Election machine of one-time multi-ticket identification system based on election

Abstract

The invention discloses an election machine of a one-time multi-ticket recognition system based on election, and belongs to the technical field of intelligent election equipment. According to the system, by integrating high-precision image acquisition, multiple identity verification and intelligent data processing technologies, an efficient and safe election process is realized. The system mainly comprises five modules: an image acquisition module in which a 20 million pixel CCD camera is matched with a multispectral light source and supports one-time processing of five votes; an identity verification module based on biological characteristics and certificate identification; the intelligent processing module adopts advanced image segmentation and OCR technology; a data security module applying AES-256 encryption and a block chain; and a cloud collaboration module. The system ensures data security through an encryption algorithm and a block chain; and real-time result query is supported, and election transparency is improved. Practical tests verify that the system can effectively solve the problems of low efficiency, poor safety and the like of traditional election, and has significant application value.

Description

Technical Field

[0001] The present invention relates to the technical field of election equipment, and in particular to an election machine based on a one-time multiple-vote recognition system. Background Art

[0002] Traditional election systems suffer from significant deficiencies in efficiency, security, resource utilization, and transparency. First, existing voting machines typically process ballots one by one, resulting in slow voting. This can easily lead to queues and congestion, especially in large-scale elections, impacting the voter experience. Manual vote counting is not only time-consuming and labor-intensive but also prone to human error, making it difficult to meet the efficiency and accuracy demands of modern elections. Second, traditional ballots lack effective anti-counterfeiting mechanisms, making them susceptible to forgery or tampering. Furthermore, the limited means of voter authentication makes it difficult to ensure the fair principle of "one person, one vote." Furthermore, the storage and transmission of ballots lack strong encryption protection, posing the risk of data leakage or tampering, undermining the credibility of the elections.

[0003] In terms of resource utilization, traditional systems have strict requirements for ballot material, size, and printing, resulting in limited flexibility. Their single-ballot processing method also results in low material utilization, increasing election costs. Transparency is also a prominent issue. The traditional vote counting process lacks a real-time public disclosure mechanism, preventing effective public oversight and easily raising questions about the election results. Existing optical recognition technology also struggles to cope with complex situations such as overlapping, stained, or tilted ballots, resulting in limited recognition accuracy and an inability to adapt to the demands of highly concurrent elections.

[0004] In response to the above problems, an election machine based on a one-time multiple-vote recognition system is urgently needed to solve these problems. Summary of the Invention

[0005] The purpose of the present invention is to solve the existing technical problems raised in the above background technology, and to provide an election machine based on a one-time multiple-vote recognition system.

[0006] The present invention achieves the above-mentioned object through the following scheme:

[0007] The election mechanism of the one-time multiple-vote recognition system based on the election includes:

[0008] Image acquisition module, used to capture high-resolution images of multiple ballots at once;

[0009] An authentication module, used to verify the unique identity of the voter;

[0010] The ballot uniqueness verification module is used to verify the uniqueness of each ballot through a unique identifier;

[0011] The data processing module is used to segment, identify and count the collected ballot images and generate voting results;

[0012] Storage and communication module, used to store ballot data and transmit it to the cloud through an encrypted channel;

[0013] Among them, the image acquisition module transmits image data to the data processing module through a high-speed interface, and the data format is RGB or grayscale image; the data processing module realizes multiple ballot recognition at one time through image segmentation algorithm, and the recognition efficiency meets the requirements of processing at least 3 ballots per second; when the ballot is placed incorrectly, the system guides the user to re-operate through a prompt interface.

[0014] As a preferred technical solution of the present invention, the image acquisition module includes a high-resolution CCD camera and a multi-spectral light source. The resolution of the CCD camera is not less than 20 million pixels and supports the acquisition of images containing at least 5 ballots.

[0015] Among them, the multi-spectral light source provides 400-700nm visible light and 850nm infrared light, automatically adjusting the brightness to enhance the contrast of the QR code or watermark on the ballot; the CCD camera supports the acquisition of ballots with a size range of A6 to A4, and the materials include paper and translucent media.

