Multi-service access intelligent power terminal system

Abstract

The invention discloses a multi-service access intelligent power terminal system, which relates to the technical field of power service terminals and comprises an interactive perception module, a service acceptance module, an information service and payment module and a system management module. The interactive sensing module is used for receiving a touch instruction, displaying interface information and voice interaction to a user and receiving a collected sensing signal; the business acceptance module is used for accepting a power business application initiated by a user and processing a power business process; the information service and payment module is used for providing information query service for the user, processing electric charge and paying application business charge; and the system management module is used for terminal self-starting and self-checking, software and hardware state monitoring, system log recording, software version management and security encryption processing.

Description

Technical Field

[0001] The present invention relates to the technical field of power service terminals, and in particular to a smart power terminal system with multi-service access. Background Art

[0002] In recent years, digital transformation and a user-centric approach have been key development priorities for many traditional industries, including the power grid services industry. As users' expectations for a better public service experience increase, State Grid's traditional offline service model faces challenges in terms of responsiveness and coverage. As the core operator of electric power infrastructure, State Grid has gradually built and improved an intelligent, intensive, and user-friendly terminal service system to meet society's demand for more efficient power services. However, the transition from a business-centric to a user-centric service model requires intelligent and practical terminal systems. Developing more adaptable terminal systems is a promising application direction.

[0003] At present, the Chinese invention patent application with application number CN201910450745.0 discloses a vertical online State Grid cloud service intelligent terminal device, which includes: it can be placed in the State Grid power business hall to replace the manual window to provide electricity services to users. The external shell of the device is divided into three parts from top to bottom: display area, operation area and support area. The display area and the support area are on the same plane, and the operation area protrudes outward. The inside of the shell is provided with a control mainboard and a functional module. The control mainboard and the functional module are connected through a bus. The outside of the shell is provided with a human-computer interaction module, a monitoring module, a voice interaction module, an information reading module and a payment module. It also includes an external connection module, which is a 4G communication module installed with a SIM card. The present invention has a high degree of self-service function, is connected to the online State Grid cloud data, can complete all business operations of the manual window, is not affected by time, and can provide 24-hour service, which greatly meets user needs.

[0004] However, this application can only be placed in the State Grid Power Business Office to provide services, with large restrictions on the area of ​​use, and relies on State Grid cloud encryption, and the local terminal security is insufficient. Summary of the Invention

[0005] The technical problem solved by the present invention is that the application scenarios of existing State Grid power service terminals are usually limited to State Grid business halls. Users still need to rush to the State Grid business hall to handle power business, which has the problem of high time cost. If the existing State Grid power service terminals are placed outside the State Grid business hall, there are problems such as excessive size and high energy consumption burden, and at the same time, insufficient consideration is given to the risks of data leakage and data attacks. In addition, compared with mature functions such as face recognition and AI voice recognition, the high-definition scanners equipped in existing State Grid power service terminals are greatly affected by light when in use, and the scanning effect of paper documents is not good.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions: A multi-service access smart power terminal system, comprising: Interactive perception module, business acceptance module, information service and payment module and system management module; The interactive perception module includes a user interface unit, a voice interaction unit, and a multimodal input perception unit, which are used to receive touch commands, display interface information to the user, perform voice interaction, and receive collected perception signals; The business acceptance module includes an account management unit, a basic business management unit, and a special business management unit, and is used to accept power business applications initiated by users and process power business processes; The information service and payment module includes an information inquiry service unit, an electricity fee payment processing unit, and a publicity and display unit, which are used to provide information inquiry services to users and process the payment of electricity fees and application fees; The system management module includes a system operation monitoring unit, a software operation and maintenance management unit, and a security encryption unit, which are used for terminal self-startup and self-test, software and hardware status monitoring, system log recording, software version management, and security encryption processing.

[0007] Preferably, the interaction perception module includes a user interface unit, a voice interaction unit and a multimodal input perception unit, which are used to receive touch instructions, display interface information to the user, voice interaction and receive collected perception signals; The user interface unit is used to display information on the 14-inch interactive screen and receive user touch commands through the interactive screen; The interface information includes business processes, query results, promotional content and payment interface; The voice interaction unit is used to receive and analyze the user's voice input through the microphone end of the sound hole in the AI ​​voice recognition array, and play sound prompts to the user through the speaker end of the sound hole.

