matrix of the feedback current by STM and performing normalization processing to form the STM image;

[0069] Preprocessing the STM image to reduce the noise of the STM image;

[0070] Separating the foreground and background of the STM image; and

[0071] Detecting the circular molecules of the STM image.

[0072] Specifically, as Figure 3 shown, in the STM image recognition method provided by the embodiments of the present application, first, a scanning tunneling microscope obtains a two-dimensional matrix of the feedback current and performs normalization processing on the data to form an STM image. Then, the contrast of the STM image is improved and the image noise is reduced through adaptive gray histogram equalization and adaptive median filtering techniques. Next, the foreground and background of the STM image are separated, and the STM image is binarized to facilitate subsequent circular feature detection. Then, the molecular circular features of the STM image are recognized by, for example, the Hough circle transform. Finally, the detection result is marked on the input image for direct manual observation of the detection result.

[0073] In the STM image recognition method provided by the embodiments of the present application, in image data processing, the molecular detection part of the adopted model mainly uses the Hough circle transform algorithm, and the circular molecule recognition part combines the attention mechanism. The Hough circle transform can resist the noise in the STM image. By mapping the data to the parameter space and judging the circular features through the accumulator to count the local maximum value, it can reduce the influence of image noise. The attention mechanism can assign higher weights to important regions of the image during training, thus avoiding interference from other regions. Secondly, the attention mechanism can perform parallel computing, and the Hough circle algorithm can process multiple circular recognitions simultaneously, with certain parallel characteristics, which can improve the computing efficiency of the algorithm. The attention mechanism allows the network to directly consider the information of all positions in the STM image when calculating the response at a certain position, so that it does not need to rely on gradually expanding local operations and can directly capture long-range dependencies. The Hough circle algorithm can detect circles in the STM image by mapping points in the image space to the parameter space. The above transformation of the parameter space enables the Hough circle algorithm to capture long-range dependencies between different regions in the image. Compared with traditional models such as CNN and RNN, the attention mechanism has fewer parameters and lower model complexity, thus reducing the requirement for computing power. The Hough circle algorithm improves the parameter efficiency and operation efficiency by reducing the dimension of the Hough space.

[0074] Further, in some embodiments of the present application, separating the foreground and background of the STM image includes:

[0075] Performing clustering processing on the STM image through a Gaussian mixture model to separate the foreground and background of the STM image.

[0076] In addition, in the STM image recognition method provided by the embodiments of the present application, a Gaussian mixture model can be used to separate the foreground and background of the image, which improves the contrast of the image, reduces interference and enhances the circular features of the data, thereby effectively improving the detection accuracy of the Hough circle transform.

[0077] Further, in some embodiments of the present application, detecting circular molecules in the STM image includes:

[0078] Performing binarization processing on the STM image;

[0079] Performing gradient detection on the binarized STM image, establishing a circumferential equation for each edge pixel of the STM image and transforming it into the parameter space; and

[0080] Judging the edge pixels, and if it is greater than a preset threshold, it is determined as a circular molecule in the STM image.

[0081] The STM image recognition method provided by the embodiments of the present application preferably uses the Hough circle transform algorithm to recognize circular molecules in the STM image. That is, first, the STM image needs to be binarized, and the gradient of the binarized STM image is detected. The equation of the circumference is established for each edge pixel of the STM image and transformed into the parameter space. The possible center of the circle is determined or voted on. When the set threshold is met, it is recognized as the center of the circle or circular molecule in the STM image. The Hough circle algorithm can detect circles in the STM image by mapping points in the image space to the parameter space. The transformation of the above parameter space enables the Hough circle algorithm to capture the long-distance dependencies between different regions in the image, thereby improving the accuracy of circular molecule recognition.

[0082] Figure 4 It is the structural block diagram of the STM image recognition device 200 provided by the embodiments of the present application. As Figure 4 shown, the STM image recognition device 200 of the embodiments of the present application includes: an image acquisition module 210, a feature recognition module 220, a frequency-domain image acquisition module 230, and a feature learning module 240, where:

[0083] The image acquisition module 210 is configured to acquire an STM image;

[0084] The feature recognition module 220 is configured to recognize the circular molecule features of the STM image;

[0085] The frequency-domain image acquisition module 230 is configured to convert the STM image including the circular molecule features to obtain a spatial image and a frequency-domain image corresponding to the STM image; and

[0086] The feature learning module 240 is configured to input the spatial image and the frequency-domain image into a convolutional neural network to learn the features of the STM image.

