Vision change monitoring system based on image processing

Abstract

The invention discloses a vision change monitoring system based on image processing, relates to the technical field of image processing and vision monitoring, and provides the following scheme: the vision change monitoring system comprises an image acquisition module, an environment adaptation module, an image processing module, a vision evaluation module, a data storage and analysis module and an early warning module, the specific structure is as follows: the image acquisition module is used for performing data information acquisition processing on an eye image and a visual mark visual image of a testee; according to the method, full-link closed loop of environment perception-parameter adjustment-attitude guidance-scene adaptation is realized, the standardization degree of the monitoring environment is improved, monitoring errors caused by environment interference are reduced, accurate calculation of naked eye / corrected vision is realized, and vision change reasons and risk points are deeply analyzed, so that the visual acuity of the user is improved, and the visual acuity of the user is improved. Meanwhile, value-added services of group comparison and trend prediction are provided for users, and the application scene of the system is expanded.

Description

TECHNICAL FIELD

[0001] The application relates to the technical field of image processing and vision monitoring, and particularly relates to a vision change monitoring system based on image processing. BACKGROUND

[0002] With the popularity of electronic devices and the change of living habits, the incidence of vision problems such as myopia and hyperopia is increasing year by year, and the vision health of the youth group has become the focus of social attention. The traditional vision monitoring method mainly relies on a standard visual acuity chart, and the vision level is judged by the subjective identification of the visual acuity of the tested person. The method has the following defects: first, the subjectivity is strong, and the judgment deviation and psychological state of the tested person will affect the accuracy of the monitoring result; second, the monitoring process needs to be operated by a professional, and cannot realize self-monitoring and normalization; third, only the visual acuity value can be obtained, and it is difficult to capture the subtle trend of vision change, and it is difficult to provide comprehensive data support for vision intervention; fourth, the contact type monitoring auxiliary tool (such as an optometric lens) may have a cross-infection risk, and is not suitable for infants, the elderly and other special groups. In the prior art, part of the vision monitoring schemes based on image recognition can only realize the vision evaluation under a single visual acuity, and do not consider the dynamic characteristics of the eyes, the environmental light interference and other factors, the monitoring precision is low, and meanwhile, the historical data analysis and trend prediction functions are lacked, so that the long-term vision tracking demand cannot be met, and therefore, an objective, accurate, convenient and dynamic tracking vision change monitoring system which can overcome the above defects is urgently needed. SUMMARY

[0003] The vision change monitoring system based on image processing provided by the application solves the above defects in the prior art.

[0004] In order to achieve the above purpose, the application adopts the following technical scheme: A vision change monitoring system based on image processing, comprising an image acquisition module, an environment adaptation module, an image processing module, a vision evaluation module, a data storage and analysis module and an early warning module, and the specific structure is as follows: The image acquisition module is used for collecting and processing the data information of the eye images and the visual acuity marking images of the tested person; The environment adaptation module is used for dynamically adjusting the monitoring environment parameters and the posture of the tested person, and ensuring the standardization of the monitoring conditions; The image processing module is used for deeply processing, feature mining and feature extraction of the collected eye images and visual acuity marking image data; The vision evaluation module is used for accurately calculating the visual acuity value and deeply analyzing the vision change condition based on the image processing result; The data storage and analysis module is used for securely storing data and analysis results, and supports deep analysis processing. The early warning module is used for issuing corresponding early warning prompt processing according to the visual assessment result.

[0005] Further, the image acquisition module includes a high-definition camera, an eye positioning sensor, and a visual target display unit. The resolution of the high-definition camera is not less than 1080P, and the frame rate is not less than 30 fps, which is used for capturing eye details such as pupil diameter, eyelid state, and eye rotation angle. The visual target display unit can display various types of visual targets (such as E-shaped visual targets, logarithmic visual targets, and dynamic visual targets), and supports adaptive adjustment of visual target size, definition, and display position. The eye positioning sensor realizes accurate positioning of the subject's eye and the camera through infrared positioning technology, ensuring the stability of image acquisition.

[0006] Further, the environment adaptation module includes an environment perception unit, a parameter adjustment unit, a posture guidance unit, and a scene adaptation unit, which work cooperatively. The environment perception unit serves as the perception core of environment adaptation, and includes a light sensor, a color temperature sensor, a humidity sensor, and a dust concentration sensor. The light sensor has a sampling frequency of ≥1 Hz, and can collect environmental light intensity in real time with an accuracy of ±20 lux, providing a data basis for brightness adjustment. The color temperature sensor detects the color temperature of the environment (range 2700K-6500K), and triggers a color temperature compensation instruction when the color temperature deviates from the optimal visual target observation range (4000K-5000K). The humidity sensor and the dust concentration sensor respectively monitor the relative humidity and the PM2.5 concentration of the environment, with a humidity warning threshold of <30% or >70% and a concentration warning threshold of >75 μg / m 3 , to avoid the influence of extreme environments on the subject's eye state and image acquisition clarity. The parameter adjustment unit performs adjustment actions based on environment perception data, including a brightness compensation subunit, a color temperature calibration subunit, and an environment purification linkage subunit. The brightness compensation subunit is internally provided with an LED light compensation array (color temperature adjustable), which is used for timely light compensation processing of the collection environment. When the light intensity is <500 lux, the light compensation intensity increases by 15% for every 100 lux decrease in light intensity, and the light compensation calculation formula is as follows: ; wherein I is the environmental light intensity (unit: lux), which is collected by the light sensor (sampling frequency ≥1 Hz, accuracy ±20 lux). 500 lux is the supplemental lighting activation threshold; 1500 lux is the anti-glare activation threshold; When the light intensity is greater than 1500 lux, activate the anti-glare filter (with adjustable transmittance) to reduce the interference of ambient light on the target display.

[0007] Furthermore, the color temperature calibration subunit controls the difference between the target's emitted color temperature and the ambient light color temperature by adjusting the backlight color temperature of the target display unit, keeping it within ±500K to avoid target recognition deviation caused by color shift. The color temperature calibration control formula is as follows: ; in, The backlight color temperature of the target display unit (adjustable range 2700K-6500K); Ambient light color temperature (detected by a color temperature sensor); This is the maximum permissible color temperature difference. The environmental purification linkage subunit is used to automatically send instructions to associated humidifiers and air purifiers when humidity or dust concentration exceeds the warning threshold, adjusting environmental parameters to a suitable range (humidity 40%-60%, PM2.5 ≤ 50 μg / m³). 3 ); The posture guidance unit is used to ensure that the relative position between the subject and the device meets the monitoring standards, and includes a distance sensing subunit, a sitting posture detection subunit, and a voice and visual guidance subunit. The distance sensing subunit uses an infrared ranging sensor (measuring range 30-100cm, accuracy ±1cm) to detect the distance between the subject's eyes and the visual target display unit in real time. When the distance is <50cm or >70cm, a distance deviation signal is generated. The sitting posture detection subunit captures the upper body posture of the subject through the auxiliary camera of the image acquisition module, and uses a skeletal key point recognition algorithm to determine whether there is a hunched back and head tilt >15° poor sitting posture. If so, a posture deviation signal is generated. After receiving distance and posture deviation signals, the voice and visual guidance subunit guides the test subject to correct their position and posture until they meet the standard through voice (such as "Please adjust your sitting posture backward and maintain a distance of 60cm") and dynamic guidance icons (such as arrows indicating the adjustment direction) displayed by the visual target unit. The scene adaptation unit optimizes environmental parameters for different usage scenarios (home, school, hospital), and includes a scene mode storage subunit and a parameter one-click switching subunit; The scene mode storage subunit presets three types of mode parameters: home mode (wider ambient light tolerance range: 400-1600 lux), school screening mode (fast adjustment, adaptable to continuous monitoring of multiple people, adjustment response time < 2 seconds), and hospital professional mode (strictly follows medical-grade monitoring standards, light / color temperature fluctuations are controlled within ±5%). The parameter one-click switching subunit allows users to select scene modes through the system interface or hardware buttons, and automatically calls the corresponding parameters after switching, eliminating the need for repeated manual adjustment.