[0008] Preferably, the multimodal input perception unit is used to receive perception signals collected by the physical input device of the terminal; The physical input devices include a binocular camera, a barcode / QR code scanner, an ID card reader, and a foldable high-definition camera; The user's facial data is collected through a face recognition binocular camera, the barcode information provided by the user is read through a barcode / QR code scanner, the user's ID card information is read through an ID card reader, and the paper document information provided by the user is obtained through a foldable high-definition high-definition scanner; The barcode information includes the account number QR code, online banking payment QR code, payment notice barcode, business acceptance form number barcode and contract number; The paper document information includes identity documents, address ownership certificate, business authorization letter, business application form and bank account information.

[0009] Preferably, the foldable high-definition scanner is a scanner with an improved OCR module. The processing logic for obtaining paper document information provided by the user through the foldable high-definition scanner includes: A foldable high-definition camera captures document images. The initial image data is analyzed with the UDoc-GAN model to obtain the camera's LED fill light parameters. Light compensation is performed based on the camera's LED fill light parameters. After light compensation, the RGB camera captures the fill light image data. The processing flow includes: Initial image data is obtained by shooting with a foldable high-definition camera. The LED fill light parameters of the camera are numerically normalized and encoded to obtain a fill light parameter feature vector. Dimension expansion is performed through a fully connected layer to obtain a two-dimensional fill light constraint feature map with the same size as the initial image data. The LED fill light parameters of the camera include the current color temperature, brightness intensity and shooting angle of the LED fill light; The cross-modal attention layer in the UDoc-GAN model interacts with the initial image data and the two-dimensional fill-light constraint feature map to obtain a fused feature map. The first convolutional layer calculates the initial image features on the initial image data, and the fully connected layer calculates the query vector (Q), key vector (K), and value vector (V) on the two-dimensional fill-light constraint feature map. The attention weights are calculated based on the query vector, key vector, and value vector. The fused feature map is then weighted element-by-element on the initial image features based on the attention weights. The fused feature map is downsampled by the encoder: the resolution is halved each time through the convolution layer and the downsampling of step 2, the number of channels is doubled, and the LeakyReLU function is activated to obtain shallow features and deep features; The decoder performs upsampling on the fused feature map: the deep features are restored to resolution through the transposed convolution layer, the upsampled features are concatenated with the shallow features of the encoder at the same resolution through skip connections, and the illumination correction features are obtained through convolution, instance normalization, and LeakyReLU processing with residual blocks. The illumination correction features are finally upsampled, and the transposed convolution is used to restore the illumination correction features to the input image size. The number of channels is compressed to 3 through 1×1 convolution, and the feature value mapping is completed through the tanh activation function. The pixel value is linearly transformed to [0, 255] to obtain the illumination enhanced image.

[0010] The lighting enhanced image is an image with shadows and reflections removed and text edges and contrast enhanced.

[0011] Preferably, the illumination enhanced image is grayscaled and Gaussian filtered to obtain a grayscale denoised image, and the grayscale denoised image is subjected to line recognition by a Hough transform layer of HoughNet to obtain a line segment data set; The line segment data set includes the starting point coordinates, end point coordinates and corresponding confidence of each line segment; Filter the segments according to the confidence level of the segment data set, perform angle correction on the segments that are close to horizontal or vertical, merge the collinear segments and calculate the intersection points of all horizontal and vertical segments to obtain the initial grid points. Identifying a minimum rectangular area based on the initial grid points to generate a cell, and extracting the bounding box coordinates of the cell based on the line segment data set; Remove duplicate cells and sort them in logical order of text to obtain a cell coordinate grid, where the text logical order is from top to bottom and from left to right; Based on high-quality images and cell coordinate grids, the OCR engine is used to identify the OCR text content and coordinates of each cell. The GNN graph neural network is used to map the OCR text content, logical row, column index, merge cell flag, and field name corresponding to each cell to obtain structured text table data in JSON format.