[0016] As a preferred technical solution of the present invention, the identity verification module adopts at least one of the following verification methods: fingerprint recognition, ID card scanning, and face recognition, with the verification priority being fingerprint, ID card, and face; the module generates a unique identity (ID) for the voter, which is defined as:

[0017] ID=H(V d )

[0018] Where: V d represents the voter's biometric data (such as fingerprint feature points) or ID data; H(·) represents the SHA-256 hash function, which generates a 256-bit unique identifier; when verification fails, the system automatically switches to the backup verification method. If the number of failures exceeds 3, the log is recorded and the operation is suspended; biometric data ν d It is deleted immediately after being generated, and only the hash value (ID) is stored, which complies with privacy protection regulations; the (ID) is bound to the unique identifier of the ballot and stored in the storage module.

[0019] As a preferred technical solution of the present invention, the ballot uniqueness verification module generates a unique identifier T for each ballot through QR code or watermark technology. k , defined as:

[0020] T k =H(ID||N s ||T s )

[0021] Where: (ID) is the voter’s unique identity; N s T is the ballot serial number, indicating the ballot issuance number; s Generate a timestamp for the ballot, accurate to milliseconds; (||) represents the data concatenation operation; H(·) is the SHA-256 hash function;

[0022] The module compares T k Detect duplicate ballots with historical records in the storage module; extract T through DWT-based watermark when the QR code is damaged k If it still fails, mark the ballot as invalid and record it in the log; system maintenance T k Index table, by timestamp T s Distinguish hash collisions.

[0023] As a preferred technical solution of the present invention, the data processing module includes an image segmentation algorithm for segmenting an image containing multiple ballots into a single ballot sub-image. The specific steps are:

[0024] Preprocess the input image (I(x,y)) and apply Gaussian filtering to remove noise:

[0025] I f (x,y)=I(x,y)*G(σ),σ=1.5

[0026] Where: (1(x,y)) is the grayscale value of the original image; G(σ) is the Gaussian kernel with standard deviation σ; * represents the convolution operation;

[0027] Apply the Canny edge detection algorithm to extract the ballot boundary and generate a binary edge map (E(x,y));

[0028] Through the K-means clustering algorithm, the edge map is divided into (N) sub-images S i (i=1,2,...,N), the initial center is the geometric center of the edge graph, the number of iterations does not exceed 10, and the convergence condition is that the variance change within the cluster is less than 0.01; when the ballots overlap or tilt, the angle is corrected by Hough transform; the segmented sub-images meet the following requirements:

[0029]

[0030] Where Area(·) represents the area of the image. If the segmentation fails, the system prompts the user to re-place the ballot.

[0031] As a preferred technical solution of the present invention, the data processing module further includes a ballot content recognition algorithm for identifying the content of the ballot from the sub-image S i Extract voting information, specifically:

[0032] Pair image Si Decode the QR code and extract the unique identification T of the ballot k , the QR code adopts the QR code standard and supports an error correction rate of up to 30%;

[0033] Apply optical character recognition (OCR) to voting areas using a convolutional neural network (CNN) model. The training dataset contains 100,000 ballot samples and supports both handwritten and printed fonts.

[0034] The recognition result is represented as vector V i ={T k ,C j}, where: T k It is the unique identifier of the ballot; C j is the candidate or option number, j = 1, 2, ..., M, where M is the total number of candidates. The system uses a multi-spectral light source and an adaptive threshold algorithm to handle uneven lighting or soiled ballots, and the recognition accuracy meets the following requirements:

[0035]

[0036] Among them, N correct is the number of correctly identified ballots, N t otal is the total number of votes.

[0037] As a preferred technical solution of the present invention, the data processing module performs voting statistics based on the recognition result ν i Generate a statistics vector (R), defined as:

[0038]

[0039] Where: r j Indicates candidate or option c j The number of votes; δ(C j ,V i ) is the indicator function, if V i Contains C j , then δ=1, otherwise δ=0; (N) is the total number of valid votes, (M) is the total number of candidates;

[0040] When the recognition result v i If a ballot contains invalid options (such as multiple choices or blanks), the system will mark it as an invalid ballot, exclude it from the count (R), and record it in the log. After the count is completed, a cross-check is performed to ensure that:

[0041]

[0042] The statistical vector (R) is updated in real time and stored in the storage module.

