Intelligent rotation verification code feature insight and content accurate identification method

Abstract

The invention relates to the field of content identification, in particular to an intelligent rotation verification code feature insight and content accurate identification method, which comprises the following steps: screening unverified intelligent rotation verification codes, collecting aiming track images of the intelligent rotation verification codes, determining track features of the aiming track images, and drawing an aiming intention graph. And marking the aiming intention graph according to the track characteristics, comparing the aiming intention graph with a coincidence threshold value, screening, reserving the intelligent rotation verification code exceeding the coincidence threshold value, performing secondary verification by taking the intelligent rotation verification code as a test verification code, collecting corresponding verification time, comparing the verification time with a time threshold value, and obtaining a verification result of the intelligent rotation verification code. The test verification codes lower than the time threshold value are reserved and serve as sample verification codes, the image quality parameters of the sample verification codes are extracted and serve as sample parameters to be applied to all the intelligent rotation verification codes, image recognition and user behavior analysis are combined, the safety and user experience of the intelligent rotation verification codes are improved, malicious machine operation is effectively prevented, and the user experience is improved. And the account security of the user is protected.

Description

Technical Field

[0001] The present invention relates to the field of content recognition, and in particular to a method for insight into the characteristics of a smart verification code and accurate content recognition. Background Art

[0002] Smart Verification Code technology is mainly based on image and text recognition. Users need to identify and drag the movable slider in the Smart Verification Code to distinguish between humans and robots. With the development of image recognition technology, simple Smart Verification Codes are gradually being cracked. Smart Verification Code feature insights and content accurate recognition methods have begun to evolve, and artificial intelligence has been introduced for optimization. By analyzing the behavioral data of the user to be verified, it is determined whether the user is a real human to improve the user experience.

[0003] With the development of artificial intelligence technology, Zhixuan's verification code feature insight and content accurate identification methods are facing new changes and challenges. Its application scope will become wider, and its human-computer interaction capabilities will gradually increase.

[0004] Chinese patent application publication number: CN110348450A discloses a security assessment method for an image verification code, including: obtaining an image verification code to be evaluated; performing denoising on the image verification code to be evaluated to obtain an image to be identified; processing the image to be identified using an image recognition model to obtain a recognition result; calculating a first matching degree between a target object contained in the image to be identified and the recognition result; and determining a security level of the image verification code to be evaluated based on the first matching degree. The disclosure also provides a security assessment device and a computer system for an image verification code.

[0005] Chinese patent application publication number: CN104125234A discloses a dynamic image security verification method, which pre-designs background, mask interference element images and word library for producing Zhixuan verification code; establishes a system domain name library for calling interface and generates a unique identifier for each domain name; in the verification process, first judges whether the user or IP is banned, if not banned, then randomly obtains the final generated verification image from the background, mask interference element image, and word library and generates a verification acquisition record, and returns the unique identifier and image address of Zhixuan verification code after the verification record is successfully generated; after the user enters the verification information on the verification image and uploads the unique identifier of Zhixuan verification code and the user input information, judges the timeliness of verification, whether the verification information is correct, whether the domain name of the user exists, and returns the processing result. This invention can effectively improve the ability to defend against mechanical data capture, and the system security is higher.

[0006] However, the above method has the following problem: it cannot analyze the user's behavior trend in combination with image recognition. Summary of the invention

[0007] To this end, the present invention provides a method for insight into the characteristics of smart verification codes and accurate content recognition, so as to overcome the problem in the prior art that it is impossible to analyze user behavior trends in combination with image recognition.

[0008] To achieve the above purpose, the present invention provides a method for insight into the characteristics of smart verification codes and accurate content recognition, including:

[0009] Screening the Zhixuan verification code that has not passed the verification, and collecting the aiming track image of the Zhixuan verification code, wherein,

[0010] The smart spin verification code includes a fixed background and a movable slider, and the fixed background is provided with a notch position;

[0011] When the movable slider is aligned with the notch, it is determined that the smart rotation verification code has passed the verification;

[0012] The aiming trajectory image is a trajectory image of the user to be verified dragging the movable slider from the initial position to the final position;

[0013] Determining the trajectory features of the aiming trajectory image, drawing an aiming intention map, and marking the aiming intention map according to the trajectory features;

[0014] The aiming intention graph is compared with the overlap threshold and screened, and the smart rotation verification codes exceeding the overlap threshold are retained and used as test verification codes for secondary verification, and the corresponding verification time is collected, wherein,

[0015] The overlap threshold is the minimum standard for the aiming trajectory image to overlap with the correct trajectory;

[0016] The verification time is compared with a time threshold, and the test verification code below the time threshold is retained as a sample verification code, wherein:

[0017] The verification time is the time required for the user to be verified to complete the dragging process and pass the secondary verification;

[0018] The image quality parameters of the sample verification code are extracted and applied as sample parameters to all smart verification codes, wherein:

[0019] The image quality parameter is the resolution corresponding to the sample verification code.