[0012] Preferably, the business acceptance module includes an account management unit, a basic business management unit and a special business management unit, which are used to accept power business applications initiated by users and process power business processes; The account management unit is responsible for user registration, login, and real-name authentication on the State Grid Cloud, and provides users with household head identity verification, account transfer operations, account cancellation operations, identity information changes, account division / merger, and associated account number binding / unbinding services; The basic service management unit is responsible for handling basic electricity services initiated by users, including new installations, capacity expansion / reduction, temporary electricity application, suspension / resumption of electricity use, smart meter fee control agreement signing, electricity price plan changes, electronic invoice subscriptions, and SMS reminder service activation / cancellation; The special business management unit is responsible for handling special electricity business initiated by users and guiding users through the approval or survey process. The special electricity business includes distributed photovoltaic grid connection, charging pile registration, electric heating dedicated line registration, historical electricity fee dispute review, fault repair upgrade processing and electricity support qualification approval.

[0013] Preferably, the information service and payment module includes an information query service unit, an electricity fee payment processing unit and a publicity and display unit, which are used to provide information query services to users and process the payment of electricity fees and application fees; The information query service unit is used to query and display the user's service records, electricity bills, payment records, energy efficiency bill analysis, power grid maintenance and power start and stop records according to the user's request; The electricity bill payment processing unit is used to complete the user's electricity bill payment and service fee payment operations by identifying the payment code through a barcode / QR code scanner, recording the payment results and generating receipts; The publicity and display unit is responsible for managing and displaying State Grid's publicity information, including State Grid Corporation news, promotional activities, business change notifications and electricity policy introductions.

[0014] Preferably, the system management module includes a system operation monitoring unit, a software operation and maintenance management unit, and a security encryption unit, which are responsible for terminal self-startup and self-test, software and hardware status monitoring, system log recording, software version management, and security encryption processing; The system operation monitoring unit is responsible for the terminal's startup and self-test, continuously monitors the health status of hardware and software during system operation, and records system operation logs, user operation logs, and exception logs; The software operation and maintenance management unit is responsible for monitoring the version of terminal application software and controlling the update and rollback operations of terminal application software.

[0015] Preferably, the security encryption unit is used to encrypt plaintext business data to obtain a secure transmission message, and the processing logic includes: By performing SM4-CTR encryption on the plaintext service data, a key stream is obtained based on the randomly generated 128-bit SM4 session key, initialization vector IV, and counter Counter. The plaintext service data is XORed with the key stream to obtain the encrypted service data. The plain text business data includes user operation data output by the interactive perception module, the business acceptance module and the information service and payment module. The operation metadata is signed based on the HMAC-SHA256 key derivation rule of the State Grid Cloud to obtain a signed log. The log hash is calculated using the SM3 algorithm and encapsulated to obtain a secure transmission message. The operation metadata includes a timestamp, a terminal system ID, and an operation type; The log hash is used for integrity comparison with the State Grid Cloud Audit System.

[0016] Preferably, the security encryption unit is also used for key management and user identity authentication, and the processing logic includes: The user's ID number is used as the user identifier of the SM9 algorithm, and the user's private key is derived from the pre-shared SM9 master key. The pre-shared SM9 master key is pre-distributed by the State Grid Cloud KMS. The user's private key is signed with SM9 using the random challenge value generated by the terminal to obtain the authentication credential, and the authentication request data is simultaneously signed with HMAC-SHA256 to generate the authentication log; Generate a 128-bit SM4 session key using a local random number generator. All 128-bit SM4 session keys are stored in OTP memory for physical isolation. A unique fingerprint of the terminal device is generated based on the SRAM PUF. The unique fingerprint of the terminal device and the real-time timestamp are used as the key derivation salt value to generate the signing key. The signing key is recalculated every hour using the KSS4 hash chain combined with the real-time timestamp, and the old key is destroyed. The timestamp is synchronized with the State Grid cloud time server through the NTP protocol to ensure that the error between the locally generated real-time timestamp and the cloud time is less than the preset time redundancy.