[0008] Furthermore, the image processing module includes an image preprocessing unit, a region segmentation unit, a dynamic feature extraction unit, a static feature extraction unit, and a feature optimization unit, with each unit progressively enhancing the others. The image preprocessing unit is used to eliminate image noise and interference, laying the foundation for subsequent processing. It includes a noise filtering subunit, an illumination equalization subunit, and a distortion correction subunit. The noise filtering subunit employs a three-stage filtering algorithm consisting of Gaussian filtering (removing high-frequency noise), median filtering (removing salt-and-pepper noise), and bilateral filtering (preserving edge details) to denoise the eye image, thereby improving the image signal-to-noise ratio to ≥35dB. The illumination equalization subunit is used to address the problem of uneven brightness caused by iris shadows and eyelid occlusion in eye images. It uses the Retinex algorithm to optimize local illumination, so that the brightness difference between the pupil area and the iris area is controlled within ±15 gray values. The distortion correction subunit corrects lens distortion (radial distortion and tangential distortion) based on the camera's intrinsic parameter matrix (obtained in advance using Zhang's calibration method), ensuring that the geometric distortion rate of the eye region is <2%. The distortion correction formula is as follows: ; in, These are the pixel coordinates of the image before correction; These are the corrected pixel coordinates; The square of the distance from the pixel to the center of the image; Radial distortion coefficient; The tangential distortion coefficient; The region segmentation unit is used to accurately separate key areas of the eye, including a pupil segmentation subunit, an iris segmentation subunit, and an eyelid and periocular skin segmentation subunit. The pupil segmentation subunit is used to combine threshold segmentation (based on gray value differences, pupil gray value is usually <50) with a circular fitting algorithm to extract the pupil outline and calculate the pupil center coordinates and diameter with an accuracy of ±0.1mm. The iris segmentation subunit uses the Canny edge detection algorithm to identify the boundary between the iris and sclera (the gray value of the iris is usually between 80 and 120), excludes the area obscured by the pupil and eyelid, and obtains the texture features and area parameters of the complete iris region. The eyelid and periocular skin segmentation subunit separates the upper / lower eyelids and periocular skin regions through skin color model (YCrCb color space) and edge contour analysis, and calculates the eyelid opening and closing degree. The normal range is 0.8-1.0. If it is lower than 0.6, it is judged as eyelid occlusion, and the image needs to be re-acquired. The formula for calculating the eyelid opening and closing degree is as follows: ; in, The vertical distance between the lowest point of the upper eyelid and the highest point of the lower eyelid (in pixels, converted to physical distance by image scaling). This is the actual diameter of the iris region (approximately 12 mm for a normal adult, used as a baseline value). Normal range: 0.8-1.0.

[0009] Furthermore, the dynamic feature extraction unit is used to capture the dynamic changes when the subject gazes at the target, and includes an eye movement tracking subunit, a pupil dynamic monitoring subunit, and a blink behavior analysis subunit; The eye movement tracking subunit calculates the eye rotation angle based on the pupil center coordinate changes of continuous frame images using optical flow. The horizontal rotation range is -45° to +45°, and the vertical rotation range is -30° to +30°. It generates an eye movement trajectory and analyzes the delay time of the eye following the visual target. The normal delay is ≤0.2 seconds, and an excessively long delay indicates visual abnormality. The pupil dynamic monitoring subunit is used to record the pupil diameter change curve when the optotype is switched (such as from a large optotype to a small optotype), calculate the pupil constriction rate (normal ≥0.5mm / s) and constriction amplitude (normal ≥0.3mm), and reflect the eye's accommodation function; The blinking behavior analysis subunit is used to statistically analyze the blinking frequency (normal range: 10-15 times / minute) and blinking duration (normal ≤0.3 seconds) during the monitoring process, and to exclude invalid monitoring data of frequent blinking (>20 times / minute) and excessively long blinking time; The static feature extraction unit is used to extract static physiological parameters of the eye, including an iris texture extraction subunit, a corneal reflective point detection subunit, and an axial length estimation subunit. The iris texture extraction subunit uses a Gabor filter to extract texture feature points of the iris (≥50 feature points are extracted from each image) for subject identification (to prevent others from taking the test on their behalf), and at the same time analyzes the uniformity of the iris texture (abnormal texture may indicate eye diseases). The corneal reflection point detection subunit identifies the reflection point formed by ambient light on the corneal surface, calculates the distance between the reflection point and the center of the pupil (normally ≤1mm), and determines whether there is astigmatism in the cornea (a distance deviation >1.5mm indicates the risk of astigmatism). The axial length estimation subunit estimates the axial length based on the iris diameter (approximately 12mm in normal adults) and the image pixel ratio, combined with a corneal curvature estimation model, to provide a preliminary estimate of the axial length (error ±0.3mm), thus providing a reference for determining the type of myopia (axial myopia / refractive myopia). The feature optimization unit improves the effectiveness and reliability of feature data and includes an anomaly data removal subunit, a feature fusion subunit, and a feature standardization subunit. The abnormal data removal subunit adopts the 3σ criterion to remove data that exceed the normal physiological range in dynamic features, such as eye rotation angle > 60° and pupil contraction amplitude < 0.1mm, and retains valid samples; The feature fusion subunit integrates multi-dimensional features into a comprehensive eye feature vector through a weighted fusion algorithm (dynamic feature weight 0.6, static feature weight 0.4), reducing the limitations of a single feature. The feature standardization subunit maps the feature vector to the [0, 1] interval (using the Min-Max standardization algorithm), making the feature data of different subjects comparable and facilitating subsequent matching with the vision-related feature library.

[0010] Furthermore, the vision assessment module includes a basic vision calculation unit, a corrected vision correction unit, a vision change analysis unit, a vision risk assessment unit, and a multi-dimensional assessment report generation unit. The basic vision calculation unit calculates uncorrected visual acuity based on image processing results, and includes a target response analysis subunit, a vision conversion subunit, and a vision accuracy calibration subunit. The target response analysis subunit is used to analyze the subject's fixation characteristics for targets of different sizes / directions (e.g., fixation duration ≥ 0.5 seconds and no obvious eye movement is considered valid recognition) and determine the largest recognizable target (i.e. the smallest target that the subject can accurately recognize). The visual acuity conversion subunit is used to select the corresponding conversion formula based on the optotype type to calculate the initial uncorrected visual acuity value. The E optotype uses the logarithmic visual acuity chart formula as follows: (Formula follows). ; Among them, the target spacing is the actual physical distance (unit: meters) of the currently displayed target, which is output by the target display unit according to a preset ratio; 5 meters is the standard observation distance reference value for the logarithmic visual acuity chart; The dynamic target uses a custom formula as follows: ; The target movement speed is the current movement rate of the dynamic target (unit: cm / s), which is output in real time by the target display unit; The maximum recognition speed is the maximum target movement speed that the subject (or group of the same age) can recognize (unit: cm / s) preset by the system, based on historical data or group benchmark values; The visual acuity calibration subunit is used to introduce an eye dynamic feature correction coefficient to calibrate the initial visual acuity value, thereby reducing the error range to within ±0.03 visual acuity values. The calculation formula is as follows: ; The initial visual acuity value is the original visual acuity result obtained through the above optotype conversion formula; The correction coefficient for eye dynamic features is set based on the results of dynamic feature detection. It is 1.0 under normal conditions, 0.95 when the pupil contraction rate is low (<0.5mm / s), and 0.9 when the eye tracking delay is too long (>0.2 seconds). Other abnormal conditions are adjusted according to the corresponding proportion. The corrected visual acuity correction unit is used to calculate corrected visual acuity for subjects who wear glasses and contact lenses, and includes a corrected parameter acquisition subunit, a parameter verification subunit, and a corrected visual acuity calculation subunit; The correction parameter acquisition subunit supports two acquisition methods: manual input (glasses power, astigmatism power, and axis) and image recognition (extracting lens optical parameters by taking a picture of the glasses lens with a camera, with an accuracy of ±25 degrees). The parameter verification subunit is used to compare the collected parameters with a common correction parameter library. If the deviation exceeds the normal range (e.g., myopia > 1000 degrees), the user is prompted to reconfirm or calibrate with professional equipment. The corrected visual acuity calculation subunit uses an optical correction model to calculate the corrected visual acuity value and labels the correction method (frames / contact lenses) and parameter confidence level (high / medium / low), as shown in the following formula: .

