CN114743187A - Automatic login method, system, equipment and storage medium for bank security control - Google Patents

Automatic login method, system, equipment and storage medium for bank security control Download PDF

Info

Publication number
CN114743187A
CN114743187A CN202210331616.1A CN202210331616A CN114743187A CN 114743187 A CN114743187 A CN 114743187A CN 202210331616 A CN202210331616 A CN 202210331616A CN 114743187 A CN114743187 A CN 114743187A
Authority
CN
China
Prior art keywords
button
image
security control
control keyboard
neural network
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN202210331616.1A
Other languages
Chinese (zh)
Inventor
吴昆�
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Mairong Intelligent Technology Shanghai Co ltd
Original Assignee
Mairong Intelligent Technology Shanghai Co ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Mairong Intelligent Technology Shanghai Co ltd filed Critical Mairong Intelligent Technology Shanghai Co ltd
Priority to CN202210331616.1A priority Critical patent/CN114743187A/en
Publication of CN114743187A publication Critical patent/CN114743187A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
    • G06F16/33—Querying
    • G06F16/332—Query formulation
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
    • G06F16/36—Creation of semantic tools, e.g. ontology or thesauri
    • G06F16/374—Thesaurus
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/243—Classification techniques relating to the number of classes
    • G06F18/2431—Multiple classes
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F21/00—Security arrangements for protecting computers, components thereof, programs or data against unauthorised activity
    • G06F21/30—Authentication, i.e. establishing the identity or authorisation of security principals
    • G06F21/45—Structures or tools for the administration of authentication
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/048—Activation functions
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • General Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • Computational Linguistics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Artificial Intelligence (AREA)
  • Software Systems (AREA)
  • Mathematical Physics (AREA)
  • Evolutionary Computation (AREA)
  • Computing Systems (AREA)
  • Molecular Biology (AREA)
  • General Health & Medical Sciences (AREA)
  • Biophysics (AREA)
  • Biomedical Technology (AREA)
  • Health & Medical Sciences (AREA)
  • Databases & Information Systems (AREA)
  • Computer Security & Cryptography (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Evolutionary Biology (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Computer Hardware Design (AREA)
  • Character Discrimination (AREA)

Abstract

The invention discloses a method, a system, equipment and a storage medium for automatically logging in a bank security control, wherein the method comprises the following steps: generating a security control keyboard at a web end and acquiring a webpage screenshot; constructing a network image segmentation model based on UNet, and segmenting a security control keyboard image from a webpage screenshot; segmenting all button images from the security control keyboard image, and acquiring the positions of the button images in the security control keyboard; constructing a character recognition neural network model, inputting the button image into the character recognition neural network model, and outputting to obtain characters corresponding to the button image; constructing a button-character dictionary by combining the positions of the button images in the safety control keyboard and characters corresponding to the button images; and searching the corresponding button position of the character to be input in the safety control keyboard according to the button-character dictionary, and triggering click operation on the button position to finish automatic login. The invention can effectively improve the efficiency of automatic login of the bank security control.