[0017] Beneficial effects of the present invention: The application area of ​​the multi-service access smart power terminal system proposed in this application is not limited to the State Grid business hall, and can be directly arranged in places outside the State Grid business hall, such as supermarket agent electricity purchasing points, street community service centers and village network co-construction windows. This application is aimed at application scenarios in small desktop spaces, and sets the high-definition scanner to folding mode. At the same time, taking into account that the application scenarios of this application are different from those of the State Grid business hall, and in view of the complex lighting conditions and uneven user operation levels, this application proposes a scanning and recognition method for a folding scanner to achieve adaptive and efficient recognition of paper materials. In addition, this application enhances the data encryption level at the local end of the system, which can fully deal with the risks of data leakage and attacks caused by being placed outside the business hall. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 A schematic diagram of a process flow of a multi-service access smart power terminal system provided by one embodiment of the present invention; Figure 2 A directional diagram of a multi-service access smart power terminal system provided by an embodiment of the present invention; Figure 3 A functional diagram of a multi-service access smart power terminal system provided by an embodiment of the present invention; Figure 4 A schematic diagram of the operating status of a multi-service access smart power terminal system provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0019] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the drawings. It is obvious that the described embodiments are only part of the embodiments of the present invention, but not all of the embodiments.

[0020] Reference Figure 1-4, as an embodiment of the present invention, provides a multi-service access smart power terminal system, including: Interactive perception module, business acceptance module, information service and payment module and system management module; The interactive perception module includes a user interface unit, a voice interaction unit, and a multimodal input perception unit, which are used to receive touch commands, display interface information to the user, perform voice interaction, and receive collected perception signals; The business acceptance module includes an account management unit, a basic business management unit, and a special business management unit, and is used to accept power business applications initiated by users and process power business processes; The information service and payment module includes an information inquiry service unit, an electricity fee payment processing unit, and a publicity and display unit, which are used to provide information inquiry services to users and process the payment of electricity fees and application fees; The system management module includes a system operation monitoring unit, a software operation and maintenance management unit, and a security encryption unit, which are used for terminal self-startup and self-test, software and hardware status monitoring, system log recording, software version management, and security encryption processing.

[0021] In this embodiment, the interaction perception module includes a user interface unit, a voice interaction unit, and a multimodal input perception unit, which are used to receive touch commands, display interface information to the user, voice interaction, and receive collected perception signals; The user interface unit is used to display information on the 14-inch interactive screen and receive user touch commands through the interactive screen; Interface information includes business processes, query results, promotional content, and payment interface; The voice interaction unit is used to receive and analyze the user's voice input through the microphone end of the sound hole in the AI ​​voice recognition array, and play sound prompts to the user through the speaker end of the sound hole.

[0022] In this embodiment, the multimodal input perception unit is used to receive perception signals collected by the physical input device of the terminal; Physical input devices include binocular cameras, barcode / QR code scanners, ID card readers, and foldable high-definition cameras; The user's facial data is collected through a face recognition binocular camera, the barcode information provided by the user is read through a barcode / QR code scanner, the user's ID card information is read through an ID card reader, and the paper document information provided by the user is obtained through a foldable high-definition high-definition scanner; The barcode information includes the account number QR code, online banking payment QR code, payment notice barcode, business acceptance form number barcode and contract number; Paper documents include identity documents, proof of address ownership, business authorization letter, business application form and bank account information.