[0011] Furthermore, the vision change analysis unit is used to trace historical data and analyze vision change trends, and includes a data filtering subunit, a change rate calculation subunit, a trend fitting subunit, and a preliminary analysis subunit for the causes of change; The data screening subunit retrieves monitoring data from the past 1-3 years from the data storage module, and screens data samples with consistent monitoring environment (light / distance error <10%) and stable subject condition (no record of eye disease) to ensure the accuracy of the analysis; The rate of change calculation subunit uses a sliding window algorithm (window size = 3 monitoring times) to calculate the rate of visual acuity change over different time periods (e.g., monthly or quarterly changes), distinguishing between short-term fluctuations (monthly change ± 0.02) and long-term trends (quarterly change > 0.05). The formula for calculating the rate of visual acuity change is as follows: ; Among them, the visual acuity value within the window is the visual acuity value of three consecutive monitoring tests that meet the requirements of consistent monitoring environment and stable condition of the subject after screening. The time interval is the time difference between the first and last monitoring within the window (unit: month / quarter, set according to analysis needs); The trend fitting subunit uses a quadratic function to fit the filtered sample data and determine the trend type, as shown in the following formula: ; Where y is the visual acuity value (dependent variable); x represents time (independent variable, unit: month, with the first monitoring time as the starting point x=0); a, b, and c are fitting coefficients, which are calculated using the least squares method on the filtered historical data (last 1-3 years). Stable (a≈0, b≈0); Slow decline (a < 0, b < 0, |b| < 0.02 / month); Rapid decline (a < 0, b < 0, |b| ≥ 0.02 / month); Slowly rising (a>0, b>0, |b|<0.01 / month); The preliminary analysis subunit for the causes of the changes is used to combine eye behavior data (such as the duration of electronic device use and outdoor activity time, which requires user authorization) to analyze the causes of vision changes (such as daily electronic device use > 4 hours and outdoor activity < 1 hour, which may lead to decreased vision). The vision risk assessment unit determines the risk level based on vision value and its trend, and includes a risk indicator definition subunit, a multi-dimensional scoring subunit, and a risk level classification subunit. The risk indicator definition subunit is used to define three types of core risk indicators, namely: current vision level (referencing national standards), vision change range (change over the past 6 months), and trend stability (R-squared of the fitted curve). 2 Value, R 2 The closer to 1, the more stable the trend. The multi-dimensional scoring subunit is used to score the three types of indicators respectively, with a full score of 100 points. The current vision level is as follows: 5.0 and above gets 100 points, 4.8-4.9 gets 80 points, 4.3-4.7 gets 50 points, and ≤4.2 gets 20 points. Variation range: <0.1 gets 100 points, 0.1-0.3 gets 60 points, ≥0.3 gets 20 points; Trend stability: R 2 ≥0.8 gets 100 points, 0.6-0.8 gets 70 points, <0.6 gets 40 points. Calculate the weighted total score (weight: current level 0.4, change range 0.3, trend stability 0.3). The calculation formula is as follows: ; The risk level classification subunit is used to classify risk levels based on the total score, and the classification is as follows: Low risk (80-100 points); Medium risk (50-79 points); High risk (<50 points), with key risk points marked, such as: high risk: visual acuity decreased by 0.2 in the past 6 months, and the trend is stable; The multi-dimensional assessment report generation subunit is used to integrate assessment results and generate a visualization report, including a data visualization subunit, a suggestion generation subunit, and a report export subunit. The data visualization sub-unit uses line charts to show the trend of visual acuity changes, bar charts to compare uncorrected visual acuity and corrected visual acuity, and radar charts to present multi-dimensional risk scores, making the results intuitive and easy to understand. The suggestion generation subunit generates personalized suggestions based on the risk level and the reason for the change, which are divided into low risk: maintain current eye use habits and monitor once every 3 months; Medium risk: Reduce electronic device use to less than 2 hours per day and conduct outdoor viewing exercises 3 times a week; High risk: It is recommended to visit an ophthalmology hospital for a professional examination within one week to determine the cause of the worsening myopia; The report export sub-unit supports generating reports in PDF, Word, and Excel formats. It includes four main modules: basic information (name, age, monitoring time), vision data (uncorrected / corrected visual acuity, trend of change), risk assessment (level, key indicators), and personalized suggestions, making it convenient for users to save or share with medical personnel.

[0012] Furthermore, the data storage and analysis module includes a data classification and storage unit, a data security protection unit, a data retrieval unit, a deep data analysis unit, and a data sharing unit: The data classification storage unit is used for hierarchical storage according to data type and importance, and includes basic data sub-units, feature data sub-units, evaluation result sub-units, and behavior-related data sub-units; The basic data subunit is used to store the basic information of the test subject (name, age, gender, ID number, encrypted storage), monitoring equipment information (equipment number, calibration record) and environmental parameters (light intensity, color temperature, distance, sorted by monitoring timestamp), and is stored in a local database (SQLite), retaining data for nearly 1 year; The feature data subunit is used to store the processed eye feature data (pupil diameter curve, iris texture feature vector, eye movement trajectory coordinates), and is stored in binary format with compression (compression rate ≥50%). Nearly 3 months of data are retained locally and backed up in the cloud for a long time. The assessment results sub-unit is used to store vision assessment reports (raw data, calculation process, and final results), and is stored in JSON format. It supports quick retrieval by time range and assessment type (naked / corrected), and is updated synchronously on both local and cloud platforms. The behavior-related data subunit is used to store user-authorized eye behavior data (electronic device usage time, reading distance, sleep time), which is stored in association with vision data by timestamp to provide data support for in-depth analysis. It is stored only in the cloud (encrypted) and retains nearly 3 years of data. The data security protection unit is used to protect data privacy and integrity, and includes a data encryption subunit, an access control subunit, a data backup and recovery subunit, and an abnormal access monitoring subunit. The data encryption subunit employs dual-layer encryption, including transmission encryption (using the HTTPS protocol, with the key updated hourly) and storage encryption (sensitive data such as ID card numbers and eye images are encrypted using AES-256, while ordinary data such as environmental parameters are encrypted using DES). The key is set by the user and stored locally to prevent data leakage from the cloud. The access control subunit sets up three levels of access permissions, specifically: User level: Users can view their own data and export reports; Family level: Parents can view their children's data, but authorization from the children is required; Medical-grade: Doctors can view authorized patient data, which requires hospital qualification certification. Each access records the visitor's identity, time, and operation content. The data backup and recovery subunit is used for daily automatic backup of local data (incremental backup, only backing up changed data), and cloud data is stored with 3 copies (distributed to servers in different regions), supporting manual / automatic recovery (recovery time < 10 minutes) to prevent data loss; The abnormal access monitoring subunit uses AI behavior analysis algorithms to identify abnormal access behaviors (such as frequent login attempts from different IP addresses and multiple exports of large amounts of data within a short period of time), triggering alerts (freezing accounts, sending SMS verification), and recording abnormal logs for administrators to view.