[0020] Furthermore, the step of collecting the aiming trajectory image of the smart rotation verification code includes:

[0021] Determining an initial position and a final position of the movable slider;

[0022] Determine whether the final position is a gap position, and filter the correct trajectory;

[0023] Drawing the aiming track image according to the dragging record of the movable slider;

[0024] The correct trajectory is the movement trajectory of the movable slider dragged from the initial position to the notch position by the user to be verified.

[0025] Furthermore, the step of drawing the aiming intention map includes:

[0026] Preprocessing the aiming trajectory image to generate corresponding trajectory preprocessing data;

[0027] The intention learning model learns the trajectory preprocessing data and generates a corresponding aiming intention graph;

[0028] Annotating the aiming intention graph according to the trajectory characteristics;

[0029] The intention learning model is generated by training a trajectory training set formed by trajectory preprocessing data corresponding to the aiming trajectory image.

[0030] Furthermore, the step of preprocessing the aiming trajectory image includes:

[0031] The trajectory feature is determined according to the aiming trajectory image, wherein:

[0032] The trajectory features include the dragging speed, operation time and / or trajectory direction corresponding to the aiming trajectory image;

[0033] Perform image filtering on the aiming trajectory image to obtain corresponding trajectory filtering data, wherein:

[0034] The image filtering is based on the trajectory features, removing aiming trajectory images that do not meet the trajectory features;

[0035] The trajectory filtering data is segmented according to the standard learning rate to form corresponding trajectory preprocessing data, where:

[0036] The standard learning rate is a learning rate that can be recognized by the intentional learning model, and for a single learning, the corresponding standard learning rate is a single learning rate.

[0037] Further, the step of comparing the aiming intention graph with the overlap threshold and screening includes:

[0038] Obtaining each displayed value in the aiming intention graph;

[0039] The displayed value is compared with the overlap threshold, wherein,

[0040] When the displayed value is lower than the overlap threshold, the corresponding smart rotation verification code is cleared;

[0041] When the displayed value is higher than the overlap threshold, the corresponding Zhixuan verification code is retained and sent to the user to be verified as a test verification code for the secondary verification.

[0042] Further, the steps of collecting verification time include:

[0043] When the user to be verified receives the test verification code, the timer starts timing;

[0044] When the movable slider is aligned with the notch position, the timer stops timing, and the corresponding verification time is calculated and collected;

[0045] Wherein, when the movable slider fails to pass the secondary verification, the timer automatically cancels the timing and resets the time to zero.

[0046] Further, the verification time is compared with the time threshold, wherein,

[0047] The time threshold is the maximum time required for the user to be verified to complete the secondary verification, which is related to the security performance of the Zhixuan verification code.

[0048] Furthermore, when the verification time is lower than the time threshold, the corresponding test verification code is retained and marked as a sample verification code; when the verification time is higher than the time threshold, the corresponding test verification code is cleared and marked as an error verification code.

[0049] Further, the resolution of the sample verification code is obtained, the resolution of the sample verification code is improved by using image technology, and the new resolution is converted into a corresponding image quality parameter;

[0050] Wherein, the resolution includes the height and width of the sample verification code;

[0051] The image technique is to divide a single resolution into several sub-resolutions.

[0052] Further, the image quality parameters are saved and marked as sample parameters, and a setter performs batch image processing and applies the sample parameters to all smart spin verification codes;

[0053] Among them, the batch image processing is that the setter traverses all the smart rotation verification codes and updates the corresponding image quality parameters to the sample parameters.