[0043] As a preferred technical solution of the present invention, the storage and communication module uses the AES-256 encryption algorithm to encrypt the ballot data and statistical results. The encryption process is expressed as follows:

[0044] D e =AES 256 (D,K e )

[0045] Where: (D) is the original data, including the ballot sub-image S i , unique identifier T k and statistical vector (R); K e It is a 256-bit encryption key, randomly generated by a hardware security module (HSM) and updated every 24 hours; e This is encrypted data; the encrypted data is transmitted to the cloud server via the HTTPS protocol, with a transmission delay of less than 1 second; if the transmission is interrupted, the system caches the data and retransmits it after the connection is restored to ensure data integrity.

[0046] The present invention also provides an election method based on one-time multiple vote recognition, comprising the following steps:

[0047] Verify the voter's identity through the identity verification module and generate a unique identity (ID);

[0048] Capture multiple ballot images at one time through the image acquisition module;

[0049] The image is segmented by the data processing module to generate the ballot sub-image S i ;

[0050] The data processing module identifies the sub-image content and generates voting information v i ;

[0051] Verify the unique identification of the ballot T through the ballot uniqueness verification module k , to prevent duplicate voting;

[0052] The voting results are counted through the data processing module to generate a statistical vector (R);

[0053] Data is encrypted and transmitted to the cloud through the storage and communication module.

[0054] As a preferred technical solution of the present invention, the method further includes a real-time disclosure step:

[0055] The statistical vector (R) is stored in the consortium chain. Only authorized nodes can write data. The storage process satisfies:

[0056] B t =HB t-1 ||R||T s )

[0057] Among them: B t is the hash of the current block; B t-1 is the hash of the previous block; (R) is the statistical vector; T s is the timestamp in claim 4; H(·) is the SHA-256 hash function;

[0058] A public query interface is provided. Users need to enter a verification code or public key signature. The query response time is less than 0.5 seconds. It is limited to the statistical vector (R) and ballot status, and does not include the voter identity (ID).

[0059] Compared with the prior art, the present invention has the following beneficial effects:

[0060] This invention utilizes high-resolution image acquisition and intelligent image segmentation algorithms to process multiple ballots simultaneously, significantly improving voting and counting speeds. Combined with efficient optical character recognition (OCR) technology, the system can quickly and accurately tally voting results, achieving an accuracy rate exceeding 99.5%. Furthermore, automated processing reduces manual intervention, effectively avoiding errors that can occur with traditional manual counting and ensuring the reliability of election results.

[0061] The system utilizes multiple authentication mechanisms (such as fingerprint, ID card, and facial recognition) and ballot uniqueness verification technology to ensure the fairness of the "one person, one vote" system. Using encryption algorithms (such as AES-256) and blockchain technology, all ballot data is strictly protected during storage and transmission to prevent tampering or leakage. Furthermore, the QR code and watermark design on the ballot further enhances anti-counterfeiting capabilities, effectively eliminating duplicate voting or ballot forgery.

[0062] This invention supports ballots of various materials and sizes, improving resource utilization and reducing election costs. Through consortium blockchain technology, statistical results are publicly available in real time, ensuring transparency and traceability of the election process. The system also features exception handling capabilities, adapting to emergencies such as network outages and hardware failures, ensuring election stability and integrity. These features collectively promote the modernization and efficiency of election systems. BRIEF DESCRIPTION OF THE DRAWINGS

[0063] In order to more clearly illustrate the technical solutions in the present invention or the prior art, a brief introduction will be given below to the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0064] Figure 1 This is a system block diagram of an election machine based on a one-time multiple-vote recognition system of the present invention;

[0065] Figure 2 This is a flowchart of the operation of the voting machine based on the one-time multiple vote recognition system of the present invention;

[0066] Figure 3 It is a schematic diagram of the appearance of a voting machine based on a one-time multiple-vote recognition system of the present invention;

[0067] Figure 4 It is an internal schematic diagram of an election machine based on the one-time multiple vote recognition system of the present invention. DETAILED DESCRIPTION

[0068] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0069] See Figure 1 、 Figure 2 and Figure 3 This invention provides an election machine and method based on a one-time, multiple-vote recognition system, aiming to achieve an efficient, secure, and transparent election process. The following describes the implementation of this system in detail through specific examples, covering the hardware structure, algorithm flow, data processing, security mechanisms, and public procedures.