[0013] Furthermore, the data retrieval unit is used to provide efficient data query functions, including a precise retrieval subunit, a fuzzy retrieval subunit, and a batch retrieval subunit; The precise retrieval subunit supports retrieval by unique identifier (such as monitoring timestamp, report number), with a retrieval response time of <0.5 seconds and returns complete data records; The fuzzy search subunit supports keyword search (e.g., corrected visual acuity of 4.9 in a high-risk assessment in March 2024), uses a full-text search algorithm, returns the top 10 results with matching scores, and annotates the matching keywords; The batch retrieval subunit supports filtering and exporting data in batches based on conditions (such as all records with uncorrected visual acuity <4.8 from January 2023 to January 2024), and supports setting export fields (such as exporting only visual acuity values ​​and monitoring time), thereby improving retrieval efficiency; The deep data analysis unit is used to mine potential patterns based on historical data, and includes a group comparison analysis subunit, a correlation factor analysis subunit, and a prediction model subunit. The group comparison analysis subunit is used to compare the test subject's data with a group database of the same age group (such as the average vision of 12-year-old teenagers and the vision distribution of 25-30-year-old adults) to generate a group ranking (such as your vision ranking in the top 300 out of 1000 people in the same age group) to help users understand their own vision level. The correlation factor analysis subunit uses Pearson correlation analysis to calculate the correlation coefficient between eye use behavior data and visual acuity changes (e.g., the correlation coefficient r=0.7 between electronic device usage time and visual acuity decline, showing a strong positive correlation), and identifies key influencing factors; The prediction model subunit is used to train an LSTM neural network prediction model based on vision and behavioral data from the past two years to predict vision change trends over the next 6-12 months (prediction accuracy ≥85%). If current eye habits are maintained, vision may drop to 4.7 after 6 months, providing a basis for early intervention. The data sharing unit supports compliant data sharing and includes an authorization management subunit, a data anonymization subunit, and a sharing record subunit. The authorization management subunit adopts the OAuth2.0 authorization protocol. Users authorize the medical platform / school to obtain data for a specified time period through a one-time authorization code (such as authorizing XX Hospital to view the assessment report from January to March 2024). The authorization validity period can be set (1 day to 1 year) and will automatically expire upon expiration. The data desensitization subunit is used to automatically desensitize sensitive information (such as hiding the middle 6 digits of the ID number and blurring the iris texture of the eye image) during data sharing, and only retain necessary information such as vision data, evaluation results and key environmental parameters to prevent privacy leaks. The shared record subunit is used to record each data sharing behavior (including the sharing object, time, data range and authorization validity period), so that users can view and revoke authorizations that have not expired at any time and protect data control rights.

[0014] Compared with existing technologies, the beneficial effects of this invention are: 1. This invention realizes a closed-loop process of environmental perception, parameter adjustment, attitude guidance, and scene adaptation through an environmental adaptation module, which improves the standardization of the monitoring environment by 40% and effectively reduces monitoring errors caused by environmental interference. 2. This invention improves the effective extraction rate of eye features to over 95% through dynamic / static feature separation, extraction, and optimization in the image processing module, providing more comprehensive data support for vision assessment and further reducing vision calculation errors; 3. This invention not only achieves accurate calculation of uncorrected / corrected visual acuity through the multi-dimensional vision assessment module, but also deeply analyzes the causes and risk points of vision changes, significantly enhancing the practicality and guidance of the assessment results; 4. Through the security protection and in-depth analysis functions of the data storage and analysis module, this invention provides users with value-added services such as group comparison and trend prediction while ensuring data privacy, thus expanding the application scenarios of the system. In summary, this device not only achieves a closed-loop system encompassing environmental perception, parameter adjustment, posture guidance, and scene adaptation, thereby improving the standardization of the monitoring environment and reducing monitoring errors caused by environmental interference, but also enables accurate calculation of uncorrected / corrected visual acuity and in-depth analysis of the causes and risks of visual acuity changes. This significantly enhances the practicality and guidance of the assessment results. Furthermore, it provides users with value-added services such as group comparison and trend prediction, expanding the system's application scenarios. Attached Figure Description

[0015] Figure 1 This is a flowchart of the overall system of a vision change monitoring system based on image processing proposed in this invention. Figure 2 This is a block diagram of the overall system modules of a vision change monitoring system based on image processing proposed in this invention; Figure 3 This is a block diagram of the image acquisition module of a vision change monitoring system based on image processing proposed in this invention; Figure 4 This is a block diagram of the environment adaptation module of an image processing-based vision change monitoring system proposed in this invention; Figure 5 This is a block diagram of the image processing module of a vision change monitoring system based on image processing proposed in this invention; Figure 6This is a block diagram of the vision assessment module of a vision change monitoring system based on image processing proposed in this invention; Figure 7 This is a block diagram of the data storage and analysis module of a vision change monitoring system based on image processing proposed in this invention; Figure 8 This is a block diagram of the early warning module of an image processing-based vision change monitoring system proposed in this invention. Detailed Implementation

[0016] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0017] In the description of this invention, it should be understood that the terms upper, lower, front, back, left, right, top, bottom, inside, outside, etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limiting this invention.

[0018] Example Reference Figures 1-8 A vision change monitoring system based on image processing includes an image acquisition module, an environment adaptation module, an image processing module, a vision assessment module, a data storage and analysis module, and an early warning module, with the following specific structure: The image acquisition module includes a high-definition camera, an eye positioning sensor, and a target display unit. The high-definition camera has a resolution of no less than 1080P and a frame rate of no less than 30fps, used to capture eye details (such as pupil diameter, eyelid state, and eyeball rotation angle). The target display unit can display various types of targets (such as E-targets, logarithmic targets, and dynamic targets) and supports adaptive adjustment of target size, clarity, and display position. The eye positioning sensor uses infrared positioning technology to achieve precise alignment between the subject's eyes and the camera, ensuring the stability of image acquisition.

[0019] In this invention, the environment adaptation module includes an environment sensing unit, a parameter adjustment unit, a posture guidance unit, and a scene adaptation unit, all of which work collaboratively: the environment sensing unit, as the core of environment adaptation, includes a light sensor, a color temperature sensor, a humidity sensor, and a dust concentration sensor; the light sensor has a sampling frequency ≥1Hz, collects ambient light intensity in real time with an accuracy of ±20 lux, providing a data basis for brightness adjustment; the color temperature sensor detects the ambient light color temperature (range 2700K—6500K), and triggers a color temperature compensation command when the color temperature deviates from the optimal target observation range (4000K—5000K); the humidity sensor and the dust concentration sensor monitor the ambient relative humidity and PM2.5 concentration, respectively, with a humidity warning threshold of <30% or >70% and a concentration warning threshold of >75μg / m³. 3 To avoid the impact of extreme environments on the subject's eye condition and image acquisition clarity, the parameter adjustment unit performs adjustments based on environmental perception data. This includes a brightness compensation subunit, a color temperature calibration subunit, and an environmental purification linkage subunit. The brightness compensation subunit has a built-in LED supplementary lighting array (adjustable color temperature) for timely supplementary lighting of the acquisition environment. When the light intensity is <500 lux, the supplementary lighting intensity dynamically increases by 15% for every 100 lux decrease in light intensity. The light compensation calculation formula is as follows: ; Where I is the ambient light intensity (unit: lux), which is collected by a light sensor (sampling frequency ≥ 1 Hz, accuracy ± 20 lux). 500 lux is the supplemental lighting activation threshold; 1500 lux is the anti-glare activation threshold; When the light intensity is greater than 1500 lux, activate the anti-glare filter (with adjustable transmittance) to reduce the interference of ambient light on the target display.

[0020] In this invention, the color temperature calibration subunit controls the difference between the target's emitted color temperature and the ambient light color temperature by adjusting the backlight color temperature of the target display unit, keeping it within ±500K to avoid target recognition deviation caused by color shift. The color temperature calibration control formula is as follows: ; in, The backlight color temperature of the target display unit (adjustable range 2700K-6500K); Ambient light color temperature (detected by a color temperature sensor); This is the maximum permissible color temperature difference. The environmental purification linkage subunit automatically sends instructions to associated humidifiers and air purifiers when humidity or dust concentration exceeds the warning threshold, adjusting environmental parameters to a suitable range (humidity 40%–60%, PM2.5 ≤ 50 μg / m³). 3 ); The posture guidance unit ensures that the relative position between the subject and the device meets the monitoring standards. It includes a distance sensing subunit, a sitting posture detection subunit, and a voice and visual guidance subunit. The distance sensing subunit uses an infrared ranging sensor (measurement range 30-100cm, accuracy ±1cm) to detect the distance between the subject's eyes and the visual target display unit in real time. When the distance is <50cm or >70cm, a distance deviation signal is generated. The sitting posture detection subunit captures the subject's upper body posture through an auxiliary camera in the image acquisition module. It uses a skeletal key point recognition algorithm to determine whether there is poor sitting posture such as hunching or head tilt >15°. If so, a posture deviation signal is generated. After receiving the distance and posture deviation signals, the voice and visual guidance subunit guides the subject to correct their position and posture through voice (e.g., please adjust your sitting posture backward to maintain a 60cm distance) and dynamic guidance icons on the visual target display unit (e.g., arrows indicating the adjustment direction) until they meet the standards. The scene adaptation unit optimizes environmental parameters for different usage scenarios (home, school, hospital), including a scene mode storage subunit and a parameter one-click switching subunit. The scene mode storage sub-unit presets three modes: home mode (wider ambient light tolerance range: 400-1600 lux), school screening mode (fast adjustment, adaptable to continuous monitoring of multiple people, adjustment response time <2 seconds), and hospital professional mode (strictly follows medical-grade monitoring standards, light / color temperature fluctuations are controlled within ±5%). The one-click parameter switching subunit allows users to select scene modes through the system interface or hardware buttons. After switching, the corresponding parameters are automatically called, eliminating the need for repeated manual adjustments.