Description

Automatic login method, system, equipment and storage medium for bank security control
Technical Field
The invention relates to the technical field of deep learning and internet finance, in particular to a method, a system, equipment and a storage medium for automatically logging in a bank security control based on deep learning.
Background
The bank receipt is a valid certificate for providing the enterprise customer with the functions of online bank payment, transaction inquiry, downloading, printing (subsidy printing) and verification, and for each account, a corresponding receipt is provided. The bank receipt is used as an original basis for an enterprise to compile a bookkeeping voucher, and a large amount of manual labor, material resources and financial resources are consumed in the bookkeeping process due to the fact that a large amount of receipt information is input.
With the rapid development of the internet technology and the OCR technology, the RPA robot can be deployed on a computer to simulate human operation to complete bank receipt management, so that the automation of the bank receipt management process is realized, and a large amount of manpower and material resources are saved. When the receipt management robot inquires or downloads the receipt, firstly, logging in an online commercial bank; however, in order to secure the account of the user, the commercial bank usually installs a security control, and requires clicking a button on a security control keyboard to input the user password. The value displayed by the security control keyboard is different from the value finally input, and the bank background calculates the password of the user according to a certain encryption algorithm, so that common analog input cannot be used.
The existing automatic login method for the bank security control often depends on the format characteristics of a bank security control keyboard, and different character templates need to be designed according to different keyboard formats. However, the formats of the security control keyboards of all the big banks in China are different, and even the security control keyboards of different versions in the same bank have the problem of non-uniform formats. In addition, existing methods generally rely on web page source code when locating security control locations, however, different web pages are not necessarily designed in the same language and keyword. All the problems lead the automatic login process of the bank security control to depend on manual intervention too much, and the recognition efficiency is low.
Therefore, how to improve the automatic login efficiency of the bank security control becomes an important problem to be solved urgently.
Disclosure of Invention
Aiming at the technical problems, the invention provides a method, a system, equipment and a storage medium for automatically logging in a bank security control, so as to improve the efficiency of automatically logging in the bank security control.
In order to achieve the purpose, the invention adopts the following technical scheme:
according to one aspect of the invention, an automatic login method for a bank security control is provided, which comprises the following steps: generating a security control keyboard at a web end and acquiring a webpage screenshot; constructing a network image segmentation model based on UNet, and segmenting a security control keyboard image from the webpage screenshot; segmenting all button images from the security control keyboard image, and acquiring the positions of the button images in the security control keyboard; constructing a character recognition neural network model, inputting the button image into the character recognition neural network model, and outputting to obtain characters corresponding to the button image; combining the position of the button image in the safety control keyboard with the character corresponding to the button image to construct a button-character dictionary; and searching the corresponding button position of the character to be input in the safety control keyboard according to the button-character dictionary, and triggering click operation on the button position to finish automatic login.
In an embodiment of the invention, the constructing of the UNet-based network image segmentation model; the constructed network image segmentation model comprises 4 convolution blocks, 1 full-connection layer, 1 convolution layer of 3 x 3 and 1 convolution layer of 1 x 1 which are connected in sequence; the convolution block comprises 2 convolution layers of 3 multiplied by 3 and 1 pooling layer of 2 multiplied by 2 which are connected in sequence; the fully-connected layer includes 2 1 × 1 convolutional layers.
In an embodiment of the present invention, the segmenting a security control keyboard image from the screenshot of the web page; the method comprises the following steps: inputting the webpage screenshot into the network image segmentation model; extracting keyboard image features of 4 layers of safety controls from the webpage screenshot through the 4 sequentially connected volume blocks, and down-sampling an extraction result to obtain a first feature; processing the first characteristic through the full connection layer to obtain a second characteristic; performing at least one upsampling operation on the second features until the second features are restored to the size of the original image, and performing feature extraction on the 1 convolution layer with the size of 3 x 3 to obtain third features; and processing the third characteristic through the 1 convolution layer of 1 × 1 to obtain a security control keyboard image with the same size as the original image and the number of k channels.
In an embodiment of the present invention, all button images are segmented from the security control keyboard image, and the positions of the button images in the security control keyboard are obtained; the method comprises the following steps: performing binarization processing on the security control keyboard image, wherein pixel points with the gray level less than or equal to a preset threshold value in the security control keyboard image are determined as characters, the gray level of the characters is set to be 0, other pixel points are determined as a background, and the gray level of the background is set to be 255; and adopting a bidirectional projection segmentation method to segment each button image from the safety control keyboard image, and acquiring the coordinates of the button image in the safety control keyboard.
In an embodiment of the present invention, the segmenting each button image from the security control keyboard image by using a bidirectional projection segmentation method, and acquiring coordinates of the button image in the security control keyboard, includes the following steps: scanning the binarized security control keyboard image pixel by pixel from top to bottom, counting the number of elements of each column of black pixels, and storing the elements in a column group; scanning the binarized security control keyboard image pixel by pixel from left to right, counting the number of elements of black pixels of each row, and storing the number in a row array; and determining the corresponding coordinates of the characters in each button in the safety control keyboard according to the column group and the row group so as to obtain the upper, lower, left and right boundaries and the position coordinates of each button, and segmenting each button image from the safety control keyboard image.
In an embodiment of the present invention, the constructing a character recognition neural network model; wherein the constructed character recognition neural network model comprises: a feature extraction neural network, said feature extraction neural network comprising 1 convolutional layer branch of 3 × 3 and 1 convolutional layer branch of 5 × 5, each said convolutional layer branch being connected to 1 pooling layer respectively; and the double-layer feedforward neural network is connected to the characteristic extraction neural network and is used for carrying out character recognition.
In an embodiment of the present invention, the button image is input into the character recognition neural network model, and a character corresponding to the button image is output; the method comprises the following steps: inputting the button image into the feature extraction neural network, extracting features of the button image through the 3 × 3 convolutional layer branch to obtain a feature h3 × 3, and extracting features of the button image through the 5 × 5 convolutional layer branch to obtain a feature h5 × 5; and splicing the features H3 × 3 and the features H5 × 5 to obtain a button image feature H: inputting the button image characteristics H into the double-layer feedforward neural network, outputting to obtain a classification score vector of the button image, calculating the classification score vector, and outputting the classification probability value of the button image to determine the characters corresponding to the input button image.
According to another aspect of the invention, there is provided a bank security control automatic login system, comprising
The acquisition module is used for generating a security control keyboard at a web end and acquiring a webpage screenshot;
the first segmentation module is used for constructing a network image segmentation model based on UNet and segmenting a security control keyboard image from the webpage screenshot;
the second segmentation module is used for segmenting all the button images from the security control keyboard image and acquiring the positions of the button images in the security control keyboard;
the recognition module is used for constructing a character recognition neural network model, inputting the button image into the character recognition neural network model and outputting to obtain a character corresponding to the button image;
the building module is used for building a button-character dictionary by combining the position of the button image in the safety control keyboard and the character corresponding to the button image;