[0070] The voting machine consists of an image acquisition module, an identity authentication module, a ballot uniqueness verification module, a data processing module, and a storage and communication module, which work together to complete the collection, verification, identification, statistics and data transmission of ballots.

[0071] When implemented, the election machine Figure 3 As described, the voting machine is deployed at the polling station and is equipped with a high-resolution CCD camera (resolution of 20 million pixels), a multi-spectral light source (covering 400-700nm visible light and 850nm infrared light), a fingerprint reader, an ID card scanner, a face recognition camera, a touch screen display and a built-in processor (supporting real-time image processing and encryption operations). The system connects each module through a high-speed interface.

[0072] The voting process starts with identity verification. Voters first select the verification method on the touch screen, with fingerprint recognition being the first choice, followed by ID card scanning or face recognition. If the fingerprint feature points captured by the fingerprint reader are clear, the system extracts the feature data V d , generate a 256-bit unique identity ID = H(V d ), to protect privacy, the original feature data Vd It is deleted immediately after being generated (ID), and only the hash value (ID) is stored. If the fingerprint is blurred or the recognition fails, the system automatically switches to ID card scanning and reads the ID number as V d , regenerate (ID). If the verification fails three times in a row, the system will record the log and suspend the operation, prompting the user to contact the staff through the touch screen. Face recognition is used as the last backup method, using the camera to capture facial features and also generate (ID). After verification, (ID) is stored in the local encrypted database for subsequent ballot binding.

[0073] After completing identity verification, the voter places multiple ballots (up to 5, ranging in size from A6 to A4, and made of paper or translucent plastic) into the scanning area of the image acquisition module. The CCD camera captures a high-definition RGB image of all ballots within 1 second. The multi-spectral light source automatically adjusts the brightness, preferentially using infrared light (850nm) to enhance the contrast of the QR code or watermark, ensuring clear imaging under different lighting conditions. If the ballot material is reflective, the system uses visible light fill to reduce reflective interference. The captured image I(x,y) is transmitted to the data processing module via a high-speed interface in the format of 24-bit RGB or 8-bit grayscale image.

[0074] The data processing module first segments the image to separate multiple ballots. The system applies Gaussian filtering to the input image I(x,y), and the formula is I f (x,y)=I(x,y)*G(σ), where the Gaussian kernel standard deviation σ=1.5, which effectively removes noise.

[0075] Subsequently, the Canny edge detection algorithm was used to extract the ballot's boundaries and generate a binary edge map (E(x,y)). The parameters were set to a low threshold of 50 and a high threshold of 150 to ensure edge continuity. To address overlapping or tilted ballots, the system used a Hough transform to detect straight lines and correct the image angle.

[0076] Then, the edge image is divided into (N) sub-images S based on the K-means clustering algorithm. i (i=1,2,...,N), the initial cluster center is the geometric center of the edge graph, the upper limit of the number of iterations is 10 times, the convergence condition is that the variance change within the cluster is less than 0.01, and the segmentation result satisfies Where Area(·) is the pixel area. If the number of sub-images after segmentation does not match the expected number (for example, 6 areas are detected and the user puts in 5 ballots), the system prompts the user to re-place the ballots through the touch screen.

[0077] Each sub-image s i Entering the content recognition stage, the system first decodes the QR code on the ballot and extracts the unique identifier T kThe QR code adopts the QR code standard and supports a 3096 error correction rate, so it can be accurately decoded even if the ballot is partially damaged. k By formula T k =H(ID||N s ||T s ) is generated, where (ID) is the voter identity, N s is the ballot serial number (pre-assigned by the election agency, 32-bit integer), T s To generate a timestamp (accurate to milliseconds), SHA-256 hashing is used to ensure uniqueness. To prevent forgery, the ballot paper is embedded with an invisible watermark based on discrete wavelet transform (DWT) as a backup verification method when the QR code is damaged. If neither the QR code nor the watermark can be extracted, the T k , the system marks the ballot as invalid and records the log, and the ballot uniqueness verification module extracts T k Compare with the local database, if duplicates are found (same T k If it occurs multiple times), the system will trigger an alarm and suspend processing, record the abnormal log, and maintain T in order to avoid hash collision. k Index table, combined with timestamp T s Identify potential conflicts.