[0023] In this embodiment, the foldable high-definition scanner is a scanner with an improved OCR module. The processing logic for obtaining paper document information provided by the user through the foldable high-definition scanner includes: A foldable high-definition camera captures document images. The initial image data is analyzed with the UDoc-GAN model to obtain the camera's LED fill light parameters. Light compensation is performed based on the camera's LED fill light parameters. After light compensation, the RGB camera captures the fill light image data. The processing flow includes: Initial image data is obtained by shooting with a foldable high-definition camera. The LED fill light parameters of the camera are numerically normalized and encoded to obtain a fill light parameter feature vector. Dimension expansion is performed through a fully connected layer to obtain a two-dimensional fill light constraint feature map with the same size as the initial image data. The parameters of the LED fill light of the camera include the current color temperature, brightness intensity and shooting angle of the LED fill light; The cross-modal attention layer in the UDoc-GAN model interacts with the initial image data and the two-dimensional fill-light constraint feature map to obtain a fused feature map. The first convolutional layer calculates the initial image features on the initial image data, and the fully connected layer calculates the query vector (Q), key vector (K), and value vector (V) on the two-dimensional fill-light constraint feature map. The attention weights are calculated based on the query vector, key vector, and value vector. The fused feature map is then weighted element-by-element on the initial image features based on the attention weights. The fused feature map is downsampled by the encoder: the resolution is halved each time through the convolution layer and the downsampling of step 2, the number of channels is doubled, and the LeakyReLU function is activated to obtain shallow features and deep features; The decoder performs upsampling on the fused feature map: the deep features are restored to resolution through the transposed convolution layer, the upsampled features are concatenated with the shallow features of the encoder at the same resolution through skip connections, and the illumination correction features are obtained through convolution, instance normalization, and LeakyReLU processing with residual blocks. The illumination correction features are finally upsampled, and the transposed convolution is used to restore the illumination correction features to the input image size. The number of channels is compressed to 3 through 1×1 convolution, and the feature value mapping is completed through the tanh activation function. The pixel value is linearly transformed to [0, 255] to obtain the illumination enhanced image.

[0024] The lighting enhanced image is an image with shadows and reflections removed and text edges and contrast enhanced.

[0025] In this embodiment, the illumination enhanced image is grayscaled and Gaussian filtered to obtain a grayscale denoised image, and the grayscale denoised image is subjected to line recognition through the Hough transform layer of HoughNet to obtain a line segment data set; The line segment data set includes the starting point coordinates, end point coordinates and corresponding confidence of each line segment; Filter the segments according to the confidence level of the segment data set, perform angle correction on the segments that are close to horizontal or vertical, merge the collinear segments and calculate the intersection points of all horizontal and vertical segments to obtain the initial grid points. Identify the smallest rectangular area based on the initial grid points to generate a cell, and extract the bounding box coordinates of the cell based on the line segment data set; Remove duplicate cells and sort them in the logical order of text to obtain a cell coordinate grid. The logical order of text is from top to bottom and from left to right. Based on high-quality images and cell coordinate grids, the OCR engine is used to identify the OCR text content and coordinates of each cell. The GNN graph neural network is used to map the OCR text content, logical row, column index, merge cell flag, and field name corresponding to each cell to obtain structured text table data in JSON format.

[0026] In this embodiment, the service acceptance module includes an account management unit, a basic service management unit, and a special service management unit, which are used to accept power service applications initiated by users and process power service processes; The account management unit is responsible for user registration, login, and real-name authentication on the State Grid Cloud, and provides users with household head identity verification, account transfer operations, account cancellation operations, identity information changes, account division / merger, and associated account number binding / unbinding services; The basic business management unit is responsible for handling basic electricity services initiated by users, including new installations, capacity expansion / reduction, temporary electricity application, suspension / resumption of electricity use, signing of smart meter fee control agreements, electricity price plan changes, electronic invoice subscriptions, and activation / cancellation of SMS reminder services. The special business management unit is responsible for handling special electricity business initiated by users and guiding users through the approval or survey process. Special electricity business includes distributed photovoltaic grid connection, charging pile registration, electric heating dedicated line registration, historical electricity fee dispute review, fault repair upgrade processing and electricity support qualification approval.

[0027] In this embodiment, the information service and payment module includes an information query service unit, an electricity fee payment processing unit, and a publicity and display unit, which are used to provide information query services to users and process the payment of electricity fees and application fees; The information query service unit is used to query and display the user's service records, electricity bills, payment records, energy efficiency bill analysis, power grid maintenance and power start and stop records according to the user's request; The electricity bill payment processing unit is used to complete the user's electricity bill payment and service fee payment operations by identifying the payment code through a barcode / QR code scanner, recording the payment results and generating receipts; The publicity and display unit is responsible for managing and displaying State Grid's publicity information, including State Grid news, promotional activities, business change notifications and electricity policy introductions.