and the clicking module is used for searching the corresponding button position of the character to be input in the safety control keyboard according to the button-character dictionary, triggering clicking operation on the button position and completing automatic login.
According to another aspect of the present invention, a storage medium is provided, and the storage medium stores a computer program, and the computer program is executed by a processor to perform the steps of the automatic login method for the bank security control.
According to still another aspect of the present invention, there is provided an electronic apparatus including:
a processor;
a storage medium having stored thereon a computer program which, when executed by the processor, performs the steps of the method for automatic entry of a bank security control as described above.
The invention has the beneficial effects that:
the invention overcomes the defects of excessive dependence on manual intervention, lower identification efficiency and the like of the conventional automatic login method for the bank security control, and provides the automatic login method for the bank security control based on deep learning, which has the advantages of simple model structure, low training complexity, high calculation speed and high identification rate up to 99.99%.
Additional aspects and advantages of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention.
Drawings
The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the invention and together with the description, serve to explain the principles of the invention. It is obvious that the drawings in the following description are only some embodiments of the invention, and that for a person skilled in the art, other drawings can be derived from them without inventive effort.
Fig. 1 is a schematic flow chart of an automatic login method for a bank security control according to an embodiment of the present invention.
FIG. 2 is a flow chart illustrating an exemplary process for performing auto-login in accordance with an embodiment of the present invention.
Fig. 3 is a schematic structural diagram of a UNet-based network image segmentation model in an embodiment of the present invention.
Fig. 4 is a schematic flowchart of a process of performing binarization processing on a security control keyboard image in an embodiment of the present invention.
FIG. 5 is a schematic flow chart of a bi-directional projection segmentation method according to an embodiment of the present invention.
FIG. 6 is a flow chart of a neural network model for character recognition according to an embodiment of the present invention.
Fig. 7 is a block diagram of an automatic login system for a bank security control in an embodiment of the present invention.
Fig. 8 is a schematic structural diagram of an electronic device according to an embodiment of the invention.
Fig. 9 is a schematic structural diagram of a computer-readable storage medium according to an embodiment of the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more clearly understood, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. It is to be understood that the described embodiments are merely exemplary of the invention, and not restrictive of the full scope of the invention. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.
The terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the application. As used in this application and the appended claims, the singular forms "a", "an", and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise. It should also be understood that the term "and/or" as used herein refers to and encompasses any and all possible combinations of one or more of the associated listed items.
In some of the flows described in the present specification and claims and in the above figures, a number of operations are included that occur in a particular order, but it should be clearly understood that these operations may be performed out of order or in parallel as they occur herein, with the order of the operations, e.g., S100, S200, etc., merely being used to distinguish between various operations, and the order of the operations itself does not represent any order of performance. Additionally, the flows may include more or fewer operations, and the operations may be performed sequentially or in parallel.
The method, system, device and storage medium for automatically logging in a deep learning-based bank security control claimed by the present invention will be further described in detail with reference to the accompanying drawings and specific embodiments. Fig. 1 is a schematic flowchart of a method, a system, a device, and a storage medium for automatically logging in a bank security control according to an embodiment of the present invention. FIG. 2 is a flow chart illustrating an exemplary process for performing auto-login in accordance with an embodiment of the present invention. Fig. 3 is a schematic structural diagram of a UNet-based network image segmentation model in an embodiment of the present invention. Fig. 4 is a schematic flowchart of a process of performing binarization processing on a security control keyboard image in an embodiment of the present invention. FIG. 5 is a schematic flow chart of a bi-directional projection segmentation method according to an embodiment of the present invention. FIG. 6 is a flow chart of a neural network model for character recognition according to an embodiment of the present invention. Fig. 7 is a block diagram of an automatic login system for a bank security control in an embodiment of the present invention. Fig. 8 is a schematic structural diagram of an electronic device according to an embodiment of the invention. Fig. 9 is a schematic structural diagram of a computer-readable storage medium according to an embodiment of the present invention.
According to one aspect of the invention, an automatic login method for a bank security control is provided.
As shown in fig. 1, the method for automatically logging in a bank security control based on deep learning according to the present invention may include the following steps:
s100, generating a security control keyboard at a web end and acquiring a webpage screenshot; specifically, after a bank landing page is visited and a security control keyboard is generated, the whole bank landing page is captured and stored as a picture;
s200, constructing a network image segmentation model based on UNet, and segmenting a security control keyboard image from the webpage screenshot; specifically, a network image segmentation model based on UNet is constructed, and a picture segmentation algorithm based on UNet is adopted to segment the network image from a webpage screenshot of the whole bank landing page to obtain a security control keyboard image which is stored as a picture;
s300, segmenting all button images from the security control keyboard image, and acquiring the positions of the button images in the security control keyboard;
s400, constructing a character recognition neural network model, inputting the button image into the character recognition neural network model, and outputting to obtain a character corresponding to the button image;
s500, combining the position of the button image in the safety control keyboard with the character corresponding to the button image to construct a button-character dictionary;
and S600, searching a corresponding button position of the character to be input in the safety control keyboard according to the button-character dictionary, and triggering click operation on the button position to finish automatic login. Specifically, as shown in fig. 2, it is a schematic flow chart of completing automatic login in an embodiment of the present invention. After the button-character corresponding dictionary is constructed, the position coordinates of the corresponding characters in the keyboard can be obtained according to the position of each button in the keyboard; when a password needs to be input, each character in the password string is read firstly, a method of sequentially traversing the password string can be adopted, then the button position of each character to be input of the password string in the safety control keyboard is searched through the button-character dictionary, and the click operation is triggered at the button position corresponding to the character of the password string to be input, so that the password input is completed. It should be noted that the triggering click operation is automatically completed by the browser, manual intervention is not required, a tool for testing the Selenium application program is run in the browser, and when a script is executed, the browser can automatically perform the click operation according to the script code, so that automatic access and input of a webpage can be facilitated.
The invention takes the webpage screenshot of the whole bank login page as input data, utilizes a network image segmentation model based on UNet to segment the keyboard image of the security control from the webpage screenshot of the whole bank login page, utilizes an image binarization method and a two-way projection segmentation method to automatically segment each button from the keyboard of the security control, and introduces a convolutional neural network to identify characters in the buttons, so as to obtain the corresponding position of each character in the keyboard, thereby improving the accuracy and efficiency of image segmentation and identification to the maximum extent.
In an embodiment of the present invention, a security control keyboard is generated at a web end and a web screenshot is obtained; the method comprises the following steps: automatically accessing a bank login page; automatically inputting a login user name; switching to a password input field and waiting for generation of a security control keyboard; and acquiring a webpage screenshot. In specific implementation, a tool for testing a Selenium application program can be run in a browser, a bank login page is automatically entered according to a link, a login account is input in a user name column through simulation input, then the user is switched to a password input column to wait for the completion of the keyboard loading of a security control, and a webpage screenshot is obtained. It should be noted that the security control login is a form in which the bank encrypts the user password in order to secure the user account. The safety control is a virtual keyboard, namely the safety control keyboard, the positions of all keys in the keyboard are disordered or randomly generated, a user cannot use the physical keyboard for inputting, and only can click corresponding characters through a mouse for inputting.