[0078] Voting content recognition is accomplished through optical character recognition (OCR). The system applies a convolutional neural network (CNN)-based OCR model to the voting area of sub-image S; the training data set contains 100,000 ballot samples, covering both handwritten and printed fonts. OCR recognizes candidate numbers or options (such as "Candidate A" or "Option 1") and outputs C j (j=1,2,...,M), where (M) is the total number of candidates. The recognition results form a vector V i ={T k ,C j To cope with uneven lighting or contamination, the system combines multi-spectral light source and adaptive threshold algorithm to optimize image quality, with recognition accuracy reaching If the recognition result is invalid (such as multiple selections or blanks), the system marks the ballot as invalid and does not count it. Voting statistics are based on all valid votes. i , generate statistical vector R = {r j |r j =∑ i =1 N δ(C j ,V i ),j=1,2,...,M}, where r j For candidate c j The number of votes, δ(C j ,V i ) is the indicator function (if V i Contains C j, then δ=1, otherwise δ=0), after the statistics are completed, the system performs cross-check to ensure Σ j =1 M r j ≤N, where (N) is the total number of valid votes. The statistical vector (R) is displayed on the touch screen in real time and stored in a local encrypted database. Invalid vote information (such as T k and invalid reasons) are recorded in the log for subsequent audit.

[0079] All data, including sub-image S i , unique identifier T k 、Voting Information i And statistical vector (R), processed by the storage and communication module, the system uses AES-256 encryption algorithm, the formula is D e =AES 256 (D,K e ), where (D) is the original data, K e It is a 256-bit key, randomly generated by a hardware security module (HSM) and updated every 24 hours.

[0080] Encrypted data D e The data is transmitted to the cloud server via the HTTPS protocol with a transmission delay of less than 1 second. If the network is interrupted, the system will cache the data locally (with a capacity of 100,000 votes) and retransmit it after the connection is restored. The key distribution uses elliptic curve cryptography (ECC) to ensure security. To improve transparency, the system stores the statistical vector (R) in a consortium chain (such as Hyperledger Fabric). Only authorized nodes (such as election agencies) can write data. The blockchain storage formula is B t =HB t-1 ||R||T s ), where B t-1 is the hash of the previous block, T s The timestamp and SHA-256 hash ensure that the data cannot be tampered with. The public accesses the statistical results through the query interface and needs to enter a verification code or public key signature. The query response time is less than 0.5 seconds. It is limited to (R) and ballot status (valid / invalid), and does not include the voter identity (ID). The query interface is deployed in the cloud and supports multi-language display (such as Chinese and English), which is convenient for different groups to use.

[0081] In actual testing, the voting machines operated in a local election involving 1,000 voters. Each device processed approximately 1,200 ballots per hour, with an average authentication time of 1 second, image acquisition and segmentation of 2 seconds, content recognition and counting of 1 second, and data transmission of 0.5 seconds, for a total of 4.5 seconds, meeting efficiency requirements. The system successfully detected and marked 5 duplicate ballots and 10 invalid ballots, achieving an accuracy rate of 99.8%. The blockchain query interface was open within 1 minute after the election, and public query responses were stable, with no record of data leaks.

[0082] To ensure the robustness of the system, various abnormal scenarios are considered during implementation. If the image acquisition module detects that the number of ballots exceeds the upper limit of 5, the system prompts the user to reduce the number of ballots. If the data processing module is suspended due to a hardware failure (such as memory overflow), the system switches to the backup processor and continues to store data locally. If the cloud server is unavailable, the system remains in offline mode, and all data is encrypted and stored locally, and uploaded in batches after the network is restored. Privacy protection runs through the entire process. Voter identity data is only stored in hashed form, and the ballot content is separated from the (ID), complying with regulations such as GDPR.