[0028] In this embodiment, the system management module includes a system operation monitoring unit, a software operation and maintenance management unit, and a security encryption unit, which are responsible for terminal self-startup and self-test, software and hardware status monitoring, system log recording, software version management, and security encryption processing; The system operation monitoring unit is responsible for the terminal's startup and self-test, continuously monitors the health status of hardware and software during system operation, and records system operation logs, user operation logs, and exception logs; The software operation and maintenance management unit is responsible for monitoring the version of terminal application software and controlling the update and rollback operations of terminal application software.

[0029] In this embodiment, the security encryption unit is used to encrypt plaintext service data to obtain a secure transmission message. The processing logic includes: By performing SM4-CTR encryption on the plaintext service data, a key stream is obtained based on the randomly generated 128-bit SM4 session key, initialization vector IV, and counter Counter. The plaintext service data is XORed with the key stream to obtain the encrypted service data. Plain text business data includes user operation data output by the interactive perception module, business acceptance module, and information service and payment module. The operation metadata is signed based on the HMAC-SHA256 key derivation rule of the State Grid Cloud to obtain a signed log. The log hash is calculated using the SM3 algorithm and encapsulated to obtain a secure transmission message. Operation metadata includes timestamp, end system ID, and operation type; Log hash is used for integrity comparison with the State Grid Cloud Audit System.

[0030] In this embodiment, the security encryption unit is also used for key management and user identity authentication. The processing logic includes: The user's ID number is used as the user identifier of the SM9 algorithm, and the user's private key is derived from the pre-shared SM9 master key. The pre-shared SM9 master key is pre-distributed by the State Grid Cloud KMS. The user's private key is signed with SM9 using the random challenge value generated by the terminal to obtain the authentication credential, and the authentication request data is simultaneously signed with HMAC-SHA256 to generate the authentication log; Generate a 128-bit SM4 session key using a local random number generator. All 128-bit SM4 session keys are stored in OTP memory for physical isolation. A unique fingerprint of the terminal device is generated based on the SRAM PUF. The unique fingerprint of the terminal device and the real-time timestamp are used as the key derivation salt value to generate the signing key. The signing key is recalculated every hour using the KSS4 hash chain combined with the real-time timestamp, and the old key is destroyed. The timestamp is synchronized with the State Grid cloud time server through the NTP protocol to ensure that the error between the locally generated real-time timestamp and the cloud time is less than the preset time redundancy.

[0031] Those skilled in the art will appreciate that embodiments of the present invention may provide methods, systems, or computer program products. Therefore, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media containing computer-usable program code. The storage medium may be implemented by any type of volatile or non-volatile storage device, or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk. These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0032] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A multi-service access smart power terminal system, characterized in that: include: Interactive perception module, business acceptance module, information service and payment module and system management module; The interactive perception module includes a user interface unit, a voice interaction unit, and a multimodal input perception unit, which are used to receive touch commands, display interface information to the user, perform voice interaction, and receive collected perception signals; The business acceptance module includes an account management unit, a basic business management unit, and a special business management unit, and is used to accept power business applications initiated by users and process power business processes; The information service and payment module includes an information inquiry service unit, an electricity fee payment processing unit, and a publicity and display unit, which are used to provide information inquiry services to users and process the payment of electricity fees and application fees; The system management module includes a system operation monitoring unit, a software operation and maintenance management unit, and a security encryption unit, which are used for terminal self-startup and self-test, software and hardware status monitoring, system log recording, software version management, and security encryption processing.

2. The multi-service access smart power terminal system according to claim 1, characterized in that: The interactive perception module includes a user interface unit, a voice interaction unit, and a multimodal input perception unit, which are used to receive touch commands, display interface information to the user, perform voice interaction, and receive collected perception signals; The user interface unit is used to display information on the 14-inch interactive screen and receive user touch commands through the interactive screen; The interface information includes business processes, query results, promotional content and payment interface; The voice interaction unit is used to receive and analyze the user's voice input through the microphone end of the sound hole in the AI ​​voice recognition array, and play sound prompts to the user through the speaker end of the sound hole.