In one embodiment of the invention, the automatic access bank login page; the method comprises the following steps: and requesting to access the bank login page in a GET request mode by using a Selenium Webdriver tool driver browser in a Selenium frame.
The invention can automatically access the bank login page by using the Selenium frame. Specifically, the invention uses a Selenium WebDriver tool in a Selenium framework to drive a browser, and requests to access the bank login page in a GET request mode. As will be appreciated by those skilled in the art, the Selenium framework is a framework for Web application testing, the bottom layer of which uses JavaScript to simulate the operation of a user on a browser. The tool for testing the Selenium application program directly runs in the browser, and when the script is executed, the browser can automatically perform operations such as clicking, inputting, verifying and the like according to the script code, so that the automatic access of the webpage can be facilitated.
In an embodiment of the invention, the constructing of the UNet-based network image segmentation model; the constructed network image segmentation model comprises 4 convolution blocks, 1 full-connection layer, 1 convolution layer of 3 multiplied by 3 and 1 convolution layer of 1 multiplied by 1 which are connected in sequence; the convolution block comprises 2 convolution layers of 3 multiplied by 3 and 1 pooling layer of 2 multiplied by 2 which are connected in sequence; the fully-connected layer includes 2 1 × 1 convolutional layers.
In an embodiment of the present invention, the segmenting a security control keyboard image from the screenshot of the web page; the method comprises the following steps: inputting the webpage screenshot into the network image segmentation model; extracting keyboard image features of 4 layers of safety controls from the webpage screenshot through the 4 sequentially connected volume blocks, and down-sampling an extraction result to obtain a first feature; processing the first characteristic through the full connection layer to obtain a second characteristic; performing at least one upsampling operation on the second features until the second features are restored to the size of the original image, and performing feature extraction on the 1 convolution layer with the size of 3 x 3 to obtain third features; and processing the third characteristic through the 1 convolution layer of 1 × 1 to obtain a security control keyboard image with the same size as the original image and the number of k channels.
The method can use a Selenium frame to capture and store the whole bank login page as a picture, and then uses a network image segmentation model based on UNet to segment the security control keyboard from the whole bank login page. When needing to be explained, the invention uses a network image segmentation model based on UNet as a basic model, and utilizes a coding-decoding structure to segment the whole bank login page; the attention mechanism is adopted to distribute larger weight to pixels in the safe control keyboard area, and the interference of other parts in the page is reduced; finally, the security control keyboard picture is obtained by segmentation from the webpage screenshot, and by adopting the method, bank login pages do not need to be distinguished, and webpage source codes do not need to be relied on, so that the identification efficiency can be effectively improved.
In particular, the invention employs a UNet-based network image segmentation model with an attention mechanism to segment the security control keyboard image from the screenshot. Fig. 3 is a schematic structural diagram of a UNet-based network image segmentation model according to an embodiment of the present invention. Inputting the webpage screenshot into the network image segmentation model, firstly extracting features through a multilayer convolutional neural network, wherein the multilayer convolutional neural network comprises 4 convolutional blocks which are sequentially connected, and each convolutional block comprises 2 convolutional layers of 3 x 3 and 1 pooling layer of 2 x 2 which are sequentially connected; the present invention extracts the features of each layer using 2 3 x 3 convolutional layers and 12 x 2 pooling layer. In one embodiment of the present invention, the number of filters used in each layer of the feature extraction of the rolling block is increased gradually, and is 64, 128, 256 and 512. Let initial input be x0Then, the feature extraction calculation formula of the ith layer is:
ci＝conv3*3(conv3*3(xi))#(1)
xi+1＝maxpooling2*2(ci)#(2)
and after feature extraction and downsampling of the same structure are carried out for four times through the multilayer convolution neural network, a first feature is obtained. The fully-connected layer processes the first feature to obtain a second feature, specifically, in an embodiment of the present invention, the fully-connected layer processes the extracted feature by using 2 convolution layers of 1 × 1 to obtain the second feature, where the formula is as follows:
d4＝conv1*1(con v1*1(x4))#(3)
then, performing at least one upsampling operation on the second features, wherein the upsampling operation can be performed by a bilinear interpolation method, in the upsampling process, the result of upsampling on each layer is fused with the features of the corresponding layer number in the four-time feature extraction performed by the multilayer convolutional neural network, the fusion mode is that an attention weight is calculated by using the result of upsampling and the feature map of the corresponding layer number in the four-time feature extraction, then the feature after attention weight distribution is obtained by multiplying the result of upsampling and the feature map, then the feature is connected with the result of upsampling in series, the upsampling process is repeated until the second features are restored to the size of the original image, and then the feature extraction is performed by using a convolution layer of 3 x 3 to obtain third features; wherein the calculation formula in the up-sampling process is as follows:
gi＝upsampling(di+1)#(4)
di＝conv3*3([gi；ai(gi，ci)])#(5)
and finally, processing the third characteristic through 1 convolution layer of 1 multiplied by 1 to obtain a security control keyboard image with the same size as the original image and the number of k channels. Wherein k is the number of categories to be classified, each pixel point on the graph is classified by utilizing softmax, whether the pixel point is a table area or not is judged, and the formula is as follows:
C＝softmax(conv1*1(d0))#(6)
it should be noted that, except for the field of security control keyboards, the background and other parts of the screenshot of the web page are not the areas concerned by the present invention, so that in the process of feature fusion of the up-sampling, the attention mechanism is adopted to set a threshold to filter the features in the feature extraction stage, so that the model can allocate more weight to the security control keyboard areas. In the attention mechanism layer, firstly, the ith layer characteristic diagram g obtained by up-sampling in the decoding process is obtainediThe ith layer feature map c corresponding to the above four times of feature extractioniPerforming cascade connection, changing the output filter into 1 by a convolution layer of 1 × 1, and mapping to [0, 1 ] by a sigmoid activation function]And c is compared withiThe values of (b) are multiplied element by element and output. Note that the output of the power machine module will be in the network with giTo carry outAnd connecting in series and sending to the next convolutional neural network. The calculation process of the attention mechanism is as follows;
ti＝ReLU(con v1*1([gi；ci]))#(7)
Ai＝σ(conv1*1(ti))#(8)
Figure BDA0003575260540000091
the invention can finally learn the shape of the keyboard region of the safety control through the coding-decoding process of the UNet model with attention.
In an embodiment of the present invention, the image of all the buttons is segmented from the image of the security control keyboard, and the positions of the button images in the security control keyboard are obtained; the method comprises the following steps: performing binarization processing on the security control keyboard image, wherein pixels with the gray level less than or equal to a preset threshold value in the security control keyboard image are judged as characters, the gray level of the characters is set to be 0, other pixel points are judged as a background, and the gray level of the background is set to be 255; and adopting a bidirectional projection segmentation method to segment each button image from the safety control keyboard image, and acquiring the coordinates of the button image in the safety control keyboard.
In an embodiment of the present invention, the segmenting each button image from the security control keyboard image by using a bidirectional projection segmentation method, and acquiring coordinates of the button image in the security control keyboard, includes the following steps: scanning the binarized security control keyboard image pixel by pixel according to the sequence from top to bottom, counting the number of elements of black pixels in each column, and storing the elements in a column group; scanning the binarized security control keyboard image pixel by pixel according to the sequence from left to right, counting the number of elements of black pixels of each row, and storing the number in a row array; and determining the corresponding coordinates of the characters in each button in the safety control keyboard according to the row array and the column array so as to obtain the upper, lower, left and right boundaries and the position coordinates of each button, and segmenting each button image from the safety control keyboard image.