[0083] like Figure 4 As shown, an election machine with a built-in multi-vote recognition system is shown, wherein a paper path 2 transmits ballots, a ballot information collection component 3 obtains ballot information, and the collected ballot information is processed by a control center 1 with a display screen to obtain election results; at the same time, the election results are printed by a printer 4.

[0084] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present invention.

Claims

1. An election machine based on a one-time multiple vote recognition system, characterized in that: include: Image acquisition module, used to capture high-resolution images of multiple ballots at once; An authentication module, used to verify the unique identity of the voter; The ballot uniqueness verification module is used to verify the uniqueness of each ballot through a unique identifier; The data processing module is used to segment, identify and count the collected ballot images and generate voting results; Storage and communication module, used to store ballot data and transmit it to the cloud through an encrypted channel; Among them, the image acquisition module transmits image data to the data processing module through a high-speed interface, and the data format is RGB or grayscale image; the data processing module realizes multiple ballot recognition at one time through image segmentation algorithm, and the recognition efficiency meets the requirements of processing at least 3 ballots per second; when the ballot is placed incorrectly, the system guides the user to re-operate through a prompt interface.

2. The voting machine based on the one-time multiple vote recognition system of claim 1, characterized in that: The image acquisition module includes a high-resolution CCD camera and a multi-spectral light source. The resolution of the CCD camera is not less than 20 million pixels and supports the acquisition of images containing at least 5 ballots. Among them, the multi-spectral light source provides 400-700nm visible light and 850nm infrared light, automatically adjusting the brightness to enhance the contrast of the QR code or watermark on the ballot; the CCD camera supports the acquisition of ballots with a size range of A6 to A4, and the materials include paper and translucent media.

3. The voting machine based on the one-time multiple vote recognition system of claim 1, characterized in that: The identity verification module uses at least one of the following verification methods: fingerprint recognition, ID card scanning, and face recognition, with the verification priority being fingerprint, ID card, and face. The module generates a unique identity (ID) for the voter, which is defined as: ID=H(V d ) Where: V d represents the voter's biometric data (such as fingerprint feature points) or ID data; H(·) represents the SHA-256 hash function, which generates a 256-bit unique identifier; when verification fails, the system automatically switches to the backup verification method. If the number of failures exceeds 3, the log is recorded and the operation is suspended; biometric data ν d It is deleted immediately after being generated, and only the hash value (ID) is stored, which complies with privacy protection regulations; the (ID) is bound to the unique identifier of the ballot and stored in the storage module.

4. The voting machine based on the one-time multiple vote recognition system of claim 1, characterized in that: The ballot uniqueness verification module generates a unique identifier T for each ballot through QR code or watermark technology. k , defined as: T k =H(ID||N s ||T s ) Where: (ID) is the voter’s unique identity; N s T is the ballot serial number, indicating the ballot issuance number; s Generate a timestamp for the ballot, accurate to milliseconds; (||) represents the data concatenation operation; H(·) is the SHA-256 hash function; The module compares T k Detect duplicate ballots with historical records in the storage module; extract T through DWT-based watermark when the QR code is damaged k If it still fails, mark the ballot as invalid and record it in the log; system maintenance T k Index table, by timestamp T s Distinguish hash collisions.

5. The voting machine based on the one-time multiple vote recognition system of claim 1, characterized in that: The data processing module includes an image segmentation algorithm for segmenting an image containing multiple ballots into individual ballot sub-images. The specific steps are as follows: Preprocess the input image (I(x,y)) and apply Gaussian filtering to remove noise: I f (x,y)=I(x,y)*G(σ),σ=1.5 Where: (1(x,y)) is the grayscale value of the original image; G(σ) is the Gaussian kernel with standard deviation σ; * represents the convolution operation; Apply the Canny edge detection algorithm to extract the ballot boundary and generate a binary edge map (E(x,y)); Through the K-means clustering algorithm, the edge map is divided into (N) sub-images S i (i=1,2,...,N), the initial center is the geometric center of the edge graph, the number of iterations does not exceed 10, and the convergence condition is that the variance change within the cluster is less than 0.01; when the ballots overlap or tilt, the angle is corrected by Hough transform; the segmented sub-images meet the following requirements: Where Area(·) represents the area of the image. If the segmentation fails, the system prompts the user to re-place the ballot.