3. The multi-service access smart power terminal system according to claim 1, characterized in that: The multimodal input perception unit is used to receive perception signals collected by the terminal's physical input device; The physical input devices include a binocular camera, a barcode / QR code scanner, an ID card reader, and a foldable high-definition camera; The user's facial data is collected through a face recognition binocular camera, the barcode information provided by the user is read through a barcode / QR code scanner, the user's ID card information is read through an ID card reader, and the paper document information provided by the user is obtained through a foldable high-definition high-definition scanner; The barcode information includes the account number QR code, online banking payment QR code, payment notice barcode, business acceptance form number barcode and contract number; The paper document information includes identity documents, address ownership certificate, business authorization letter, business application form and bank account information.

4. The multi-service access smart power terminal system according to claim 3, characterized in that: The foldable high-definition scanner is a scanner with an improved OCR module. The processing logic for obtaining paper document information provided by the user through the foldable high-definition scanner includes: A foldable high-definition camera captures document images. The initial image data is analyzed with the UDoc-GAN model to obtain the camera's LED fill light parameters. Light compensation is performed based on the camera's LED fill light parameters. After light compensation, the RGB camera captures the fill light image data. The processing flow includes: Initial image data is obtained by shooting with a foldable high-definition camera. The LED fill light parameters of the camera are numerically normalized and encoded to obtain a fill light parameter feature vector. Dimension expansion is performed through a fully connected layer to obtain a two-dimensional fill light constraint feature map with the same size as the initial image data. The LED fill light parameters of the camera include the current color temperature, brightness intensity and shooting angle of the LED fill light; The cross-modal attention layer in the UDoc-GAN model interacts with the initial image data and the two-dimensional fill-light constraint feature map to obtain a fused feature map. The first convolutional layer calculates the initial image features on the initial image data, and the fully connected layer calculates the query vector (Q), key vector (K), and value vector (V) on the two-dimensional fill-light constraint feature map. The attention weights are calculated based on the query vector, key vector, and value vector. The fused feature map is then weighted element-by-element on the initial image features based on the attention weights. The fused feature map is downsampled by the encoder: the resolution is halved each time through the convolution layer and the downsampling of step 2, the number of channels is doubled, and the LeakyReLU function is activated to obtain shallow features and deep features; The decoder performs upsampling on the fused feature map: the deep features are restored to resolution through the transposed convolution layer, the upsampled features are concatenated with the shallow features of the encoder at the same resolution through skip connections, and the illumination correction features are obtained through convolution, instance normalization, and LeakyReLU processing with residual blocks. The illumination correction features are finally upsampled, and the transposed convolution is used to restore the illumination correction features to the input image size. The number of channels is compressed to 3 through 1×1 convolution, and the feature value mapping is completed through the tanh activation function. The pixel value is linearly transformed to [0, 255] to obtain the illumination enhanced image.

5. The multi-service access smart power terminal system according to claim 4, characterized in that: The illumination enhanced image is grayscaled and Gaussian filtered to obtain a grayscale denoised image. The grayscale denoised image is then subjected to line recognition through the Hough transform layer of HoughNet to obtain a line segment data set. The line segment data set includes the starting point coordinates, end point coordinates and corresponding confidence of each line segment; Filter the segments according to the confidence level of the segment data set, perform angle correction on the segments that are close to horizontal or vertical, merge the collinear segments and calculate the intersection points of all horizontal and vertical segments to obtain the initial grid points. Identifying a minimum rectangular area based on the initial grid points to generate a cell, and extracting the bounding box coordinates of the cell based on the line segment data set; Remove duplicate cells and sort them in logical order of text to obtain a cell coordinate grid, where the text logical order is from top to bottom and from left to right; Based on high-quality images and cell coordinate grids, the OCR engine is used to identify the OCR text content and coordinates of each cell. The GNN graph neural network is used to map the OCR text content, logical row, column index, merge cell flag, and field name corresponding to each cell to obtain structured text table data in JSON format.