[0054] Compared with the prior art, the present invention screens the Zhixuan verification codes that have not passed the verification, collects the aiming trajectory images of the Zhixuan verification codes, determines the trajectory features of the aiming trajectory images, draws an aiming intention map, and marks the aiming intention map according to the trajectory features, compares the aiming intention map with the overlap threshold and screens them, retains the Zhixuan verification codes that exceed the overlap threshold and uses them as test verification codes for secondary verification, collects the corresponding verification time, compares the verification time with the time threshold, retains the test verification codes that are lower than the time threshold and uses them as sample verification codes, extracts the image quality parameters of the sample verification codes, and applies them as sample parameters to all Zhixuan verification codes. By combining image recognition and user behavior analysis, the security and user experience of the Zhixuan verification code are improved, malicious machine operation is effectively prevented, and the user's account security is protected.

[0055] Furthermore, by filtering the correct trajectory and drawing the aiming trajectory image according to the dragging record of the movable slider, the user's behavior trajectory during the verification process can be accurately captured and analyzed, thereby improving the security of the Zhixuan verification code and the accuracy of recognition.

[0056] Furthermore, by preprocessing the aiming trajectory image to generate corresponding trajectory preprocessing data, using the intention learning model to learn the trajectory preprocessing data and generating the corresponding aiming intention graph, it helps to simplify the image data, making it easier to analyze and process, facilitating further analysis of user behavior and providing feedback during the verification process, thereby improving the accuracy of the intention learning model's predictions.

[0057] Furthermore, by determining the trajectory features of the aiming trajectory image, the aiming trajectory image is filtered, and the filtered trajectory filtering data is segmented according to the standard learning rate to form trajectory preprocessing data suitable for model learning, and structured data is provided to the intention learning model so that the model can more accurately learn and identify the aiming behavior of the user to be verified, which helps the model to better understand and learn the behavior patterns of the user to be verified, effectively reduce misjudgments, and improve the user experience of the Zhixuan verification code.

[0058] Furthermore, by comparing the aiming intention graph with the overlap threshold and screening it, key data for evaluating the degree of match between user operations and expected trajectories can be obtained, which helps to filter out possible automated attacks or erroneous operations, further confirm the user's identity, and ensure that only the Zhixuan verification code that highly matches the expected trajectory will be used for secondary verification, thereby improving the security and efficiency of the verification process, effectively reducing false positives and negatives, and improving the overall effectiveness of the Zhixuan verification code system.

[0059] Furthermore, by calculating and collecting the verification time of the secondary verification, comparing the verification time with the time threshold, retaining the sample verification code and clearing the wrong verification code, the security and accuracy of the Zhixuan verification code are further enhanced. By setting a reasonable time threshold, the system can more effectively identify and distinguish between real users and robot operations, thereby improving the security of the Zhixuan verification code and ensuring that only real users can pass the verification and automated attacks are effectively blocked.

[0060] Furthermore, by optimizing the image quality of the sample verification code and applying it as a standard parameter to all Zhixuan verification codes, the consistency and high quality of Zhixuan verification codes can be ensured, which not only helps to improve the security of Zhixuan verification codes and reduce the misrecognition rate, but also enhances the user's visual experience, making Zhixuan verification codes easier to be correctly identified and processed. Through batch processing, time and resources can be saved and the efficiency of the Zhixuan verification code management system can be improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0061] Figure 1 This is a flow chart of a method for insight into the characteristics of smart verification codes and accurate content identification of the present invention;

[0062] Figure 2 A flowchart of the aiming trajectory image for collecting the smart spin verification code of the present invention;

[0063] Figure 3 Draw a flow chart of the aiming intention diagram for the present invention;

[0064] Figure 4 The present invention is a flow chart of preprocessing the aiming trajectory image. DETAILED DESCRIPTION

[0065] In order to make the objects and advantages of the present invention more clearly understood, the present invention is further described below in conjunction with embodiments; it should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0066] The preferred embodiments of the present invention are described below with reference to the accompanying drawings. It should be understood by those skilled in the art that these embodiments are only used to explain the technical principles of the present invention and are not intended to limit the protection scope of the present invention.

[0067] It should be noted that, in the description of the present invention, terms such as "up", "down", "left", "right", "inside" and "outside" indicating directions or positional relationships are based on the directions or positional relationships shown in the drawings. This is merely for the convenience of description and does not indicate or imply that the device or element must have a specific orientation, be constructed and operated in a specific orientation. Therefore, it cannot be understood as a limitation on the present invention.