[0021] In this invention, the image processing module includes an image preprocessing unit, a region segmentation unit, a dynamic feature extraction unit, a static feature extraction unit, and a feature optimization unit, with each unit progressively building upon the previous one. The image preprocessing unit is used to eliminate image noise and interference, laying the foundation for subsequent processing. It includes a noise filtering subunit, an illumination equalization subunit, and a distortion correction subunit. The noise filtering subunit uses a three-stage filtering algorithm: Gaussian filtering (removing high-frequency noise), median filtering (removing salt-and-pepper noise), and bilateral filtering (preserving edge details) to denoise the eye image, improving the signal-to-noise ratio to ≥35dB. The illumination equalization subunit addresses the uneven brightness caused by iris shadows and eyelid occlusion in the eye image by using the Retinex algorithm to optimize local illumination, keeping the brightness difference between the pupil and iris regions within ±15 grayscale values. The distortion correction subunit corrects lens distortion (radial and tangential distortion) based on the camera's intrinsic parameter matrix (obtained in advance using Zhang's calibration method), ensuring that the geometric distortion rate of the eye region is <2%. The distortion correction formula is as follows: ; in, These are the pixel coordinates of the image before correction; These are the corrected pixel coordinates; The square of the distance from the pixel to the center of the image; Radial distortion coefficient; The tangential distortion coefficient; The region segmentation unit is used to accurately separate key areas of the eye, including a pupil segmentation subunit, an iris segmentation subunit, and an eyelid and periocular skin segmentation subunit. The pupil segmentation subunit combines threshold segmentation (based on grayscale differences, where pupil grayscale values ​​are typically <50) with a circular fitting algorithm to extract the pupil contour and calculate the pupil center coordinates and diameter, with an accuracy of ±0.1mm. The iris segmentation subunit uses the Canny edge detection algorithm to identify the boundary between the iris and sclera (iris grayscale values ​​are typically 80-120), excluding areas obscured by the pupil and eyelids, and obtaining the texture features and area parameters of the complete iris region. The eyelid and periocular skin segmentation subunit separates the upper / lower eyelids and periocular skin regions through a skin color model (YCrCb color space) and edge contour analysis, calculating the eyelid opening degree. The normal range is 0.8-1.0; values ​​below 0.6 are considered eyelid occlusion, requiring image re-acquisition. The formula for calculating the eyelid opening degree is as follows: ; in, The vertical distance between the lowest point of the upper eyelid and the highest point of the lower eyelid (in pixels, converted to physical distance by image scaling). This is the actual diameter of the iris region (approximately 12 mm for a normal adult, used as a baseline value). Normal range: 0.8-1.0.

[0022] In this invention, the dynamic feature extraction unit is used to capture the dynamic changes of the subject when gazing at the visual target. It includes an eye movement tracking subunit, a pupil dynamic monitoring subunit, and a blink behavior analysis subunit. The eye movement tracking subunit calculates the eye rotation angle using optical flow based on the pupil center coordinate changes in consecutive frame images. The horizontal rotation range is -45° to +45°, and the vertical rotation range is -30° to +30°. It generates an eye movement trajectory and analyzes the delay time of eye tracking of the visual target. A normal delay is ≤0.2 seconds; excessive delay indicates visual abnormality. The pupil dynamic monitoring subunit records visual target switching (e.g., from large...) The pupil diameter change curve when the target becomes smaller (normal ≥0.5mm / s) is calculated to measure the pupil constriction rate (normal ≥0.3mm) and constriction amplitude (normal ≥0.3mm), reflecting the eye's accommodative function. The blink behavior analysis subunit is used to statistically analyze the blink frequency (normal range: 10-15 times / minute) and blink duration (normal ≤0.3 seconds) during the monitoring process, excluding invalid monitoring data of frequent blinking (>20 times / minute) and excessively long eye closure time. The static feature extraction unit is used to extract static physiological parameters of the eye, including the iris texture extraction subunit, the corneal reflective point detection subunit, and the axial length estimation subunit. The iris texture extraction subunit uses a Gabor filter to extract texture feature points of the iris (≥50 feature points per image) for subject identification (to prevent proxy testing) and analyzes the uniformity of the iris texture (abnormal texture may indicate eye diseases). The corneal reflective point detection subunit identifies reflective points formed by ambient light on the corneal surface, calculates the distance between the reflective point and the pupil center (normal ≤1mm), and determines whether astigmatism exists (distance deviation >1.5mm indicates astigmatism risk). The axial length estimation subunit, based on the iris diameter (normal adult approximately 12mm) and image pixel ratio, combined with a corneal curvature estimation model, preliminarily estimates the axial length (error ±0.3mm), providing a reference for determining myopia type (axial myopia / refractive myopia). The feature optimization unit improves the effectiveness and reliability of feature data, including an anomaly removal subunit, a feature fusion subunit, and a feature standardization subunit. The anomaly removal subunit uses the 3σ criterion to remove data in dynamic features where the eye rotation angle is greater than 60° and the pupil constriction amplitude is less than 0.1mm, which are outside the normal physiological range, while retaining valid samples. The feature fusion subunit integrates multi-dimensional features into a comprehensive eye feature vector through a weighted fusion algorithm (dynamic feature weight 0.6, static feature weight 0.4), reducing the limitations of single features. The feature standardization subunit maps the feature vector to the [0, 1] interval (using the Min-Max standardization algorithm), making the feature data of different subjects comparable and facilitating subsequent matching with the vision-related feature library.

[0023] In this invention, the vision assessment module includes a basic vision calculation unit, a corrected vision correction unit, a vision change analysis unit, a vision risk assessment unit, and a multi-dimensional assessment report generation unit. The basic vision calculation unit calculates uncorrected visual acuity based on image processing results and includes a target response analysis subunit, a vision conversion subunit, and a vision accuracy calibration subunit. The target response analysis subunit analyzes the subject's fixation characteristics on targets of different sizes / directions (e.g., fixation duration ≥ 0.5 seconds and no significant eye movement is considered valid recognition), determining the largest recognizable target (i.e., the smallest target the subject can accurately identify). The vision conversion subunit selects the corresponding conversion formula based on the target type to calculate the initial uncorrected visual acuity value. The E-target uses the following logarithmic visual acuity chart formula: [Formula omitted]. ; Among them, the target spacing is the actual physical distance (unit: meters) of the currently displayed target, which is output by the target display unit according to a preset ratio; 5 meters is the standard observation distance reference value for the logarithmic visual acuity chart; The dynamic target uses a custom formula as follows: ; The target movement speed is the current movement rate of the dynamic target (unit: cm / s), which is output in real time by the target display unit; The maximum recognition speed is the maximum target movement speed that the subject (or group of the same age) can recognize (unit: cm / s) preset by the system, based on historical data or group benchmark values; The visual acuity calibration subunit is used to introduce eye dynamic feature correction coefficients to calibrate the initial visual acuity value, reducing the error range to within ±0.03 visual acuity values. The calculation formula is as follows: ; The initial visual acuity value is the original visual acuity result obtained through the above optotype conversion formula; The correction coefficient for eye dynamic features is set based on the results of dynamic feature detection. It is 1.0 under normal conditions, 0.95 when the pupil contraction rate is low (<0.5mm / s), and 0.9 when the eye tracking delay is too long (>0.2 seconds). Other abnormal conditions are adjusted according to the corresponding proportion. The corrected visual acuity correction unit is used to calculate corrected visual acuity for test subjects who wear glasses and contact lenses. It includes a parameter acquisition subunit, a parameter verification subunit, and a corrected visual acuity calculation subunit. The parameter acquisition subunit supports two acquisition methods: manual input (glasses power, astigmatism power, and axis) and image recognition (extracting lens optical parameters from a camera image of the glasses lens, with an accuracy of ±25 degrees). The parameter verification subunit compares the acquired parameters with a common correction parameter library. If the deviation exceeds the normal range (e.g., myopia > 1000 degrees), the user is prompted to reconfirm or undergo calibration using professional equipment. The corrected visual acuity calculation subunit uses an optical correction model to calculate the corrected visual acuity value and labels the correction method (glasses / contact lenses) and parameter confidence level (high / medium / low), as shown in the following formula: .