The invention can adopt binarization and two-way projection algorithms to segment each button from the keyboard image of the safety control. Specifically, as shown in fig. 4, it is a flowchart of performing binarization processing on a security control keyboard image in an embodiment of the present invention. According to the method, after the security control keyboard picture is obtained through network image segmentation based on UNet, the security control keyboard picture to be processed is scanned pixel by pixel, the gray parameter threshold value can be set to be 10 according to needs, pixel points with the gray level smaller than or equal to 10 in the security control keyboard picture are judged to be characters, the gray value of the characters is set to be 0, other pixel points are judged to be a background, the gray value of the background is set to be 255, and therefore the security control keyboard picture after binarization processing can be obtained. Then, a bidirectional projection segmentation method is adopted to segment each button image from the security control keyboard image, and coordinates of the button image in the security control keyboard are obtained, specifically, as shown in fig. 5, the method is a schematic flow chart of the bidirectional projection segmentation method in one embodiment of the present invention, the security control keyboard image after binarization is scanned pixel by pixel according to a sequence from top to bottom, the number of elements of each column of black pixels is counted, and the counted number is stored in a column group; scanning the binarized security control keyboard image pixel by pixel according to the sequence from left to right, counting the number of elements of black pixels of each row, and storing the number in a row array; and determining the corresponding coordinates of the characters in each button in the safety control keyboard according to the row array and the column array so as to obtain the upper, lower, left and right boundaries and the position coordinates of each button, and segmenting each button image from the safety control keyboard image.
In an embodiment of the present invention, the constructing a character recognition neural network model; wherein the constructed character recognition neural network model comprises: a feature extraction neural network, said feature extraction neural network comprising 1 convolutional layer branch of 3 × 3 and 1 convolutional layer branch of 5 × 5, each said convolutional layer branch being connected to 1 pooling layer respectively; and the double-layer feedforward neural network is connected to the characteristic extraction neural network and is used for carrying out character recognition.
In an embodiment of the present invention, the button image is input into the character recognition neural network model, and the character in the button image is output; the method comprises the following steps: inputting the button image into the feature extraction neural network, and extracting the features of the button image through the 3 x 3 convolutional layer branch to obtain the features h3*3And extracting the characteristics of the button image through the 5 multiplied by 5 convolutional layer branch to obtain the characteristics h5*5(ii) a For the feature h3*3And the characteristic h5*5And splicing to obtain a button image characteristic H: inputting the button image characteristics H into the double-layer feedforward neural network, outputting to obtain a classification score vector of the button image, calculating the classification score vector by a ReLU activation function, and outputting the classification probability value of the button image to confirm the input characters corresponding to the button image.
The invention can adopt a convolution neural network model, namely the characteristic extraction neural network to extract image characteristics, and then uses a double-layer feedforward neural network to carry out character recognition, namely can carry out character recognition on the button image containing the characters after being segmented so as to determine the corresponding characters in each button. Specifically, as shown in fig. 6, it is a schematic flow chart of the character recognition neural network model in an embodiment of the present invention. The constructed character recognition neural network model comprises the following steps: a feature extraction neural network, said feature extraction neural network comprising 1 convolutional layer branch of 3 × 3 and 1 convolutional layer branch of 5 × 5, each said convolutional layer branch being connected to 1 pooling layer respectively; and the double-layer feedforward neural network is connected to the characteristic extraction neural network and is used for carrying out character recognition. Inputting the button image into the feature extraction neural network, and extracting the features of the button image through the 3 x 3 convolutional layer branch to obtain the features h3*3Extracting the characteristics of the button image through the 5 multiplied by 5 convolutional layer branch to obtain the characteristics h5*5(ii) a In the specific implementation, in the button image feature extraction stage, the button image input to be processed firstly passes through a feature extraction device comprising 2 convolution layers and corresponding pooling layersCarrying out feature extraction through a network, wherein the feature extraction neural network comprises 1 convolutional layer branch of 3 multiplied by 3 and 1 convolutional layer branch of 5 multiplied by 5, each convolutional layer branch is respectively connected with 1 pooling layer, and the convolution process is as follows:
xi＝maxpooling(conv(conv(xi-1)))#(15)
wherein x isi-1For the input of the i-th layer picture, xiThe output, representing the ith convolutional neural network, is a three-dimensional tensor.
Splicing the features from two convolution kernels with different sizes in the last layer of the convolution layer, namely, splicing the features h3*3And the characteristic h5*5And splicing to finally obtain the button image characteristics H:
Figure BDA0003575260540000121
wherein h is3*3Representing a feature from a convolutional layer branch of size 3 × 3 of the convolutional kernel, h5*5Representing features from convolutional layer branches with a convolutional kernel size of 5 x 5,
Figure BDA0003575260540000122
a series operation of the representation tensors.
Inputting the button image characteristics H into the double-layer feedforward neural network for character recognition to obtain a fractional vector s of each class of characters in the picture, namely outputting the classified fractional vector s of the button image,
s＝W2(σ(W1·H))
wherein, W1And W2Is a learnable weight parameter, and σ is an activation function.
In one possible embodiment, the classification score vector is operated by using a ReLU activation function, and the classification probability value of the button image is output to confirm the input character corresponding to the button image. That is, the probability of identifying the character as each class is obtained by the score vector s through the softmax function, and the formula is as follows:
Figure BDA0003575260540000123
output probability OiAnd the largest class is used as a character recognition result to confirm the input characters corresponding to the button images.
The present invention may extract features in the button image using a convolutional neural network having a plurality of different sized convolution kernels that may extract different granularities of picture features. Specifically, the button image is subjected to feature extraction through the 3 × 3 convolutional layer branch to obtain a feature h3*3Extracting the characteristics of the button image through the 5 multiplied by 5 convolutional layer branch to obtain the characteristics h5*5. And after feature extraction, obtaining a series of vectors H containing the button image features, classifying the H by adopting a double-layer feedforward neural network, and obtaining a button-character dictionary by corresponding the class with the maximum classification probability value to the class to which the characters in the button belong.
According to another aspect of the invention, a bank security control automatic login system is provided. Fig. 7 is a block diagram of an automatic login system for a bank security control in an embodiment of the present invention. As shown in fig. 7, the system 500 for automatically logging in a bank security control in this embodiment includes an obtaining module 510, a first dividing module 520, a second dividing module 530, an identifying module 540, a building module 550, and a clicking module 560. The obtaining module 510 is configured to generate a security control keyboard at the web end and obtain a screenshot of a web page. The first segmentation module 520 is configured to construct a network image segmentation model based on UNet, and segment a security control keyboard image from the screenshot; the second segmentation module 530 is configured to segment all button images from the security control keyboard image, and obtain the positions of the button images in the security control keyboard; the recognition module 540 is configured to construct a character recognition neural network model, input the button image into the character recognition neural network model, and output a character corresponding to the button image; the building module 550 is configured to build a button-character dictionary by combining the position of the button image in the security control keyboard and the character corresponding to the button image; and the clicking module 560 is configured to search for a corresponding button position of the character to be input in the security control keyboard according to the button-character dictionary, and trigger a clicking operation on the button position to complete automatic login.