[0068] In addition, it should be noted that in the description of the present invention, unless otherwise clearly specified and limited, the terms "installed", "connected", and "connected" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be indirectly connected through an intermediate medium, or it can be the internal communication of two components. For those skilled in the art, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0069] See also Figure 1 As shown, it is a flow chart of a method for insight into the characteristics of smart verification codes and accurate content identification of the present invention, including:

[0070] Step S1, screening the Zhixuan verification codes that have not passed the verification, and collecting the aiming track images of the Zhixuan verification codes, wherein:

[0071] The smart verification code includes a fixed background and a movable slider. The fixed background is set with a gap position;

[0072] When the movable slider is aligned with the notch, it is determined that the smart rotation verification code has passed the verification;

[0073] The aiming trajectory image is a trajectory image of the user to be verified dragging the movable slider from the initial position to the final position;

[0074] Step S2, determining the trajectory features of the aiming trajectory image, drawing an aiming intention map, and marking the aiming intention map according to the trajectory features;

[0075] Step S3, compare the aiming intention graph with the overlap threshold and screen it, retain the smart rotation verification code that exceeds the overlap threshold, and use it as a test verification code for secondary verification, and collect the corresponding verification time, where:

[0076] The overlap threshold is the minimum standard for the aiming trajectory image to overlap with the correct trajectory;

[0077] Step S4, compare the verification time with the time threshold, retain the test verification code below the time threshold and use it as a sample verification code, where:

[0078] Verification time refers to the time required for the user to be verified to complete the dragging process and pass the secondary verification;

[0079] Step S5, extracting the image quality parameters of the sample verification code and applying them as sample parameters to all smart verification codes, wherein:

[0080] The image quality parameter is the resolution corresponding to the sample verification code.

[0081] By screening the Zhixuan verification codes that have not passed the verification and collecting the aiming trajectory images of the Zhixuan verification codes, determining the trajectory features of the aiming trajectory images, drawing the aiming intention map, and marking the aiming intention map according to the trajectory features, the aiming intention map is compared with the overlap threshold and screened, the Zhixuan verification codes that exceed the overlap threshold are retained and used as test verification codes for secondary verification, the corresponding verification time is collected, the verification time is compared with the time threshold, the test verification codes that are lower than the time threshold are retained and used as sample verification codes, the image quality parameters of the sample verification codes are extracted, and applied as sample parameters to all Zhixuan verification codes, and by combining image recognition and user behavior analysis, the security and user experience of the Zhixuan verification code are improved, malicious machine operations are effectively prevented, and the user's account security is protected.

[0082] See also Figure 2 As shown, it is a flow chart of collecting the aiming trajectory image of the smart rotation verification code of the present invention, including:

[0083] Step S11, determining the initial position and final position of the movable slider;

[0084] Step S12, determining whether the final position is a gap position, and filtering the correct trajectory;

[0085] Step S13, drawing an aiming track image according to the dragging record of the movable slider;

[0086] The correct trajectory is the moving trajectory of the movable slider dragged from the initial position to the gap position by the user to be verified.

[0087] In the specific implementation, it is checked whether the final position of the movable slider matches the gap position in the smart spin verification code. If the final position is correct (that is, the movable slider is dragged to the gap position), it is considered to be a correct track. If not, this track is considered to be an error track and can be filtered out.

[0088] When recording the dragging record, all position change data of the user to be verified in the process of dragging the movable slider are collected, and the trajectory image of the user dragging the slider is drawn using this data, that is, the aiming trajectory image. By recording and analyzing the behavior trajectory of the user to be verified dragging the movable slider, the behavior pattern of the user to be verified during the verification process can be captured. The filtering of the correct trajectory helps to identify and exclude automated attacks or abnormal user behavior, because automated scripts usually cannot accurately simulate the dragging behavior of human users.

[0089] By filtering the correct trajectory and drawing the aiming trajectory image according to the dragging record of the movable slider, the user's behavior trajectory during the verification process can be accurately captured and analyzed, thereby improving the security of the Zhixuan verification code and the accuracy of recognition.

[0090] See also Figure 3 As shown, it is a flow chart of drawing a sighting intention diagram according to the present invention, including:

[0091] Step St1, preprocessing the aiming trajectory image to generate corresponding trajectory preprocessing data;

[0092] Step St2, the intention learning model learns the trajectory preprocessing data and generates a corresponding aiming intention graph;

[0093] Step St3, marking the aiming intention graph according to the trajectory features;

[0094] Among them, the intention learning model is trained and generated by training a trajectory training set formed by trajectory preprocessing data corresponding to the aiming trajectory image.