6. The voting machine based on the one-time multiple vote recognition system of claim 5, characterized in that: The data processing module further includes a ballot content recognition algorithm for identifying the content of the ballot from the sub-image S i Extract voting information, specifically: Pair image S i Decode the QR code and extract the unique identification T of the ballot k , the QR code adopts the QR code standard and supports an error correction rate of up to 30%; Apply optical character recognition (OCR) to voting areas using a convolutional neural network (CNN) model. The training dataset contains 100,000 ballot samples and supports both handwritten and printed fonts. The recognition result is represented as vector V i ={T k ,C j }, where: T k It is the unique identifier of the ballot; C j is the candidate or option number, j = 1, 2, ..., M, where M is the total number of candidates. The system uses a multi-spectral light source and an adaptive threshold algorithm to handle uneven lighting or soiled ballots, and the recognition accuracy meets the following requirements: Among them, N correct is the number of correctly identified ballots, N t otal is the total number of votes.

7. The voting machine based on the one-time multiple vote recognition system of claim 6, characterized in that: The data processing module performs voting statistics based on the recognition result ν i Generate a statistics vector (R), defined as: Where: r j Indicates candidate or option c j The number of votes; δ(C j ,V i ) is the indicator function, if V i Contains C j , then δ=1, otherwise δ=0; (N) is the total number of valid votes, (M) is the total number of candidates; When the recognition result v i If a ballot contains invalid options (such as multiple choices or blanks), the system will mark it as an invalid ballot, exclude it from the count (R), and record it in the log. After the count is completed, a cross-check is performed to ensure that: The statistical vector (R) is updated in real time and stored in the storage module.

8. The voting machine based on the one-time multiple vote recognition system of claim 1, characterized in that: The storage and communication module uses the AES-256 encryption algorithm to encrypt the ballot data and statistical results. The encryption process is expressed as follows: D e =AES 256 (D,K e ) Where: (D) is the original data, including the ballot sub-image S i , unique identifier T k and statistical vector (R); K e It is a 256-bit encryption key, randomly generated by a hardware security module (HSM) and updated every 24 hours; e This is encrypted data; the encrypted data is transmitted to the cloud server via the HTTPS protocol, with a transmission delay of less than 1 second; if the transmission is interrupted, the system caches the data and retransmits it after the connection is restored to ensure data integrity.

9. The voting machine based on the one-time multiple vote recognition system of claim 1, characterized in that: include: A control center (1), a paper path (2), a ballot information collection component (3) and a printer (4); the paper path (2) transmits ballots, the ballot information collection component (3) obtains ballot information, and processes the collected ballot information through the control center (1) with a display screen to obtain election results; and the election results are printed through the printer (4); The control center (1) has a built-in one-time multiple ticket recognition system.

10. An election method based on one-time multiple vote recognition, applied to an election machine based on one-time multiple vote recognition system according to any one of claims 1 to 9, characterized in that: The following steps are involved: Verify the voter's identity through the identity verification module and generate a unique identity (ID); Capture multiple ballot images at one time through the image acquisition module; The image is segmented by the data processing module to generate the ballot sub-image S i ; The data processing module identifies the sub-image content and generates voting information v i ; Verify the unique identification of the ballot T through the ballot uniqueness verification module k , to prevent duplicate voting; The voting results are counted through the data processing module to generate a statistical vector (R); Encrypt and transmit data to the cloud through storage and communication modules; The method further comprises the step of publishing in real time: The statistical vector (R) is stored in the consortium chain. Only authorized nodes can write data. The storage process satisfies: B t =HB t-1 ||R||T s ) Among them: B t is the hash of the current block; B t-1 is the hash of the previous block; (R) is the statistical vector; T s is the timestamp in claim 4; H(·) is the SHA-256 hash function; A public query interface is provided. Users need to enter a verification code or public key signature. The query response time is less than 0.5 seconds. It is limited to the statistical vector (R) and ballot status, and does not include the voter identity (ID).

Images