[0024] In this invention, the vision change analysis unit is used to trace historical data and analyze vision change trends. It includes a data screening subunit, a change rate calculation subunit, a trend fitting subunit, and a preliminary analysis subunit for the causes of change. The data screening subunit retrieves monitoring data from the past 1-3 years from the data storage module, selecting data samples with consistent monitoring environments (light / distance error <10%) and stable subject conditions (no record of eye diseases) to ensure analysis accuracy. The change rate calculation subunit uses a sliding window algorithm (window size = 3 monitoring sessions) to calculate the vision change rate over different time periods (e.g., monthly or quarterly changes), distinguishing between short-term fluctuations (monthly change ±0.02) and long-term trends (quarterly change >0.05). The formula for calculating the vision change rate is as follows: ; Among them, the visual acuity value within the window is the visual acuity value of three consecutive monitoring tests that meet the requirements of consistent monitoring environment and stable condition of the subject after screening. The time interval is the time difference between the first and last monitoring within the window (unit: month / quarter, set according to analysis needs); The trend fitting subunit uses a quadratic function to fit the filtered sample data and determine the trend type, as shown in the following formula: ; Where y is the visual acuity value (dependent variable); x represents time (independent variable, unit: month, with the first monitoring time as the starting point x=0); a, b, and c are fitting coefficients, which are calculated using the least squares method on the filtered historical data (last 1-3 years). Stable (a≈0, b≈0); Slow decline (a < 0, b < 0, |b| < 0.02 / month); Rapid decline (a < 0, b < 0, |b| ≥ 0.02 / month); Slowly rising (a>0, b>0, |b|<0.01 / month); The preliminary analysis sub-unit for the causes of changes is used to analyze the reasons for changes in vision by combining eye-use behavior data (such as the duration of electronic device use and outdoor activity time, which requires user authorization). For example, daily electronic device use > 4 hours and outdoor activity < 1 hour may lead to decreased vision. The vision risk assessment sub-unit determines the risk level based on vision values ​​and trends, and includes a risk indicator definition sub-unit, a multi-dimensional scoring sub-unit, and a risk level classification sub-unit. The risk indicator definition sub-unit is used to define three types of core risk indicators: current vision level (referencing national standards), the magnitude of vision change (change over the past 6 months), and trend stability (R-squared of the fitted curve). 2 Value, R 2 The closer to 1, the more stable the trend. The multi-dimensional scoring sub-unit is used to score the three types of indicators separately, with a full score of 100 points. Current vision level: 5.0 and above gets 100 points, 4.8-4.9 gets 80 points, 4.3-4.7 gets 50 points, and ≤4.2 gets 20 points. Variation range: <0.1 gets 100 points, 0.1-0.3 gets 60 points, ≥0.3 gets 20 points; Trend stability: R 2 ≥0.8 gets 100 points, 0.6-0.8 gets 70 points, <0.6 gets 40 points. Calculate the weighted total score (weight: current level 0.4, change range 0.3, trend stability 0.3). The calculation formula is as follows: ; The risk level classification subunit is used to classify risk levels based on the total score. The risk levels are classified as follows: Low risk (80-100 points); Medium risk (50-79 points); High risk (<50 points), with key risk points marked, such as: high risk: visual acuity decreased by 0.2 in the past 6 months, and the trend is stable; The multi-dimensional assessment report generation sub-unit is used to integrate assessment results and generate a visual report, including a data visualization sub-unit, a suggestion generation sub-unit, and a report export sub-unit. The data visualization sub-unit uses line charts to show the trend of vision changes, bar charts to compare uncorrected and corrected visual acuity, and radar charts to present multi-dimensional risk scores, making the results intuitive and easy to understand. It is recommended to generate personalized suggestions based on the risk level and the reason for the change, which are divided into low risk: maintain current eye use habits and monitor once every 3 months; Medium risk: Reduce electronic device use to less than 2 hours per day and conduct outdoor viewing exercises 3 times a week; High risk: It is recommended to visit an ophthalmology hospital for a professional examination within one week to determine the cause of the worsening myopia; The report export sub-unit supports generating reports in PDF, Word, and Excel formats. It includes four main modules: basic information (name, age, monitoring time), vision data (uncorrected / corrected visual acuity, trend of change), risk assessment (level, key indicators), and personalized suggestions, making it convenient for users to save or share with medical personnel.

[0025] In this invention, the data storage and analysis module includes a data classification and storage unit, a data security protection unit, a data retrieval unit, a deep data analysis unit, and a data sharing unit. The data classification storage unit is used for hierarchical storage according to data type and importance, and includes basic data sub-units, feature data sub-units, evaluation result sub-units, and behavior-related data sub-units; The basic data sub-unit is used to store the basic information of the test subjects (name, age, gender, ID number, encrypted storage), monitoring equipment information (equipment number, calibration record) and environmental parameters (light intensity, color temperature, distance, sorted by monitoring timestamp), and uses a local database (SQLite) for storage, retaining data for nearly 1 year; The feature data sub-unit is used to store the processed eye feature data (pupil diameter curve, iris texture feature vector, eye movement trajectory coordinates), which is compressed and stored in binary format (compression rate ≥50%). Nearly 3 months of data are retained locally and backed up in the cloud for a long time. The assessment results sub-unit is used to store vision assessment reports (raw data, calculation process, and final results). It is stored in JSON format and supports quick retrieval by time range and assessment type (naked / corrected). It is updated synchronously on both local and cloud platforms. The behavior-related data sub-unit is used to store user-authorized eye behavior data (electronic device usage time, reading distance, sleep time), which is stored in association with vision data by timestamp to provide data support for in-depth analysis. It is stored only in the cloud (encrypted) and retains nearly 3 years of data. The data security protection unit is used to protect data privacy and integrity, and includes a data encryption subunit, an access control subunit, a data backup and recovery subunit, and an abnormal access monitoring subunit. The data encryption subunit employs dual-layer encryption, including transmission encryption (using the HTTPS protocol, with the key updated hourly) and storage encryption (sensitive data such as ID card numbers and eye images are encrypted using AES-256, while ordinary data such as environmental parameters are encrypted using DES). The key is set by the user and stored locally to prevent data leakage from the cloud. The access control subunit is configured with three levels of access permissions, specifically: User level: Users can view their own data and export reports; Family level: Parents can view their children's data, but authorization from the children is required; Medical-grade: Doctors can view authorized patient data, which requires hospital qualification certification. Each access records the visitor's identity, time, and operation content. The data backup and recovery subunit is used for daily automatic backup of local data (incremental backup, only backing up changed data), and cloud data is stored with 3 copies (distributed to servers in different regions), supporting manual / automatic recovery (recovery time < 10 minutes) to prevent data loss; The abnormal access monitoring subunit uses AI behavior analysis algorithms to identify abnormal access behaviors (such as frequent login attempts from IP addresses in different locations and multiple exports of large amounts of data within a short period of time), triggering alerts (freezing accounts, sending SMS verification), and recording abnormal logs for administrators to view.