[0095] In the specific implementation, the trajectory preprocessing data is learned using the intention learning model. The intention learning model is generated by training the trajectory training set through a machine learning algorithm, such as a deep learning network. The training set consists of a large number of aiming trajectory images and their corresponding trajectory preprocessing data. Through training, the model can learn the characteristics of the trajectory data and predict the user's aiming intention. Annotating the aiming intention map according to the trajectory characteristics involves identifying and marking the key points or areas in the aiming intention map for further analysis and verification.

[0096] By preprocessing the aiming trajectory image and generating the corresponding trajectory preprocessing data, the intention learning model is used to learn the trajectory preprocessing data and generate the corresponding aiming intention graph, which helps to simplify the image data and make it easier to analyze and process, facilitate further analysis of user behavior, and provide feedback during the verification process, thereby improving the accuracy of the intention learning model prediction.

[0097] See also Figure 4 As shown, it is a flow chart of the present invention for preprocessing the aiming trajectory image, including:

[0098] Step Sp1, determining the trajectory features according to the aiming trajectory image, wherein:

[0099] The trajectory characteristics include the dragging speed, operation time and / or trajectory direction corresponding to the aiming trajectory image;

[0100] Step Sp2, performing image filtering on the aiming trajectory image to obtain corresponding trajectory filtering data, wherein:

[0101] Image filtering is based on trajectory features, and aiming trajectory images that do not meet the trajectory features are removed;

[0102] Step Sp3, segmenting the trajectory filtering data according to the standard learning rate to form corresponding trajectory preprocessing data, where:

[0103] The standard learning rate is a learning rate that the intentional learning model can recognize, and for single-shot learning, its corresponding standard learning rate is a single learning rate.

[0104] In a specific implementation, when the standard learning rate is between 160 and 180 times per second, the intentional learning model has the best analysis effect on trajectory preprocessing data. Preferably, the standard learning rate is 170 times per second.

[0105] The purpose of filtering the aiming trajectory images is to remove those images that do not meet the specific trajectory characteristics to ensure the data quality of subsequent processing. For example, if the dragging speed of an image is too fast or too slow, or the operation time is abnormal, then the image may be filtered out. After filtering, the trajectory filtering data is segmented according to the standard learning rate to form trajectory preprocessing data. The standard learning rate here refers to the learning rate that the intention learning model can recognize and process. For single learning, the corresponding standard learning rate is a single learning rate, which means that the amount of data for each learning is fixed so that the model can learn and predict effectively.

[0106] By determining the trajectory features of the aiming trajectory image, filtering the aiming trajectory image, and segmenting the filtered trajectory filtering data according to the standard learning rate, trajectory preprocessing data suitable for model learning is formed, and structured data is provided to the intention learning model so that the model can more accurately learn and identify the aiming behavior of the user to be verified, which helps the model to better understand and learn the behavior patterns of the user to be verified, effectively reduce misjudgments, and improve the user experience of the Zhixuan verification code.

[0107] Specifically, the steps of comparing the aiming intention map with the coincidence threshold and screening include:

[0108] Obtain each displayed value in the aiming intention graph;

[0109] The displayed value is compared to the coincidence threshold, where

[0110] When the displayed value is lower than the overlap threshold, the corresponding smart rotation verification code will be cleared;

[0111] When the displayed value is higher than the overlap threshold, the corresponding Zhixuan verification code is retained and sent to the user to be verified as a test verification code for secondary verification.

[0112] In some possible implementations, the overlap threshold is set to 80%. If the A value in the aiming intention graph is 70%, which is lower than the overlap threshold, it will be considered that the user's operation does not match the expected trajectory enough, so the corresponding A smart rotation verification code will be cleared, which helps to filter out possible automated attacks or erroneous operations.

[0113] If the B value in the aiming intention graph is 90%, which is higher than the overlap threshold, the corresponding B smart rotation verification code will be retained and sent to the user to be verified as a test verification code for secondary verification. This step is to further confirm the user's identity.

[0114] In the specific implementation, various display values ​​are obtained from the aiming intention map. These values ​​represent the trajectory characteristics of the user's operation, such as dragging speed, operation time or trajectory direction, etc. These characteristics are used to evaluate the degree of match between the user's operation and the expected trajectory. The displayed value is compared with the preset overlap threshold, which is a pre-set standard used to judge the degree of match between the user's operation trajectory and the expected trajectory. Only the Zhixuan verification code that highly matches the expected trajectory will be used for secondary verification, which reduces unnecessary verification steps and improves the efficiency of the verification process.