According to the invention, the webpage screenshot of the whole bank login page is acquired as input data through the acquisition module 510, the security control keyboard image is segmented from the webpage screenshot of the whole bank login page through the first segmentation module 520, each button is automatically segmented from the security control keyboard through the second segmentation module 530, the characters in the buttons are identified through the identification module 540, the corresponding position of each character in the keyboard is obtained, and the accuracy and efficiency of image segmentation and identification can be improved to the maximum extent. The button-character dictionary is constructed by the construction module 550, and the click module 560 automatically inputs the password according to the button-character dictionary. The method can automatically segment and identify characters in different security control keyboards during implementation, does not need to distinguish bank login pages and templates, can meet the requirement of automatic login of security controls of banks with different formats, simultaneously supports security spaces of various large banks and various versions, does not need manual intervention, and has the advantages of high efficiency and accuracy.
According to still another aspect of the present invention, there is provided an electronic apparatus including:
a processor;
a storage medium having stored thereon a computer program which, when executed by the processor, performs the steps of the method for automatic entry of a bank security control as described above.
An electronic device 600 in one embodiment in accordance with the application is described below with reference to fig. 8. The electronic device 600 shown in fig. 8 is only an example, and should not bring any limitation to the functions and the scope of use of the embodiments of the present application.
As shown in fig. 8, the electronic device 600 is embodied in the form of a general purpose computing device. The components of device 600 may include, but are not limited to: at least one processor 610, at least one memory unit 620, a bus 630 connecting the various system components (including the memory unit 620 and the processor 610), a display unit 640, and the like.
Wherein the storage unit stores program code executable by the processor 610 to cause the processor 610 to perform the steps according to various exemplary embodiments of the present application described in the above-mentioned electronic prescription flow processing method section of the present specification. For example, the processor 610 may perform the steps in the above method.
The storage unit 620 may include storage media in the form of volatile storage units, such as a random access memory unit (RAM)6201 and/or a cache storage unit 6202, and may further include a read-only memory unit (ROM) 6203.
The memory unit 620 may also include a program/utility 6204 having a set (at least one) of program modules 6205, such program modules 6205 including, but not limited to: an operating system, one or more application programs, other program modules, and program data, each of which, or some combination thereof, may comprise an implementation of a network environment.
Bus 630 may be one or more of several types of bus structures, including a memory unit bus or memory unit controller, a peripheral bus, an accelerated graphics port, a processor, or a local bus using any of a variety of bus architectures.
The electronic device 600 may also communicate with one or more external devices 700 (e.g., keyboard, pointing device, bluetooth device, etc.), with one or more devices that enable a tenant to interact with the electronic device 600, and/or with any devices (e.g., router, modem, etc.) that enable the electronic device 600 to communicate with one or more other computing devices. Such communication may occur via input/output (I/O) interface 650. Also, the electronic device 600 may communicate with one or more networks (e.g., a Local Area Network (LAN), a Wide Area Network (WAN) and/or a public network, such as the Internet) via the network 102 adapter 660. The network adapter 660 may communicate with the other modules of the electronic device 600 via the bus 630. It should be understood that although not shown in the figures, other hardware and/or software modules may be used in conjunction with the electronic device 600, including but not limited to: microcode, device drivers, redundant processors, external disk drive arrays, RAID systems, tape drives, and data backup storage systems, among others.
The method can automatically segment and identify characters in different security control keyboards during implementation, does not need to distinguish bank login pages and templates, can meet the requirement of automatic login of security controls of banks with different formats, simultaneously supports security spaces of various large banks and various versions, does not need manual intervention, and has the advantages of high efficiency and accuracy.
According to another aspect of the present invention, a storage medium is provided, and the storage medium stores a computer program, and the computer program is executed by a processor to perform the steps of the automatic login method for the bank security control.
Referring to fig. 9, in one embodiment, a program product 800 for implementing the bank security control auto-login method may employ a portable compact disc read only memory (CD-ROM) and include program code, and may be run on a server. However, those skilled in the art will appreciate that the program product referred to herein is not limited to such, but may be any tangible medium that can contain, or store a program for use by or in connection with an instruction execution system, apparatus, or device.
It should be noted that the present application may be implemented in software and/or a combination of software and hardware, for example, as an Application Specific Integrated Circuit (ASIC), a general purpose computer or any other similar hardware device. In one embodiment, the software programs of the present application may be executed by a processor to implement the above steps or functions. Likewise, the software programs (including associated data structures) of the present application may be stored in a computer readable recording medium, such as RAM memory, magnetic or optical drive or diskette and the like. Additionally, some of the steps or functions of the present application may be implemented in hardware, for example, as circuitry that cooperates with the processor to perform various steps or functions.
In addition, some of the present application can be applied as a computer program product, such as computer program instructions, which when executed by a computer, can invoke or provide the method or technical solution according to the present application through the operation of the computer. Those skilled in the art will appreciate that the forms of computer program instructions that reside on a computer storage medium include, but are not limited to, source files, executable files, installation package files, and the like, and that the manner in which the computer program instructions are executed by a computer includes, but is not limited to: the computer directly executes the instruction, or the computer compiles the instruction and then executes the corresponding compiled program, or the computer reads and executes the instruction, or the computer reads and installs the instruction and then executes the corresponding installed program. In this regard, computer storage media may be any available computer readable storage media or communication media that can be accessed by a computer.
Communication media includes media by which communication signals, including, for example, computer readable instructions, data structures, program modules, or other data, are transmitted from one system to another. Communication media may include conductive transmission media such as cables and wires, e.g., fiber optics, coaxial, and the like, and wireless (non-conductive transmission) media capable of propagating energy waves such as acoustic, electromagnetic, RF, microwave, and infrared. Computer readable instructions, data structures, program modules, or other data may be embodied in a modulated data signal, for example, in a wireless medium such as a carrier wave or similar mechanism such as is embodied as part of spread spectrum techniques. The term "modulated data signal" means a signal that has one or more of its characteristics set or changed in such a manner as to encode information in the signal. The modulation may be analog, digital or hybrid modulation techniques.
By way of example, and not limitation, computer-readable storage media may include volatile and nonvolatile, removable and non-removable media implemented in any method or technology for storage of information such as computer-readable instructions, data structures, program modules or other data. For example, computer-readable storage media include, but are not limited to, volatile memory such as random access memory (RAM, DRAM, SRAM); and non-volatile memory such as flash memory, various read-only memories (ROM, PROM, EPROM, EEPROM), magnetic and ferromagnetic/ferroelectric memories (MRAM, FeRAM); and magnetic and optical storage devices (hard disk, tape, CD, DVD); or other now known media or later developed that can store computer-readable information or data for use by a computer system.
In summary, the invention overcomes the defects of excessive dependence on manual intervention, low recognition efficiency and the like of the existing bank security control automatic login method, and provides the bank security control automatic login method based on deep learning, which has the advantages of simple method model structure, low training complexity, high calculation speed and high recognition rate up to 99.99%.
The foregoing is a more detailed description of the invention in connection with specific preferred embodiments and it is not intended that the invention be limited to these specific details. For those skilled in the art to which the invention pertains, several simple deductions or substitutions can be made without departing from the spirit of the invention, and all shall be considered as belonging to the protection scope of the invention.