[0026] In this invention, the data retrieval unit is used to provide efficient data query functions, including a precise retrieval subunit, a fuzzy retrieval subunit, and a batch retrieval subunit; The precise retrieval sub-unit supports retrieval by unique identifier (such as monitoring timestamp, report number), with a retrieval response time of less than 0.5 seconds, and returns complete data records; The fuzzy search sub-unit supports keyword retrieval (e.g., corrected visual acuity of 4.9 in a high-risk assessment in March 2024), uses a full-text search algorithm, returns the top 10 results with matching scores, and annotates the matching keywords; The batch search sub-unit supports filtering and exporting data in batches based on conditions (such as all records with uncorrected visual acuity <4.8 from January 2023 to January 2024), and supports setting export fields (such as exporting only visual acuity values ​​and monitoring time), thereby improving search efficiency; The deep data analysis unit is used to mine potential patterns based on historical data, and includes a group comparison analysis subunit, a correlation factor analysis subunit, and a prediction model subunit. The group comparison analysis subunit is used to compare the test subject's data with a database of groups of the same age (such as the average vision of 12-year-old teenagers and the vision distribution of 25-30-year-old adults) to generate a group ranking (such as your vision ranking in the top 300 out of 1000 people in the same age group) to help users understand their own vision level. The correlation factor analysis sub-unit uses Pearson correlation analysis to calculate the correlation coefficient between eye use behavior data and visual acuity changes (e.g., the correlation coefficient r=0.7 between electronic device usage time and visual acuity decline, showing a strong positive correlation), and identifies key influencing factors; The prediction model subunit is used to train an LSTM neural network prediction model based on vision and behavioral data from the past two years to predict vision change trends over the next 6-12 months (prediction accuracy ≥85%). If current eye habits are maintained, vision may drop to 4.7 after 6 months, providing a basis for early intervention. The data sharing unit supports compliant data sharing and includes an authorization management subunit, a data anonymization subunit, and a sharing record subunit. The authorization management sub-unit adopts the OAuth2.0 authorization protocol. Users authorize the medical platform / school to obtain data for a specified time period through a one-time authorization code (e.g., authorize XX hospital to view the assessment report from January to March 2024). The authorization validity period can be set (1 day to 1 year) and will automatically expire upon expiration. The data desensitization subunit is used to automatically desensitize sensitive information (such as hiding the middle 6 digits of the ID number and blurring the iris texture in eye images) during data sharing, retaining only necessary information such as vision data, assessment results and key environmental parameters to prevent privacy leaks. The shared record sub-unit is used to record every data sharing behavior (including the sharing object, time, data scope and authorization validity period), so that users can view and revoke authorizations that have not expired at any time, thus ensuring data control.

[0027] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. An image processing based vision change monitoring system comprising of image acquisition module, environment adaptation module, image processing module, vision assessment module, data storage and analysis module and warning module, characterized in that, The specific structure is as follows: The image acquisition module is used for collecting and processing data information of the eye image and the fixation image of the target; The environment adaptation module is used for dynamically adjusting the monitoring environment parameters and the posture of the target to ensure the standardization of the monitoring conditions; The image processing module is used for deep processing, feature mining and feature extraction of the collected eye image and fixation image data; The visual acuity evaluation module is used for accurately calculating the visual acuity value and deeply analyzing the visual acuity change based on the image processing result; The data storage and analysis module is used for safely storing data and analysis results, and supporting deep analysis processing; The early warning module is used for issuing corresponding early warning prompt processing according to the visual acuity evaluation result.

2. The image processing based vision change monitoring system according to claim 1, wherein, The image acquisition module includes a high-definition camera, an eye positioning sensor and a target display unit; The high-definition camera is used for capturing eye details; The target display unit is used for displaying various types of videos, and supports adaptive adjustment of target size, definition and display position; The eye positioning sensor realizes accurate positioning of the eye of the target and the camera through infrared positioning technology, and ensures the stability of image acquisition.

3. The image processing based vision change monitoring system according to claim 1, wherein, The environment adaptation module includes an environment perception unit, a parameter adjustment unit, a posture guiding unit and a scene adaptation unit: The environment perception unit is the perception core of environment adaptation, including a light sensor, a color temperature sensor, a humidity sensor and a dust concentration sensor; The light sensor is used for real-time collection of environmental light intensity to provide data basis for brightness adjustment; The color temperature sensor is used for detecting the color temperature of the environment, and triggering a color temperature compensation instruction when the color temperature deviates from the optimal target observation range; The humidity sensor and the dust concentration sensor respectively monitor the relative humidity and the PM2.5 concentration of the environment to avoid the influence of extreme environment on the eye state of the target and the definition of image acquisition; The parameter adjustment unit executes adjustment actions based on environment perception data, including a brightness compensation subunit, a color temperature calibration subunit and an environment purification linkage subunit; The brightness compensation subunit is internally provided with an LED light compensation array for timely light compensation processing of the collection environment.

4. The image processing based vision change monitoring system of claim 3, wherein, The color temperature calibration subunit adjusts the backlight color temperature of the target display unit to control the difference between the target light color temperature and the ambient light color temperature; The environment purification linkage subunit is used for automatically sending instructions to the associated humidifier and air purifier when the humidity or dust concentration exceeds the early warning threshold to adjust the environmental parameters to the appropriate range; The posture guiding unit is used to ensure that the relative position of the target and the device meets the monitoring standard, including a distance sensing subunit, a sitting posture detection subunit, a voice and visual guiding subunit; The distance sensing subunit uses an infrared distance measuring sensor to detect the distance between the eye of the target and the target display unit in real time; The sitting posture detection subunit captures the upper body posture of the target through the auxiliary camera of the image acquisition module, uses a skeleton key point recognition algorithm to judge the sitting posture, and generates a posture deviation signal according to the judgment result; The voice and visual guiding subunit receives the distance and posture deviation signals, and guides the target to correct the position and posture through the dynamic guiding icons of the voice and target display unit until the standard is met. The scene adaptation unit is used for optimizing the environmental parameters for different use scenarios, and includes a scene mode storage subunit and a parameter one-key switching subunit. The scene mode storage subunit presets three types of mode parameters, including a home mode, a school screening mode and a hospital professional mode. The parameter one-key switching subunit supports a user to select a scene mode through a system interface or a hardware key, and automatically calls corresponding parameters after switching, without the need for repeated manual adjustment.

5. The image processing based vision change monitoring system according to claim 1, wherein, The image processing module includes an image preprocessing unit, a region segmentation unit, a dynamic feature extraction unit, a static feature extraction unit and a feature optimization unit, and each unit is progressively arranged. The image preprocessing unit is used for eliminating image noise and interference, and laying a foundation for subsequent processing, and includes a noise filtering subunit, an illumination balancing subunit and a distortion correction subunit. The noise filtering subunit is used for performing noise reduction processing on an eye image, so that the image signal-to-noise ratio is improved. The illumination balancing subunit is used for controlling the luminance difference between a pupil region and an iris region in view of the problem of uneven brightness caused by iris shadows and eyelid occlusion in the eye image. The distortion correction subunit corrects lens distortion based on an intrinsic matrix of a camera, and ensures the geometric shape distortion rate of the eye region. The region segmentation unit is used for accurately separating key regions of the eye, and includes a pupil segmentation subunit, an iris segmentation subunit and an eyelid and periorbital skin segmentation subunit. The pupil segmentation subunit is used for extracting a pupil contour by combining threshold segmentation and circular fitting algorithm, and calculating pupil center coordinates and a diameter. The iris segmentation subunit identifies the boundary of the iris and the sclera by an edge detection algorithm, excludes the pupil and the eyelid occlusion region, and obtains texture features and area parameters of the complete iris region. The eyelid and periorbital skin segmentation subunit separates the upper and lower eyelids and the periorbital skin region by skin color model and edge contour analysis, and calculates the eyelid opening degree.