[0115] By comparing the aiming intention graph with the overlap threshold and screening it, we can obtain key data to evaluate the degree of match between user operations and expected trajectories, which helps to filter out possible automated attacks or erroneous operations, further confirm the user's identity, and ensure that only Zhixuan verification codes that highly match the expected trajectory are used for secondary verification, thereby improving the security and efficiency of the verification process, effectively reducing false positives and negatives, and improving the overall effectiveness of the Zhixuan verification code system.

[0116] Specifically, the steps to collect verification time include:

[0117] When the user to be verified receives the test verification code, the timer starts counting;

[0118] When the movable slider is aligned with the notch position, the timer stops timing, and the corresponding verification time is calculated and collected;

[0119] Among them, when the movable slider fails to pass the secondary verification, the timer automatically cancels the timing and resets the time to zero.

[0120] In some possible implementations, if the user to be verified receives the test verification code at 12:00, the timer starts timing, and if the user to be verified drags the movable slider to align with the notch position at 12:01, the timer stops timing, and the verification time is one minute. If the user to be verified receives the test verification code at 12:00, and the secondary verification fails at 12:01, the timer automatically cancels the timing and resets the time to zero.

[0121] In a specific implementation, when the user to be verified receives the test verification code, the system starts a timer, which marks the beginning of the user's secondary verification process, and the timer records the time required for the user to complete the verification from receiving the verification code.

[0122] Specifically, the verification time is compared with a time threshold, where

[0123] The time threshold is the maximum time required for the user to be verified to complete the secondary verification, which is related to the security performance of the Zhixuan verification code.

[0124] In a specific implementation, if the user's verification time is higher than the time threshold, the system will consider that the user's operation speed is too slow, which may indicate that the user is not a human but an automated robot or script. In this case, the system will require the user to re-verify or take other security measures.

[0125] If the user's verification time is within the time threshold, the system considers the user's verification operation to be normal and can continue with the subsequent identity authentication process.

[0126] The setting of the time threshold is not static. It is adjusted according to changes in security requirements, statistics of user behavior, and changes in attack patterns. By dynamically adjusting the time threshold, the system can better adapt to different security environments and improve its adaptability and security.

[0127] By comparing the verification time with the time threshold, the system can effectively filter out automated attacks, reduce false positives and false negatives, and improve the security of the verification process. At the same time, reasonable time threshold settings can also help improve user experience and avoid normal users being mistakenly rejected due to overly strict security measures.

[0128] Specifically, when the verification time is lower than the time threshold, the corresponding test verification code is retained and marked as a sample verification code; when the verification time is higher than the time threshold, the corresponding test verification code is cleared and marked as an error verification code.

[0129] With this strategy, possible security threats can be effectively identified and handled while minimizing disruption to normal users.

[0130] Keeping sample verification codes helps the system learn and adapt to new attack patterns, while clearing incorrect verification codes helps reduce the waste of system resources. The system can dynamically adjust time thresholds and other security parameters based on the statistical data and analysis results of sample verification codes and incorrect verification codes to improve the adaptability and accuracy of the system. This data can also be used to provide feedback to users to help them understand possible security risks and best practices, which not only helps improve the security of the Zhixuan verification code system, but also optimizes the system's performance and user experience.

[0131] In some possible implementations, if the time threshold is 10s, the verification time of test verification code A is monitored to be 8s, and the verification time is lower than the time threshold, the corresponding test verification code is retained and marked as a sample verification code; if the verification time of test verification code B is monitored to be 12s, and the verification time is higher than the time threshold, the test verification code B is cleared and marked as an incorrect verification code.

[0132] By calculating and collecting the verification time of the secondary verification, comparing the verification time with the time threshold, retaining sample verification codes and clearing incorrect verification codes, the security and accuracy of Zhixuan verification codes are further enhanced. By setting a reasonable time threshold, the system can more effectively identify and distinguish between real users and robot operations, thereby improving the security of Zhixuan verification codes and ensuring that only real users can pass the verification and automated attacks are effectively blocked.

[0133] Specifically, the resolution of the sample verification code is obtained, the resolution of the sample verification code is improved by using image technology, and the new resolution is converted into a corresponding image quality parameter;

[0134] The resolution includes the height and width of the sample verification code;

[0135] The image processing technique is to divide a single resolution into several sub-resolutions.