[0087] In the STM image recognition device provided by the embodiments of the present application, first, an STM image is acquired, and the circular molecule features of the STM image are recognized. By converting the STM image including the circular molecule features, a spatial image and a frequency-domain image corresponding to the STM image are obtained. Finally, the spatial image and the frequency-domain image are input into a convolutional neural network to learn the features of the STM image. The STM image recognition method provided by the embodiments of the present application, by using a convolutional neural network with dual-channel input in the frequency domain and the spatial domain, only requires a small number of training samples, and can have a fast iteration speed to achieve an efficient molecular recognition mode for STM images with low training costs and high accuracy.

[0088] It should be noted that the specific implementation of the STM image recognition device in the embodiments of the present application is similar to that of the STM image recognition method in the embodiments of the present application. For details, please refer to the description in the method section and will not be elaborated here.

[0089] Figure 5 FIG. is a schematic structural diagram of an electronic device 300 according to an embodiment of the present application.

[0090] As Figure 5 shown, the electronic device 300 includes a central processing unit (CPU) 301, which can perform various appropriate actions and processes according to the program stored in the read-only memory (ROM) 302 or the program loaded from the storage section 302 into the random access memory (RAM) 303. In the RAM 303, various programs and data required for the operation of the electronic device 300 are also stored. The CPU 301, ROM 302, and RAM 303 are connected to each other via a bus 304. An input / output (I / O) interface 305 is also connected to the bus 304.

[0091] The following components are connected to the I / O interface 305: an input section 306 including a keyboard, a mouse, etc.; an output section 307 including a cathode ray tube (CRT), a liquid crystal display (LCD), etc. and a speaker, etc.; a storage section 308 including a hard disk, etc.; and a communication section 309 including a network interface card such as a LAN card, a modem, etc. The communication section 309 performs communication processing via a network such as the Internet. A drive 310 is also connected to the I / O interface 305 as required. A removable medium 311, such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, etc., is installed on the drive 310 as required so that a computer program read from it can be installed into the storage section 308 as required.

[0092] Particularly, according to the embodiments of the present application, the process described above with reference to the flowchart can be implemented as a computer software program. For example, the embodiments of the present application include a computer program product, which includes a computer program carried on a machine-readable medium, and the computer program includes program codes for performing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from the network through the communication section 309, and / or installed from the removable medium 311. When the computer program is executed by the central processing unit (CPU) 301, the above functions defined in the electronic device of the present application are executed.

[0093] It should be noted that the computer-readable medium shown in this application can be a computer-readable signal medium, a computer-readable storage medium, or any combination of the two. A computer-readable storage medium can be, for example, but not limited to, an electronic device, apparatus, or device of electricity, magnetism, light, electromagnetic, infrared, or semiconductor, or any combination of the above. More specific examples of the computer-readable storage medium can include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above.

[0094] In this application, a computer-readable storage medium can be any tangible medium that contains or stores a program that can be used by or in conjunction with an instruction-executing electronic device, apparatus, or device. And in this application, a computer-readable signal medium can include a data signal propagated in a baseband or as part of a carrier wave, which carries computer-readable program code. Such a propagated data signal can take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. A computer-readable signal medium can 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-executing electronic device, apparatus, or device. The program code contained on a computer-readable medium can be transmitted using any appropriate medium, including but not limited to: wireless, wire, optical fiber, RF, etc., or any suitable combination of the above.

[0095] The flowcharts and block diagrams in the accompanying drawings illustrate the possible architectures, functions, and operations of a processing receiving device, method, and computer program product according to various embodiments of this application. In this regard, each block in the flowchart or block diagram can represent a module, a program segment, or a part of code, and the foregoing module, program segment, or part of code contains one or more executable instructions for implementing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the blocks can occur in a different order than marked in the accompanying drawings. For example, two consecutive blocks shown can actually be executed substantially in parallel, and they can sometimes be executed in the reverse order, depending on the functions involved. It should also be noted that each block in the block diagram and / or flowchart, and the combination of blocks in the block diagram and / or flowchart, can be implemented by a dedicated hardware-based electronic device that executes the specified function or operation, or can be implemented by a combination of dedicated hardware and computer instructions.

[0096] The units or modules involved in the embodiments of the present application can be implemented in software or in hardware. The described units or modules can also be provided in a processor, and when the processor executes the program, it implements the STM image recognition method:

[0097] Obtain an STM image;

[0098] Identify the circular molecular features of the STM image;

[0099] Convert the STM image including the circular molecular features to obtain a spatial image and a frequency-domain image corresponding to the STM image; and

[0100] Input the spatial image and the frequency-domain image into a convolutional neural network to learn the features of the STM image.