Claims (10)

1. The automatic login method of the bank security control is characterized by comprising the following steps:
generating a security control keyboard at a web end and acquiring a webpage screenshot;
constructing a network image segmentation model based on UNet, and segmenting a security control keyboard image from the webpage screenshot;
segmenting all button images from the security control keyboard image, and acquiring the positions of the button images in the security control keyboard;
constructing a character recognition neural network model, inputting the button image into the character recognition neural network model, and outputting to obtain characters corresponding to the button image;
combining the position of the button image in the safety control keyboard with the character corresponding to the button image to construct a button-character dictionary;
and searching the corresponding button position of the character to be input in the safety control keyboard according to the button-character dictionary, and triggering click operation on the button position to finish automatic login.
2. The automatic login method of bank security control according to claim 1, wherein said building UNet based network image segmentation model; the constructed network image segmentation model comprises 4 convolution blocks, 1 full-connection layer, 1 convolution layer of 3 x 3 and 1 convolution layer of 1 x 1 which are connected in sequence; the convolution block comprises 2 convolution layers of 3 multiplied by 3 and 1 pooling layer of 2 multiplied by 2 which are connected in sequence; the fully-connected layer includes 2 1 × 1 convolutional layers.
3. The method for automatically logging in a bank security control according to claim 2, wherein the keyboard image of the security control is segmented from the screenshot of the web page; the method comprises the following steps:
inputting the webpage screenshot into the network image segmentation model;
extracting keyboard image features of 4 layers of safety controls from the webpage screenshot through the 4 sequentially connected volume blocks, and down-sampling an extraction result to obtain a first feature;
processing the first characteristic through the full connection layer to obtain a second characteristic;
performing at least one upsampling operation on the second features until the second features are restored to the size of the original image, and performing feature extraction on the 1 convolution layer with the size of 3 x 3 to obtain third features;
and processing the third feature through the 1 convolution layer of 1 × 1 to obtain a security control keyboard image with the same size as the original image and the number of k channels.
4. The automatic login method for the bank security control according to claim 1, wherein the button images are all segmented from the security control keyboard image, and the positions of the button images in the security control keyboard are obtained; the method comprises the following steps:
performing binarization processing on the security control keyboard image, wherein pixel points of which the gray scale is less than or equal to a preset threshold value in the security control keyboard image are judged as characters, the gray scale value of the characters is set to be 0, other pixel points are judged as a background, and the gray scale value of the background is set to be 255;
and adopting a bidirectional projection segmentation method to segment each button image from the safety control keyboard image, and acquiring the coordinates of the button image in the safety control keyboard.
5. The automatic login method of the bank security control according to claim 4, wherein the step of segmenting each button image from the security control keyboard image by using a two-way projection segmentation method and obtaining the coordinates of the button image in the security control keyboard comprises the following steps:
scanning the binarized security control keyboard image pixel by pixel from top to bottom, counting the number of elements of each black pixel column, and storing the number of elements in a column group;
scanning the binarized security control keyboard image pixel by pixel from left to right, counting the number of elements of black pixels of each row, and storing the elements in a row array;
and determining the corresponding coordinates of the characters in each button in the safety control keyboard according to the column group and the row group so as to obtain the upper, lower, left and right boundaries and the position coordinates of each button, and segmenting each button image from the safety control keyboard image.
6. The automatic login method for the bank security control according to claim 1, wherein the building of a character recognition neural network model; wherein the constructed character recognition neural network model comprises: a feature extraction neural network, said feature extraction neural network comprising 1 convolutional layer branch of 3 × 3 and 1 convolutional layer branch of 5 × 5, each said convolutional layer branch being connected to 1 pooling layer respectively; and the double-layer feedforward neural network is connected to the characteristic extraction neural network and is used for carrying out character recognition.
7. The automatic login method for the bank security control according to claim 6, wherein the button image is input into the character recognition neural network model and output to obtain the character corresponding to the button image; the method comprises the following steps:
inputting the button image into the feature extraction neural network, and extracting the features of the button image through the 3 x 3 convolutional layer branch to obtain the features h3*3Extracting the characteristics of the button image through the 5 multiplied by 5 convolutional layer branch to obtain the characteristics h5*5(ii) a For the feature h3*3And the characteristic h5*5And splicing to obtain a button image characteristic H:
inputting the button image features H into the double-layer feedforward neural network, outputting to obtain classification score vectors of the button images, calculating the classification score vectors, and outputting the classification probability values of the button images to determine characters corresponding to the input button images.
8. A bank security control automatic login system is characterized by comprising
The acquisition module is used for generating a security control keyboard at a web end and acquiring a webpage screenshot;
the first segmentation module is used for constructing a network image segmentation model based on UNet and segmenting a security control keyboard image from the webpage screenshot;
the second segmentation module is used for segmenting all the button images from the security control keyboard image and acquiring the positions of the button images in the security control keyboard;
the recognition module is used for constructing a character recognition neural network model, inputting the button image into the character recognition neural network model and outputting to obtain a character corresponding to the button image;
the building module is used for building a button-character dictionary by combining the position of the button image in the safety control keyboard and the character corresponding to the button image;
and the clicking module is used for searching the corresponding button position of the character to be input in the safety control keyboard according to the button-character dictionary, triggering clicking operation on the button position and completing automatic login.
9. A storage medium having stored thereon a computer program for performing the steps of the method for automatic entry of a bank security control according to any one of claims 1 to 7 when executed by a processor.
10. An electronic device, characterized in that the electronic device comprises:
a processor;
storage medium having stored thereon a computer program for performing the steps of the method for automatic login by a bank security control according to any one of claims 1 to 7 when executed by the processor.
CN202210331616.1A 2022-03-31 2022-03-31 Automatic login method, system, equipment and storage medium for bank security control Pending CN114743187A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202210331616.1A CN114743187A (en) 2022-03-31 2022-03-31 Automatic login method, system, equipment and storage medium for bank security control