[0136] In a specific implementation, it is necessary to determine the original resolution of the sample verification code, which includes the height and width of the verification code image. Resolution is an important parameter of image quality, which determines the clarity and detail of the image. The resolution of the sample verification code is improved by image technology. Preferably, image super-resolution technology can be used to improve the image resolution by machine learning algorithms. This technology may include using deep learning models, such as convolutional neural networks, to predict and enhance high-resolution details of images.

[0137] Image technology can split a single resolution into several sub-resolutions. In specific implementations, the GrabCut algorithm can be used to segment an image into different regions or objects. In addition, clustering-based segmentation methods such as K-means and SLIC algorithms can also be used to segment an image into multiple superpixels, which can be regarded as sub-resolution units.

[0138] After the resolution is increased, the new resolution can be converted into image quality parameters. These parameters include line pairs per picture height (LP / PH), lines per millimeter (L / mm), cycles per millimeter (Cyc les / mm), and line width per picture height (LW / PH). These parameters help evaluate the clarity and detail representation of the image.

[0139] Through the above steps, the resolution of the sample verification code can be effectively improved, and its new resolution can be converted into corresponding image quality parameters, thereby providing higher quality image data for subsequent image processing and recognition tasks.

[0140] Specifically, the image quality parameters are saved and marked as sample parameters, and the setter performs batch image processing and applies the sample parameters to all smart spin verification codes;

[0141] Among them, batch image processing is that the setter traverses all the smart rotation verification codes and updates the corresponding image quality parameters to sample parameters.

[0142] In the specific implementation, after improving the resolution of the sample verification code and converting it into image quality parameters, these parameters are saved, including LP / PH, L / mm, Cycles / mm and LW / PH, which represent the clarity and details of the image. The saved image quality parameters are marked as "sample parameters". These sample parameters will be used as a benchmark or reference for quality control in subsequent batch processing of Zhixuan verification code images.

[0143] Use a configurator (which may be a software tool or script to automate the process) to iterate over all the Wisdom Captcha images. This configurator will be responsible for applying the sample parameters to each captcha image.

[0144] During batch processing, the setter updates the corresponding image quality parameters to the sample parameters. This means that each Wisdom Captcha image will be adjusted according to the sample parameters to ensure that they all reach the same quality standard.

[0145] The specific operation of applying sample parameters includes adjusting the resolution, contrast, brightness, etc. of the image to ensure the clarity and recognizability of the image. This can be achieved through various image processing techniques, such as interpolation, image enhancement, etc.

[0146] After batch processing, the Zhixuan CAPTCHA images are verified to ensure that they all meet the expected image quality standards. If problems are found, they can be fed back into the processing flow, the sample parameters can be adjusted, and the batch processing can be repeated. In this way, it can be ensured that all Zhixuan CAPTCHA images have consistent and high image quality, which helps improve the recognition rate of CAPTCHAs and user experience. In addition, this also helps reduce security risks caused by image quality issues, such as automated attacks.

[0147] By optimizing the image quality of the sample verification code and applying it as a standard parameter to all Zhixuan verification codes, the consistency and high quality of Zhixuan verification codes can be ensured. This not only helps to improve the security of Zhixuan verification codes and reduce the misrecognition rate, but also enhances the user's visual experience, making Zhixuan verification codes easier to be correctly identified and processed. Through batch processing, time and resources can be saved and the efficiency of the Zhixuan verification code management system can be improved.

[0148] So far, the technical solutions of the present invention have been described in conjunction with the preferred embodiments shown in the accompanying drawings. However, it is easy for those skilled in the art to understand that the protection scope of the present invention is obviously not limited to these specific embodiments. Without departing from the principle of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will fall within the protection scope of the present invention.

[0149] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for insight into the characteristics of smart verification codes and accurate content recognition, characterized in that: include: Screening the Zhixuan verification code that has not passed the verification, and collecting the aiming track image of the Zhixuan verification code, wherein, The smart spin verification code includes a fixed background and a movable slider, and the fixed background is provided with a notch position; When the movable slider is aligned with the notch, it is determined that the smart rotation verification code has passed the verification; The aiming trajectory image is a trajectory image of the user to be verified dragging the movable slider from the initial position to the final position; Determining the trajectory features of the aiming trajectory image, drawing an aiming intention map, and marking the aiming intention map according to the trajectory features; The aiming intention graph is compared with the overlap threshold and screened, and the smart rotation verification codes exceeding the overlap threshold are retained and used as test verification codes for secondary verification, and the corresponding verification time is collected, wherein, The overlap threshold is the minimum standard for the aiming trajectory image to overlap with the correct trajectory; The verification time is compared with a time threshold, and the test verification code below the time threshold is retained as a sample verification code, wherein: The verification time is the time required for the user to be verified to complete the dragging process and pass the secondary verification; The image quality parameters of the sample verification code are extracted and applied as sample parameters to all smart verification codes, wherein: The image quality parameter is the resolution corresponding to the sample verification code.