[0101] As another aspect, the present application also provides a computer-readable storage medium, which may be included in the electronic device described in the above embodiments; or may exist alone without being assembled into the electronic device. The above computer-readable storage medium stores one or more programs, and when the foregoing programs are executed by one or more processors, they implement the STM image recognition method described in the present application:

[0102] Obtain an STM image;

[0103] Identify the circular molecular features of the STM image;

[0104] Convert the STM image including the circular molecular features to obtain a spatial image and a frequency-domain image corresponding to the STM image; and

[0105] Input the spatial image and the frequency-domain image into a convolutional neural network to learn the features of the STM image.

[0106] As another aspect, the present application also provides a computer program product, which may be included in the electronic device described in the above embodiments; or may exist alone without being assembled into the electronic device. The above computer program product stores one or more programs, and when the foregoing programs are executed by one or more processors, they implement the STM image recognition method described in the present application:

[0107] Obtain an STM image;

[0108] Identify the circular molecular features of the STM image;

[0109] Convert the STM image containing the circular molecular feature to obtain a spatial image and a frequency domain image corresponding to the STM image; and

[0110] Input the spatial image and the frequency domain image into a convolutional neural network to learn the features of the STM image.

[0111] The above are only the preferred embodiments of the present application, and do not limit the patent scope of the present application. Any equivalent structural transformation made under the application concept of the present application by using the content of the specification and drawings of the present application, or directly / indirectly applied in other related technical fields, is included in the patent protection scope of the present application.

Claims

1. A STM image recognition method, characterized in that: The following steps are involved: Acquire STM images; identifying circular molecular features of the STM image; Converting the STM image containing the circular molecular features to obtain a spatial image and a frequency domain image corresponding to the STM image; and The spatial image and the frequency domain image are input into a convolutional neural network to learn the features of the STM image.

2. The STM image recognition method according to claim 1, characterized in that: The convolutional neural network includes an ascending part, a top part and a descending part, the ascending part includes a spatial attention module, a feature expansion convolution block and a first pooling layer, the top part includes a channel attention module, the descending part includes a feature concentration convolution block, a second pooling layer and a fully connected layer, and the spatial image and the frequency domain image are input into the convolutional neural network to learn the features of the STM image, including: The spatial image and the frequency domain image are simultaneously used as inputs, the important image regions of the STM image are calibrated by the spatial attention module of the ascending part to learn the image features of the STM image, the learning of the image features of the STM image is expanded by the feature expansion convolution block, and the image size of the STM image is reduced by the first pooling layer to reduce the amount of calculation; Learning from the image features of the STM image extracted by the channel attention module of the top portion; and The features of the STM image are summarized through the feature concentration convolution block of the descending part, the image size of the STM image is reduced through the second pooling layer to reduce the amount of calculation, and the pattern of the molecular image of the STM image is determined through the fully connected layer.

3. The STM image recognition method according to claim 2, characterized in that: The convolutional neural network also includes a jump connection part, which connects the ascending part and the descending part in the first pooling layer and the second pooling layer area to learn the spatial features of the STM image.

4. The STM image recognition method according to any one of claims 1 to 3, characterized in that: The step of converting the STM image containing the circular molecular features to obtain a spatial image and a frequency domain image corresponding to the STM image comprises: The STM image containing the circular molecular features is subjected to a fast Fourier transform to obtain a spatial image and a frequency domain image corresponding to the STM image.

5. The STM image recognition method according to claim 1, characterized in that: The identifying circular molecular features of the STM image comprises: Acquiring a two-dimensional matrix of feedback current by STM and performing normalization processing to form the STM image; Preprocessing the STM image to reduce noise in the STM image; Separating the foreground and background of the STM image; and Detect circular molecules in the STM image.

6. The STM image recognition method according to claim 5, characterized in that: The foreground and background of the STM image are separated, comprising: The STM image is clustered using a Gaussian mixture model to separate the foreground and background of the STM image.

7. The STM image recognition method according to claim 5, characterized in that: The detecting of the circular molecules in the STM image comprises: Binarization processing is performed on the STM image; Performing gradient detection on the STM image after binarization, establishing a circle equation for each edge pixel of the STM image and converting it into a parameter space; and The edge pixels are judged, and if they are greater than a preset threshold, they are identified as circular molecules in the STM image.

8. An STM image recognition device, characterized in that: include: An image acquisition module, used for acquiring STM images; A feature recognition module, used for recognizing circular molecular features of the STM image; A frequency domain image acquisition module, used for converting the STM image containing the circular molecular features to acquire a spatial image and a frequency domain image corresponding to the STM image; and A feature learning module is used to input the spatial image and the frequency domain image into a convolutional neural network to learn the features of the STM image.

9. An electronic device, characterized in that: The invention comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the processor is used to implement the STM image recognition method according to any one of claims 1 to 7 when executing the program.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and the computer program is used to implement the STM image recognition method according to any one of claims 1-7.

Images