6. The image processing based vision change monitoring system of claim 5, wherein, The dynamic feature extraction unit is used for capturing the dynamic changes of the subject when staring at a visual target, and includes an eye movement tracking subunit, a pupil dynamic monitoring subunit and a blinking behavior analysis subunit. The eye movement tracking subunit calculates the eye rotation angle by using an optical flow method based on the change of the pupil center coordinates of the continuous frame images, generates an eye movement trajectory, and analyzes the delay time of the eye following the visual target. The pupil dynamic monitoring subunit records the pupil diameter change curve when the visual target is switched, calculates the pupil contraction rate and contraction amplitude, and reflects the eye adjustment function. The blinking behavior analysis subunit is used for counting the blinking frequency and blinking time length in the monitoring process, and excluding invalid monitoring data of frequent blinking and too long eye closure time. The static feature extraction unit is used for extracting eye static physiological parameters, and includes an iris texture extraction subunit, a corneal reflection point detection subunit and an eye axis length estimation subunit. The iris texture extraction subunit is used for subject identity recognition, and simultaneously analyzes the uniformity of the iris texture. The corneal reflection point detection subunit identifies the reflection point formed by the ambient light on the corneal surface, calculates the distance between the reflection point and the pupil center, and judges whether there is astigmatism in the cornea. The axial length estimation unit preliminarily estimates the axial length based on the iris diameter and the image pixel ratio, in combination with a corneal curvature estimation model, to provide a reference for myopia type judgment; The feature optimization unit is used to improve the effectiveness and reliability of the feature data, and includes an abnormal data elimination subunit, a feature fusion subunit, and a feature standardization subunit; The abnormal data elimination subunit is used to eliminate data of the eye rotation angle and the pupil contraction amplitude in the dynamic feature that are out of the normal physiological range, and to retain effective samples; The feature fusion subunit integrates the multi-dimensional features into an eye comprehensive feature vector through a weighted fusion algorithm, reducing the limitations of a single feature; The feature standardization subunit is used to make the feature data of different subjects comparable, facilitating subsequent matching with the vision-related feature library.

7. The image processing based vision change monitoring system according to claim 1, wherein, The vision assessment module includes a basic vision calculation unit, a corrected vision correction unit, a vision change analysis unit, a vision risk judgment unit, and a multi-dimensional evaluation report generation unit: The basic vision calculation unit calculates the unaided visual acuity based on the image processing results, including a visual target response analysis subunit, a vision conversion subunit, and a vision accuracy calibration subunit; The visual target response analysis subunit is used to analyze the fixation features of the subjects on visual targets of different sizes / directions, and to determine the largest recognizable visual target; The vision conversion subunit is used to calculate the initial unaided visual acuity value according to the visual target type selection corresponding conversion formula; The vision accuracy calibration subunit is used to introduce an eye dynamic feature correction coefficient to calibrate the initial visual acuity value, reducing the error range; The corrected vision correction unit is used to calculate the corrected visual acuity for subjects wearing glasses and contact lenses, including a correction parameter acquisition subunit, a parameter verification subunit, and a corrected vision calculation subunit; The correction parameter acquisition subunit supports two acquisition methods, namely manual input and image recognition; The parameter verification subunit is used to compare the acquired parameters with the common correction parameter library. If the deviation exceeds the normal range, the user is prompted to reconfirm or calibrate through professional equipment; The corrected vision calculation subunit uses an optical correction model to calculate the corrected visual acuity value, and labels the correction method and parameter reliability.

8. The image processing based vision change monitoring system of claim 7, wherein, The vision change analysis unit is used to trace historical data and analyze the vision change trend, including a data screening subunit, a change rate calculation subunit, a trend fitting subunit, and a change reason preliminary analysis subunit; The data screening subunit retrieves monitoring data from the data storage module for the past 1-3 years, screens data samples with consistent monitoring environment and stable subject state, and ensures analysis accuracy; The change rate calculation subunit is used to calculate the vision change rate in different time periods, distinguishing between short-term fluctuations and long-term trends; The trend fitting subunit uses a quadratic function fitting based on the screened sample data to determine the trend type; The change reason preliminary analysis subunit is used to analyze the reasons for the change in vision in combination with eye behavior data; The vision risk judgment unit judges the risk level based on the vision value and change trend, including a risk indicator definition subunit, a multi-dimensional scoring subunit, and a risk level division subunit; The risk indicator definition subunit is configured to define three types of core risk indicators, including current visual acuity level, visual acuity change range, and trend stability. The multi-dimensional scoring subunit is configured to score the three types of indicators respectively. The risk level division subunit is configured to divide the risk level according to the total score and mark the key risk points. The multi-dimensional evaluation report generation subunit is configured to integrate the evaluation results and generate a visual report, including a data visualization subunit, a suggestion generation subunit, and a report export subunit. The data visualization subunit uses a line chart to display the visual acuity change trend, a column chart to compare naked visual acuity and corrected visual acuity, and a radar chart to present multi-dimensional risk scores, making the results intuitive and easy to understand. The suggestion generation subunit generates personalized suggestions based on the risk level and change reasons. The report export subunit generates reports in different formats, including four modules: basic information, visual acuity data, risk assessment, and personalized suggestions, to facilitate users to retain or share with medical personnel.

9. The image processing based vision change monitoring system according to claim 1, wherein, The data storage and analysis module includes data classification storage unit, data security protection unit, data retrieval unit, deep data analysis unit and data sharing unit: The data classification storage unit is configured to store data by type and importance, including a basic data subunit, a feature data subunit, an evaluation result subunit, and a behavior correlation data subunit. The basic data subunit is configured to store the basic information of the testee, the monitoring device information and the environmental parameters. The feature data subunit is configured to store the eye feature data after image processing. The evaluation result subunit is configured to store the visual acuity evaluation report, support quick retrieval by time range and evaluation type, and update locally and in the cloud synchronously. The behavior correlation data subunit is configured to store the user-authorized eye behavior data and store it in association with the visual acuity data by timestamp, providing data support for deep analysis. The data security protection unit is configured to protect data privacy and integrity, including a data encryption subunit, an access control subunit, a data backup and recovery subunit, and an abnormal access monitoring subunit. The data encryption subunit uses double-layer encryption, including transmission encryption and storage encryption, and the key is set by the user and saved locally to prevent cloud data leakage. The access control subunit sets three levels of access permissions. The data backup and recovery subunit is configured to automatically backup local data daily to prevent data loss. The abnormal access monitoring subunit is configured to identify abnormal access behavior, trigger an alarm, and record abnormal logs for administrators to view.

10. The image processing based vision change monitoring system of claim 9, wherein, The data retrieval unit is configured to provide efficient data query functions, including a precise retrieval subunit, a fuzzy retrieval subunit, and a batch retrieval subunit. The precise retrieval subunit supports retrieval by unique identifier and provides complete data records. The fuzzy retrieval subunit supports keyword retrieval using full-text retrieval algorithms and marks matching keywords. The batch retrieval subunit supports batch data export by condition and supports setting export fields. The deep data analysis unit is configured to mine potential rules based on historical data, including a group comparison analysis subunit, an associated factor analysis subunit, and a prediction model subunit. The population comparison analysis subunit is used to compare the subject data with the population database of the same age group, generate a population ranking, and help the user understand their own vision level; The correlation factor analysis subunit is used to calculate the correlation coefficient between eye behavior data and vision changes, and identify key influencing factors; The prediction model subunit is used to train an LSTM neural network prediction model based on the vision data and behavior data of the past two years, and predict the vision change trend in the next 6-12 months; The data sharing unit supports compliant data sharing, including an authorization management subunit, a data desensitization subunit, and a sharing record subunit; The authorization management subunit is used for users to authorize the medical platform / school to obtain data for a specified period of time through a one-time authorization code, which automatically expires at the end of the period; The data desensitization subunit is used to automatically desensitize sensitive information when sharing data, and only retain necessary information such as vision data, evaluation results, and key environmental parameters to prevent privacy leaks; The sharing record subunit is used to record each data sharing behavior, allowing the user to view and revoke unexpired authorizations at any time, and ensuring data control.

Images