2. The method for insight into the characteristics of smart verification codes and accurate content recognition according to claim 1 is characterized in that: The steps of collecting the aiming trajectory image of the Zhixuan verification code include: Determining an initial position and a final position of the movable slider; Determine whether the final position is a gap position, and filter the correct trajectory; Drawing the aiming track image according to the dragging record of the movable slider; The correct trajectory is the movement trajectory of the movable slider dragged from the initial position to the notch position by the user to be verified.

3. The method for insight into the characteristics of smart verification codes and accurate content recognition according to claim 2 is characterized in that: The steps to drawing a targeting intent map include: Preprocessing the aiming trajectory image to generate corresponding trajectory preprocessing data; The intention learning model learns the trajectory preprocessing data and generates a corresponding aiming intention graph; Annotating the aiming intention graph according to the trajectory characteristics; The intention learning model is generated by training a trajectory training set formed by trajectory preprocessing data corresponding to the aiming trajectory image.

4. The method for insight into the characteristics of smart verification codes and accurate content identification according to claim 3 is characterized in that: The steps of preprocessing the aiming trajectory image include: The trajectory feature is determined according to the aiming trajectory image, wherein: The trajectory features include the dragging speed, operation time and / or trajectory direction corresponding to the aiming trajectory image; Perform image filtering on the aiming trajectory image to obtain corresponding trajectory filtering data, wherein: The image filtering is based on the trajectory features, removing aiming trajectory images that do not meet the trajectory features; The trajectory filtering data is segmented according to the standard learning rate to form corresponding trajectory preprocessing data, where: The standard learning rate is a learning rate that can be recognized by the intentional learning model, and for a single learning, the corresponding standard learning rate is a single learning rate.

5. The method for insight into the characteristics of smart verification codes and accurate content identification according to claim 4 is characterized in that: The steps of comparing the aiming intention map with the coincidence threshold and screening include: Obtaining each displayed value in the aiming intention graph; The displayed value is compared with the overlap threshold, wherein, When the displayed value is lower than the overlap threshold, the corresponding smart rotation verification code is cleared; When the displayed value is higher than the overlap threshold, the corresponding Zhixuan verification code is retained and sent to the user to be verified as a test verification code for the secondary verification.

6. The method for insight into the characteristics of smart verification codes and accurate content identification according to claim 5 is characterized in that: The steps to collect verification time include: When the user to be verified receives the test verification code, the timer starts timing; When the movable slider is aligned with the notch position, the timer stops timing, and the corresponding verification time is calculated and collected; Wherein, when the movable slider fails to pass the secondary verification, the timer automatically cancels the timing and resets the time to zero.

7. The method for insight into the characteristics of smart verification codes and accurate content identification according to claim 6 is characterized in that: The verification time is compared with the time threshold, wherein, The time threshold is the maximum time required for the user to be verified to complete the secondary verification, which is related to the security performance of the Zhixuan verification code.

8. The method for insight into the characteristics of smart verification codes and accurate content identification according to claim 7 is characterized in that: When the verification time is lower than the time threshold, the corresponding test verification code is retained and marked as a sample verification code. When the verification time is higher than the time threshold, the corresponding test verification code is cleared and marked as an error verification code.

9. The method for insight into the characteristics of smart verification codes and accurate content identification according to claim 8 is characterized in that: Obtaining the resolution of the sample verification code, improving the resolution of the sample verification code by using image technology, and converting the new resolution into a corresponding image quality parameter; Wherein, the resolution includes the height and width of the sample verification code; The image technique is to divide a single resolution into several sub-resolutions.

10. The method for insight into the characteristics of smart verification codes and accurate content recognition according to claim 9 is characterized in that: The image quality parameters are saved and marked as sample parameters, and a setter performs batch image processing and applies the sample parameters to all smart spin verification codes; Among them, the batch image processing is that the setter traverses all the smart rotation verification codes and updates the corresponding image quality parameters to the sample parameters.

Images