6. The multi-service access smart power terminal system according to claim 1, characterized in that: The business acceptance module includes an account management unit, a basic business management unit, and a special business management unit, and is used to accept power business applications initiated by users and process power business processes; The account management unit is responsible for user registration, login, and real-name authentication on the State Grid Cloud, and provides users with household head identity verification, account transfer operations, account cancellation operations, identity information changes, account division / merger, and associated account number binding / unbinding services; The basic service management unit is responsible for handling basic electricity services initiated by users, including new installations, capacity expansion / reduction, temporary electricity application, suspension / resumption of electricity use, smart meter fee control agreement signing, electricity price plan changes, electronic invoice subscriptions, and SMS reminder service activation / cancellation; The special business management unit is responsible for handling special electricity business initiated by users and guiding users through the approval or survey process. The special electricity business includes distributed photovoltaic grid connection, charging pile registration, electric heating dedicated line registration, historical electricity fee dispute review, fault repair upgrade processing and electricity support qualification approval.

7. The multi-service access smart power terminal system according to claim 1, characterized in that: The information service and payment module includes an information inquiry service unit, an electricity fee payment processing unit, and a publicity and display unit, which are used to provide information inquiry services to users and process the payment of electricity fees and application fees; The information query service unit is used to query and display the user's service records, electricity bills, payment records, energy efficiency bill analysis, power grid maintenance and power start and stop records according to the user's request; The electricity bill payment processing unit is used to complete the user's electricity bill payment and service fee payment operations by identifying the payment code through a barcode / QR code scanner, recording the payment results and generating receipts; The publicity and display unit is responsible for managing and displaying State Grid's publicity information, including State Grid Corporation news, promotional activities, business change notifications and electricity policy introductions.

8. The multi-service access smart power terminal system according to claim 1, characterized in that: The system management module includes a system operation monitoring unit, a software operation and maintenance management unit, and a security encryption unit, which are responsible for terminal self-startup and self-test, software and hardware status monitoring, system log recording, software version management, and security encryption processing; The system operation monitoring unit is responsible for the terminal's startup and self-test, continuously monitors the health status of hardware and software during system operation, and records system operation logs, user operation logs, and exception logs; The software operation and maintenance management unit is responsible for monitoring the version of terminal application software and controlling the update and rollback operations of terminal application software.

9. The multi-service access smart power terminal system according to claim 8, characterized in that: The security encryption unit is used to encrypt plaintext business data to obtain secure transmission messages. The processing logic includes: By performing SM4-CTR encryption on the plaintext service data, a key stream is obtained based on the randomly generated 128-bit SM4 session key, initialization vector IV, and counter Counter. The plaintext service data is XORed with the key stream to obtain the encrypted service data. The plain text business data includes user operation data output by the interactive perception module, the business acceptance module and the information service and payment module. The operation metadata is signed based on the HMAC-SHA256 key derivation rule of the State Grid Cloud to obtain a signed log. The log hash is calculated using the SM3 algorithm and encapsulated to obtain a secure transmission message. The operation metadata includes a timestamp, a terminal system ID, and an operation type; The log hash is used for integrity comparison with the State Grid Cloud Audit System.

10. The multi-service access smart power terminal system according to claim 9, characterized in that: The security encryption unit is also used for key management and user authentication. The processing logic includes: The user's ID number is used as the user identifier of the SM9 algorithm, and the user's private key is derived from the pre-shared SM9 master key. The pre-shared SM9 master key is pre-distributed by the State Grid Cloud KMS. The user's private key is signed with SM9 using the random challenge value generated by the terminal to obtain the authentication credential, and the authentication request data is simultaneously signed with HMAC-SHA256 to generate the authentication log; Generate a 128-bit SM4 session key using a local random number generator. All 128-bit SM4 session keys are stored in OTP memory for physical isolation. A unique fingerprint of the terminal device is generated based on the SRAM PUF. The unique fingerprint of the terminal device and the real-time timestamp are used as the key derivation salt value to generate the signing key. The signing key is recalculated every hour using the KSS4 hash chain combined with the real-time timestamp, and the old key is destroyed. The timestamp is synchronized with the State Grid cloud time server through the NTP protocol to ensure that the error between the locally generated real-time timestamp and the cloud time is less than the preset time redundancy.

Images