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202210331616.1A CN114743187A (en) 2022-03-31 2022-03-31 Automatic login method, system, equipment and storage medium for bank security control

Publications (1)

Publication Number Publication Date
CN114743187A true CN114743187A (en) 2022-07-12

Family

ID=82279028

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202210331616.1A Pending CN114743187A (en) 2022-03-31 2022-03-31 Automatic login method, system, equipment and storage medium for bank security control

Country Status (1)

Country Link
CN (1) CN114743187A (en)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115905767A (en) * 2023-01-07 2023-04-04 珠海金智维信息科技有限公司 Webpage login method and system based on fixed candidate box target detection algorithm
CN118247572A (en) * 2024-04-11 2024-06-25 阿谱斯(上海)通信技术有限公司 Intelligent digital keyboard pressing system and method based on AI deep learning

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115905767A (en) * 2023-01-07 2023-04-04 珠海金智维信息科技有限公司 Webpage login method and system based on fixed candidate box target detection algorithm
CN118247572A (en) * 2024-04-11 2024-06-25 阿谱斯(上海)通信技术有限公司 Intelligent digital keyboard pressing system and method based on AI deep learning

Similar Documents

Publication Publication Date Title
US10650042B2 (en) Image retrieval with deep local feature descriptors and attention-based keypoint descriptors
EP3844669A1 (en) Method and system for facilitating recognition of vehicle parts based on a neural network
CN108345827B (en) Method, system and neural network for identifying document direction
CN113902926A (en) General image target detection method and device based on self-attention mechanism
CN112633459B (en) Method for training neural network, data processing method and related device
CN105938559A (en) Digital image processing using convolutional neural networks
CN109034206A (en) Image classification recognition methods, device, electronic equipment and computer-readable medium
CN112257578B (en) Face key point detection method and device, electronic equipment and storage medium
CN114743187A (en) Automatic login method, system, equipment and storage medium for bank security control
CN108681746A (en) A kind of image-recognizing method, device, electronic equipment and computer-readable medium
CN114418030B (en) Image classification method, training method and device for image classification model
EP4350575A1 (en) Image classification method and related device thereof
CN109409504A (en) A kind of data processing method, device, computer and storage medium
CN115222998B (en) Image classification method
CN116363037B (en) Multi-mode image fusion method, device and equipment
EP4318322A1 (en) Data processing method and related device
CN114241411B (en) Counting model processing method and device based on target detection and computer equipment
CN111444802A (en) Face recognition method and device and intelligent terminal
CN114282258A (en) Screen capture data desensitization method and device, computer equipment and storage medium
CN116861262B (en) Perception model training method and device, electronic equipment and storage medium
CN114581177B (en) Product recommendation method, device, equipment and storage medium
CN116452802A (en) Vehicle loss detection method, device, equipment and storage medium
CN116992937A (en) Neural network model restoration method and related equipment
Fei et al. A GNN Architecture With Local and Global-Attention Feature for Image Classification
CN113487374A (en) Block E-commerce platform transaction system based on 5G network

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination