Functional vision testing method for self-monitoring of vision health

Abstract

The invention provides a functional vision testing method for visual health self-monitoring, which belongs to the technical field of intelligent vision testing and comprises the following steps: loading a multi-task self-adaptive model and parameters of a self-supervision pre-training editor to intelligent equipment of a vision tester, after loading is completed, test questions are presented according to requirements for visual acuity tests, contrast sensitivity tests and color vision tests; inputting the test content and the operation behavior of each test question into a self-supervised pre-training editor, and extracting fusion features; and based on all the fusion features under each test and in combination with the difficulty level of each test question based on the current distance between the display screen and the visual testee, obtaining a visual acuity threshold value, a contrast sensitivity curve and color vision arrangement error distribution of the visual testee, and generating a remote visual evaluation report. Self-monitoring is facilitated, the visual evaluation efficiency is improved, and remote and accurate evaluation of the illness state is facilitated.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent vision testing, and particularly relates to a functional vision testing method for self-monitoring of visual health. Background Art

[0002] Cataract is a disease that causes visual impairment due to lens opacity. With the acceleration of the social aging process, its efficient monitoring shows great significance in the field of elderly health management. Functional vision assessment, which reflects the visual state of patients in actual life, is particularly important in the clinical evaluation system of cataract severity. Existing functional vision assessment systems such as visual acuity, contrast sensitivity, moving visual acuity, and color vision can evaluate the visual function of the human eye from multiple dimensions such as form vision resolution, the ability to observe details of moving objects, and color recognition ability, so as to more comprehensively present the visual changes caused by cataracts. However, its assessment process is relatively complex and relies on professional examination equipment, which is not convenient for the elderly to self-monitor and reduces the assessment efficiency. Therefore, it is necessary to develop an inspection method based on standard quantitative indicators and combine it with portable electronic devices to achieve remote and accurate assessment of the condition.

[0003] Therefore, the present invention proposes a functional vision testing method for self-monitoring of visual health. Summary of the Invention

[0004] The present invention provides a functional vision testing method for self-monitoring of visual health to solve the above-mentioned technical problems.

[0005] The present invention provides a functional vision testing method for self-monitoring of visual health, including: Loading the parameters of the multi-task adaptive model and the self-supervised pre-training editor into the intelligent device of the vision tester, and presenting test questions according to requirements for visual acuity test, contrast sensitivity test, and color vision test respectively after the loading is completed; Based on the intelligent device, capturing the operation behaviors of the vision tester based on each test question; Inputting the test content and operation behaviors of each test question into the self-supervised pre-training editor to extract fusion features; Based on all the fusion features under each test and combining the difficulty level of each test question based on the current distance between the display screen and the vision tester, obtaining the visual acuity threshold, contrast sensitivity curve, and color vision arrangement error distribution of the vision tester, and generating a remote vision assessment report.

[0006] Preferably, it further includes: Collect the test results of each intelligent device after loading, and use the federated learning mechanism to locally train the collected data, and then update the parameters of the multi-task adaptive model and the self-supervised pre-training editor.

[0007] Preferably, for the visual acuity test, it includes: When the visual tester triggers the visual acuity test operation of the display screen, based on the display screen, output a presentation set at a difficulty level matching the current distance, where the presentation set includes several visual acuity test questions at the initial recognition difficulty level, and the visual acuity test questions are related to the size of the visual target and the movement speed of the visual target; Determine the feedback results of the visual tester for each visual acuity test question depending on the presentation set, where the feedback results are related to the operation behavior.

[0008] Preferably, after determining the feedback results of the visual tester for each visual acuity test question, it further includes: adjusting the visual target according to the feedback results and the actually changed distance, specifically including: Obtain the correct recognition rate according to the feedback results at the difficulty level; If the correct recognition rate is greater than or equal to the preset recognition rate, increase the difficulty level according to the feedback results; Among them, represents the increased difficulty level; represents the current difficulty level; represents the visual acuity difficulty adjustment step size; represents based on the feedback results Auxiliary function of the key information K and value information V inside the multi-task adaptive model; represents the preset recognition rate; represents the correct recognition rate; represents the ceiling symbol; Obtain a first new set matching the actually changed distance from the level-distance database, and conduct a visual acuity test at the next difficulty level for the visual tester, where the first new set includes several test questions related to the size and movement speed of the visual target at the updated level; If the correct recognition rate is less than the preset recognition rate, reduce the difficulty level according to the feedback results, and obtain a second new set matching the actually changed distance from the level-distance database to continue the visual acuity test.

[0009] Preferably, for the contrast sensitivity test, it includes: Set the initial contrast, and randomly present test questions consistent with the initial contrast on the display screen based on the initial contrast, and obtain the feedback vector of the visual tester for each contrast test question, where the feedback vector f = {Pj, tj}, Pj represents the judgment result of the j-th contrast test question; tj represents the answering duration of the j-th contrast test question; Input the feedback vector f into a multi-task adaptive model, fuse the bar grating information and user interaction information corresponding to the contrast test questions to form high-dimensional features; Based on all the high-dimensional features under the initial contrast, dynamically update the initial contrast and continue the test; Among them, represents all the high-dimensional features under the initial contrast , the auxiliary functions of the key information K and value information V inside the multi-task adaptive model; represents the contrast after dynamic update; represents the initial contrast; represents the contrast difficulty adjustment step size.

[0010] Preferably, for color vision testing, it includes: Capture the color block dragging behavior of the visual tester on the display screen for each color vision test question, where the color block dragging behavior includes: the entire dragging trajectory of each test color block and the individual dragging trajectories involved in the entire dragging trajectory, and the individual dragging trajectories are consistent with the number of dragging times; Analyze the dragging error probability of each test color block based on the color block dragging behavior, and generate a color gamut distribution in combination with the dragging position deviation of each test color block; Update the color vision test level according to the color gamut distribution and continue the test.

[0011] Preferably, analyzing the dragging error probability of each test color block based on the color block dragging behavior and generating a color gamut distribution in combination with the dragging position deviation of each test color block includes: Statistically analyze the time of each unit color block passing point in each entire dragging trajectory involved in each color vision test question to construct a trajectory vector , where, represents the position and residence time of the second unit color block passed in the entire dragging trajectory; represents the position and residence time of the m-th unit color block passed in the entire dragging trajectory; Screen the positions in the trajectory vector with a residence time greater than a preset time as the first positions, and split the entire drag trajectory based on the first positions; Determine the color difference between the color identification of the unit color block at the last position in each individual drag trajectory and the test identification, to obtain the color difference vector and the position difference vector corresponding to the entire drag trajectory; When the number of the individual drag trajectories is 1, keep the color difference vector and the position difference vector unchanged; When the number of the individual drag trajectories is multiple, determine the number of drags and the drag direction each time at each first position, and draw a drag vector diagram; Determine the drag invalidity between the last position in each individual drag trajectory and the first position of the next individual drag trajectory according to the drag vector diagram, and adjust the color difference and the position difference of the last position, and update the color difference vector and the position difference vector; Determine the number of test color blocks in each color gamut under each color vision test question, and construct a color difference matrix based on the final color difference vector to obtain the color difference feature vector corresponding to the color gamut. At the same time, construct a position difference matrix based on the final position difference vector to obtain the position difference feature vector corresponding to the color gamut; Obtain the drag error probability of each test color block according to the color difference vector, and combine the color difference feature vector and the position difference feature vector to obtain the error vertical value corresponding to the color gamut; Generate a color gamut distribution based on the error vertical values under different color gamuts.

[0012] Preferably, generate a remote vision assessment report, including: Based on all visual acuity test questions and the fusion features of each visual acuity test question at different difficulty levels of the visual acuity test, obtain the visual acuity threshold; Based on all contrast sensitivity test questions and the fusion features of each contrast sensitivity test question at different difficulty levels of the contrast sensitivity test, obtain the contrast sensitivity curve; Perform an average process on each color gamut error vertical value in the color gamut distribution of the color vision test questions involved at different difficulty levels of the color vision test to obtain the color vision arrangement error distribution; Generate a remote vision assessment report based on the visual acuity threshold, the contrast sensitivity curve, and the color vision arrangement error distribution.

[0013] Compared with the prior art, the beneficial effects of the present application are as follows: By loading a multi-task adaptive model and an editor into a smart device, visual monitoring is carried out from three dimensions: visual acuity, contrast sensitivity, and color vision test, which facilitates self-monitoring and improves the efficiency of visual evaluation, and is convenient for realizing remote and accurate assessment of the condition. At the same time, in the visual acuity test, the present invention provides a method for self-testing and evaluating the resolution of moving visual targets, which meets the self-monitoring needs of the elderly with driving needs in the process of social aging for their own visual function states related to movement in simulated special scenarios.

[0014] Other features and advantages of the present invention will be described in the following specification, and, in part, will become apparent from the specification or will be understood by implementing the present invention. The objectives and other advantages of the present invention can be achieved and obtained by the structures specifically pointed out in the written specification and the drawings.

[0015] The technical solution of the present invention will be further described in detail below with reference to the drawings and embodiments. Description of the Drawings

[0016] The drawings are used to provide a further understanding of the present invention, and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation to the present invention. In the drawings: Figure 1 It is a flowchart of a functional vision test method for visual health self-monitoring in an embodiment of the present invention; Figure 2 It is a structural diagram of the full-course dragging trajectory in an embodiment of the present invention; Figure 3 It is an implementation diagram of step 1 in an embodiment of the present invention; Figure 4 It is a specific flowchart of visual evaluation in an embodiment of the present invention. Detailed Embodiments

[0017] The following describes the preferred embodiments of the present invention with reference to the drawings. It should be understood that the preferred embodiments described herein are only used to illustrate and explain the present invention and are not used to limit the present invention.

[0018] The present invention provides a functional vision test method for visual health self-monitoring, as Figure 1 shown, including: Step 1: Load the parameters of the multi-task adaptive model and the self-supervised pre-training editor into the smart device of the visual tester, and after the loading is completed, present test questions according to requirements for visual acuity test, contrast sensitivity test, and color vision test respectively; Step 2: Based on the smart device, capture the operation behaviors of the visual tester based on each test question; Step 3: Input the test content and operation behavior of each test question into the self-supervised pre-trained editor to extract fused features; Step 4: Based on all the fused features under each test and combined with the difficulty level of each test question based on the current distance between the display screen and the visual tester, obtain the visual acuity threshold, contrast sensitivity curve, and color vision arrangement error distribution of the visual tester, and generate a remote visual assessment report.

[0019] In this embodiment, the intelligent device refers to a smartphone, a tablet computer, etc., which has a high-performance processor and a high-definition screen and can smoothly display various visual test questions and complex visual patterns, facilitating the tester to observe.

[0020] In this embodiment, the multi-task adaptive model can simultaneously process various different visual test tasks such as visual acuity test, contrast sensitivity test, and color vision test, and can automatically adjust the difficulty of subsequent test questions according to the actual performance of the tester during the test. For example, in the visual acuity test, if the tester continuously answers several difficult visual acuity chart questions correctly, the model will automatically select more challenging visual acuity chart questions, such as visual acuity chart symbols with smaller sizes and faster moving speeds; conversely, if the tester answers incorrectly multiple times, the model will reduce the question difficulty and provide larger, clearer, and slower moving visual acuity chart questions.

[0021] In this embodiment, the self-supervised pre-trained editor can perform pre-training through self-supervised learning to autonomously learn feature representations from data without manual annotation, so as to extract the fused features of the test content and operation behavior. Learn various features in the visual test data, such as extracting valuable features from information such as the shape and color of the visual acuity chart symbol images and the operation behavior data of the tester.

[0022] In this embodiment, the visual acuity test is used to accurately measure the ability of the visual system to distinguish details. There are "E" visual acuity chart symbols of different sizes and directions on the visual acuity chart. The tester shows their visual acuity level by identifying the direction of the visual acuity chart symbol at a certain distance from the display screen. The test system automatically calculates the size of the visual acuity chart symbol representing a certain visual angle at this test distance and presents it. When the tester can clearly distinguish the "E" visual acuity chart symbol representing a smaller visual angle, it indicates that their visual acuity is higher; conversely, if they can only see the visual acuity chart symbol with a larger visual angle, the visual acuity is lower.

[0023] In this embodiment, the contrast sensitivity test is used to evaluate the ability of the visual system to distinguish objects with different contrasts. First, a grating pattern with a relatively high contrast (e.g., 80%) is shown. At this time, the bright-dark difference between the bars in the pattern is obvious, and it is easier for the tester to see clearly. Then, the contrast is gradually reduced (e.g., reduced to 20%), and the bright-dark difference becomes smaller. The tester needs to concentrate more to distinguish the pattern. By observing how the tester distinguishes patterns with different contrasts, the contrast sensitivity level of the tester is evaluated.

[0024] In this embodiment, the color vision test is used to detect the ability of the visual system to distinguish colors.

[0025] In this embodiment, during the visual test, a series of actions are generated when the tester interacts with the intelligent device. Common operation behaviors include clicking. For example, when answering multiple-choice questions, the tester clicks on the option on the screen that they think is correct with their finger; swiping operation. When there is a large amount of content displayed on the test interface, the tester swipes on the screen with their finger to view different test questions; touch operation. In some tests that require adjusting the size or position of the visual target, the tester operates by touching the screen.

[0026] In this embodiment, the test content is the specific information contained in each test question. In the visual acuity test, the shape of the visual target (such as the letter "E", letters, numbers, etc.), size (dimensions corresponding to different viewing angles), and direction (up, down, left, right, etc.) are all test contents; in the contrast sensitivity test, the contrast values of the stripe patterns (such as 30%, 50%, etc.) and spatial frequency (the number of stripes within a unit viewing angle) belong to the test contents; in the color vision test, the color combinations of the color vision test chart (such as red-green combination, blue-yellow combination, etc.) and the arrangement of color blocks (the relative position relationship between different color blocks and the deviation degree from the standard arrangement) constitute the test contents.

[0027] In this embodiment, the fusion feature is a feature set formed by organically integrating various features of the test content and operation behavior. Taking the visual acuity test as an example, the fusion feature may include the size feature of the visual target (such as 1′ viewing angle, 5′ viewing angle, etc.), the reaction time feature of the tester clicking on the answer (the time from the display of the question to clicking on the answer, such as 2 seconds, 5 seconds, etc.), the accuracy feature of the clicking position (the deviation distance between the clicking position and the correct answer position), etc. These different types of features are fused together to form a comprehensive feature vector, which can more comprehensively reflect the performance and ability of the tester in this test question.

[0028] In this embodiment, the visual acuity threshold is a key indicator to measure the resolution ability of the visual system, which refers to the minimum visual angle at which a visual tester can accurately distinguish a visual target. For example, when a tester can clearly distinguish a visual target with a visual angle of 1′ during a visual acuity test, their visual acuity threshold is 1′; if they can only see a visual target with a visual angle of 2′, the visual acuity threshold is 2′. The smaller the visual acuity threshold, the higher the visual acuity of the tester.

[0029] In this embodiment, the contrast sensitivity curve presents the sensitivity relationship of the visual system to visual stimuli with different spatial frequencies and contrasts. It represents the spatial frequency on the abscissa (unit: cycles / degree, i.e., the logarithm of the number of light and dark stripes within each degree of visual angle), and the contrast sensitivity (the reciprocal of the lowest contrast that can be distinguished) on the ordinate. By plotting the contrast sensitivity points corresponding to different spatial frequencies and connecting these points to form a curve, it can intuitively display the contrast sensitivity function state of the visual system under different conditions. For example, a normal contrast sensitivity curve has higher sensitivity at medium and low spatial frequencies, and the sensitivity gradually decreases as the spatial frequency increases.

[0030] In this embodiment, the color vision arrangement error distribution is the distribution range of the hues of the color patches with incorrect arrangement orders when a tester arranges a series of color patches. Among them, the hue of each color patch to be arranged corresponds to a colored light with a continuously changing wavelength within the visible light spectrum range. After arranging all the color patches to be arranged in the correct order, a continuous and gradually changing color band corresponding to the visible light spectrum range will be obtained.

[0031] In this embodiment, the forms of the remote visual assessment report are diverse, including charts (such as bar charts of visual acuity thresholds, line charts of contrast sensitivity curves, radar charts of color vision error distributions, curve charts of color vision error distributions, etc.) and written explanations (interpretations of various assessment results, diagnostic suggestions, etc.). For example, the report will detail whether the tester's visual acuity threshold is within the normal range or abnormal, the differences between the contrast sensitivity curve and the normal curve, and whether the tester has color vision disorders and the specific color segments affected based on the color vision arrangement error distribution.

[0032] In this embodiment, the adaptive LogMAR visual acuity test algorithm is used to test visual acuity: an adaptive algorithm based on the LogMAR visual acuity chart. The system dynamically adjusts the size and presentation order of the visual acuity chart symbols according to the patient's real-time feedback. For example, if the patient makes a mistake at a larger symbol, the system will gradually reduce the symbol size until the smallest recognizable symbol is found. Through this dynamic adjustment mechanism, the personalization and accuracy of the test are ensured, and personalized visual acuity assessments can be provided according to the changes in the patient's visual function.

[0033] In this embodiment, an adaptive contrast sensitivity evaluation algorithm is adopted for the test comparison sensitivity: the system uses the Bayesian optimization algorithm for dynamic adjustment of the contrast sensitivity test. By analyzing the patient's feedback in real time, the system adjusts the contrast level of the test images to find the optimal parameter setting that best reflects the patient's contrast sensitivity, and can efficiently converge to the optimal solution within a small sampling range, ensuring accurate testing and rapid feedback.

[0034] In this embodiment, the Farnsworth-Munsell 100 Hue color chess is used for the test of color vision: combined with the convolutional neural network (CNN), it can accurately evaluate the patient's color vision resolution ability by analyzing the patient's sensitivity to hue changes under different lighting conditions. The adaptive mechanism automatically adjusts the hue difference according to the patient's initial performance to ensure that the test difficulty matches the patient's actual color vision level, helping to detect color vision disorders in cataract patients at an early stage.

[0035] The adaptive deep learning segmentation network is adopted: using the adaptive deep learning model, it can analyze and segment visual function data in real time and identify different visual function levels. The segmentation network provides personalized segmentation results by automatically adjusting the cataract lesion characteristics at different stages. For example, the system can automatically adjust the network weights and feature extraction methods according to different change stages of the patient's visual function, helping doctors identify the progression of the lesion and providing a reliable basis for personalized treatment plans.

[0036] In this embodiment, the implementation diagram for step 1 is as Figure 3 described.

[0037] In this embodiment, as Figure 4 shown, it is a specific visual assessment flowchart.

[0038] The beneficial effects of the above technical solutions are: by loading the multi-task adaptive model and the editor into the intelligent device, visual monitoring is carried out from three dimensions of visual acuity, contrast sensitivity, and color vision test, which is convenient for self-monitoring and improves the efficiency of visual evaluation, and is convenient for realizing remote and accurate assessment of the condition. At the same time, in the visual acuity test, the present invention provides a method for self-testing and evaluating the resolution of moving visual targets, which meets the self-monitoring needs of the elderly with driving needs in the process of social aging for their own visual function states related to movement in simulated special scenarios.

[0039] The present invention provides a functional vision test method for self-monitoring of visual health, and further includes: Collect the test results of each loaded intelligent device, and use the federated learning mechanism to perform local training on the collected data and then update the parameters of the multi-task adaptive model and the self-supervised pre-training editor.

[0040] In this embodiment, the test result is a data set generated after a visual tester completes visual acuity tests, contrast sensitivity tests, and color vision tests on an intelligent device.

[0041] In this embodiment, local training refers to the process in which an intelligent device uses the test result data collected by itself to train a multi-task adaptive model and a self-supervised pre-training editor locally. For example, based on the visual acuity, contrast sensitivity, and color vision test result data of 10 testers, the parameters of the model are adjusted to improve the prediction accuracy of the model when processing this data. This process is local training.

[0042] In this embodiment, on the intelligent device side, a lightweight machine learning framework (such as TensorFlowLite, PyTorchMobile) is used to load the parameters of the multi-task adaptive model and the self-supervised pre-training editor. According to the collected test result data, the model is trained according to a preset training algorithm (such as the stochastic gradient descent algorithm). During the training process, the update amount of the model parameters (such as the change value of the weight) is recorded. To improve the efficiency and effect of local training, a dynamic learning rate adjustment strategy is set on the intelligent device side. According to the computing resources and data volume of the device, the learning rate is automatically adjusted. When the computing resources of the device are sufficient and the data volume is large, the learning rate is appropriately increased to speed up the training; when the resources are limited or the data volume is small, the learning rate is decreased to ensure the stability of the training. At the same time, to avoid overfitting, a regularization method (such as L2 regularization) is used to constrain the model.

[0043] In this embodiment, after the server side receives the update amounts of the model parameters uploaded by each intelligent device, an aggregation algorithm (such as the FedAvg algorithm) is used to summarize and calculate these update amounts. According to the summary result, the global parameters of the multi-task adaptive model and the self-supervised pre-training editor are updated. After the update is completed, the new parameters are sent to each intelligent device through the network for the next test.

[0044] The beneficial effects of the above technical solution are: by collecting test results for local training, the accuracy of the test results in the subsequent three dimensions is further guaranteed.

[0045] The present invention provides a functional vision test method for visual health self-monitoring. For the visual acuity test, it includes: When the visual tester triggers the visual acuity test operation on the display screen, a presentation set at a difficulty level matching the current distance is output based on the display screen, where the presentation set includes several visual acuity test questions at the initial recognition difficulty level, and the visual acuity test questions are related to the size of the visual target and the movement speed of the visual target; Determine the feedback results of the visual tester for each visual acuity test question based on the presentation set, where the feedback results are related to the operation behavior.

[0046] In this embodiment, the visual tester is a person who participates in the visual acuity test and receives visual evaluation.

[0047] In this embodiment, the current distance refers to the actual distance between the display screen and the eyes of the visual tester. In actual testing, this distance can be measured by the distance sensor (such as a ToF sensor) of the intelligent device. For example, when the intelligent device is placed on the desktop and the tester is sitting in a normal posture, the distance between the device and the eyes is about 40 cm; if the tester holds the device for testing, the distance may be about 30 cm.

[0048] In this embodiment, the difficulty level is different test difficulty levels set according to the visual acuity test requirements. Generally, the higher the difficulty level, the smaller the visual target, the faster the movement speed, and the higher the requirement for the tester's vision. For example, the visual target of difficulty level 1 is larger and stationary, suitable for people with poor eyesight or those taking the test for the first time; the visual target of difficulty level 5 is extremely small and moves on the screen at a relatively fast speed, used to detect people with extremely good eyesight or those who need high-precision vision evaluation.

[0049] In this embodiment, the presentation set is a set of visual acuity test questions presented to the visual tester at a specific difficulty level. For example, in the presentation set of difficulty level 3, it may include 5 "E" character visual target questions with different directions, moderate sizes, and moving slowly at a certain speed, as well as 3 visual target questions of other shapes (such as letters, numbers).

[0050] In this embodiment, the visual target size is the size of the visual target in the visual acuity test, usually measured in visual angle (′ visual angle). For example, in the standard logarithmic visual acuity chart, the visual target size corresponding to a visual acuity of 1.0 is about 1′ visual angle; the visual target corresponding to a visual acuity of 0.1 is larger, about 10′ visual angle.

[0051] In this embodiment, the movement speed is the speed at which the visual target moves on the display screen, with the unit of degrees per second, that is, the speed at which the visual target moves relative to the tester (the center of the circle) on a circle with the tester as the center and the test distance as the radius. The movement speed of the commonly used low-speed moving visual target can be set to 10 degrees per second, the movement speed of the medium-speed moving visual target can be set to 30 degrees per second, the movement speed of the medium-high-speed moving visual target can be set to 60 degrees per second, and the movement speed of the high-speed moving visual target can be set to 90 degrees per second.

[0052] In this embodiment, in the vision test APP of the intelligent device, a distance detection and difficulty matching algorithm is written. After detecting that the tester triggers the visual acuity test operation, the current distance between the display screen and the tester's eyes is first obtained through a distance sensor. If the device is not equipped with a distance sensor, a prompt box will pop up, asking the tester to manually input the distance. According to the preset distance-difficulty mapping rule, the difficulty level matching the current distance is determined. For example, when the distance is between 30 and 40 cm, the corresponding difficulty level is 3; when the distance is greater than 40 cm, the difficulty level is reduced by one level. Then, several visual acuity test questions at this difficulty level are retrieved from the question bank to form a presentation set, which is displayed on the display screen in a certain order. During the display process, the size and movement speed of the visual target are controlled through graphics rendering technology. For example, the "E" visual target is rendered, and its size and movement trajectory are set according to the difficulty level to ensure that the visual target moves smoothly on the screen at a set speed. In terms of difficulty matching, the proportion of correctly matching the difficulty level according to the distance is 95%. Regarding the display effect of the presentation set at different difficulty levels, 90% of the testers feedback that the size and movement speed settings of the visual targets are reasonable and easy to observe and identify; 10% of the testers think that the movement speed of the visual targets at some high difficulty levels is too fast to distinguish.

[0053] In this embodiment, the feedback result is the response information made by the vision tester for each visual acuity test question, which is closely related to the tester's operation behavior. For example, when the tester clicks on the screen to select the direction of the visual target (such as up, down, left, right), it is the feedback result.

[0054] In this embodiment, the operation behavior is the action of the tester interacting with the display screen when answering the visual acuity test questions, including touching the screen, swiping the screen, etc. For example, when the tester slides their finger on the screen to adjust the clarity of the visual target, or clicks the confirmation button on the screen to submit the answer, it belongs to the operation behavior.

[0055] The beneficial effects of the above technical solution are as follows: Outputting a presentation set by matching the difficulty level based on the current distance can provide visual acuity test questions with appropriate difficulty according to the actual test environment of the tester, avoiding the questions being too difficult or too easy, and improving the accuracy and effectiveness of the test. At the same time, the dynamic settings of the size and movement speed of the visual targets simulate a more realistic visual scene, making the test results more able to reflect the tester's vision in real life. Relying on the presentation set to accurately determine the tester's feedback results provides a reliable data basis for subsequent visual acuity threshold calculation, vision assessment, etc. By recording the operation behavior and feedback results, the vision characteristics and answering habits of the tester can be further analyzed, which helps to optimize the visual acuity test questions and processes, and improve the accuracy and user experience of the test.

[0056] The present invention provides a functional vision test method for visual health self-monitoring. After determining the feedback results of the visual tester for each visual acuity test question, it further includes: adjusting the visual target according to the feedback results and the actually changed distance, specifically including: Obtaining the correct recognition rate according to the feedback results at the difficulty level; If the correct recognition rate is greater than or equal to the preset recognition rate, increasing the difficulty level according to the feedback results; wherein, represents the increased difficulty level; represents the current difficulty level; represents the visual acuity difficulty adjustment step size; represents based on the feedback results and the auxiliary function of the key information K and value information V inside the multi-task adaptive model; represents the preset recognition rate; represents the correct recognition rate; represents the ceiling symbol; Obtaining the first new set matching the actually changed distance from the level-distance database, and conducting a visual acuity test at the next difficulty level for the visual tester, where the first new set includes several visual target sizes and test questions related to the movement speed at the updated level; If the correct recognition rate is less than the preset recognition rate, reducing the difficulty level according to the feedback results, and obtaining the second new set matching the actually changed distance from the level-distance database to continue the visual acuity test.

[0057] In this embodiment, the difficulty level is different test difficulty levels preset in the visual acuity test.

[0058] In this embodiment, the correct recognition rate is the proportion of the number of visual acuity test questions correctly answered by the tester at a certain specific difficulty level to the total number of questions at that level. A calculation function is written in the Python language. By traversing the database table storing the feedback results, screening out the records of the correctly answered questions, counting their number, and then calculating with the total number of questions at that level, the correct recognition rate is obtained, and the value of the preset recognition rate is 90%.

[0059] In this embodiment, the difficulty of the visual acuity test is increased to a higher level, making the subsequent visual acuity test questions more difficult, such as reducing the size of the visual target, increasing the movement speed of the visual target, and increasing the complexity of the questions. For example, when the difficulty level is increased from level 3 to level 4, the visual target will become smaller and the movement speed will also increase, further increasing the requirements for the tester's eyesight. Conditional judgment code is written in the Java language. When the correct recognition rate meets the conditions, the variable value representing the difficulty level is incremented by 1, and relevant functions are called to reset the visual target parameters.

[0060] In this embodiment, the level-distance database is used to store the corresponding relationship between different difficulty levels and the actual distance between the display screen and the tester's eyes, as well as the database of the corresponding test question sets for each difficulty level. The database records the difficulty levels corresponding to various distance ranges, as well as the information of the test questions related to the size and movement speed of the visual target included in each difficulty level. For example, a distance of 30 - 40 cm corresponds to difficulty level 3, and this level includes multiple visual target test questions with specific sizes and movement speeds. In this embodiment, the actually changed distance is the distance that actually changes between the display screen and the tester's eyes during the test. For example, during the test, the tester picks up the tablet computer originally placed on the table and brings it closer to the eyes, and at this time the distance changes from 40 cm to 30 cm.

[0061] In this embodiment, the first new set is a set of new visual acuity test questions obtained from the level-distance database according to the actually changed distance and the increased difficulty level. This set contains several test questions related to the size and movement speed of the visual target at the updated level, and is used to conduct a re-visual acuity test on the tester. For example, when the difficulty level is increased from 3 to 4 and the actual distance becomes 30 cm, the set of visual acuity test questions suitable for difficulty level 4 and matching the 30 cm distance obtained from the database is the first new set. The SQLite database is used to store the level-distance data, and the database query operation is performed through the sqlite3 library of Python to obtain the corresponding test question set. It should be noted that the principle of the second new set is similar to that of the first new set, and will not be elaborated here.

[0062] In this embodiment, when py ≤ pz < 1, through more refined calculations, considering the gap between the correct recognition rate and the preset recognition rate, the difficulty is adjusted proportionally, taking into account the situation where the tester has a certain correct recognition rate but does not reach perfection, making the difficulty adjustment fit the actual level of the tester, and avoiding excessive or insufficient adjustment. The value of py is 0.9.

[0063] The auxiliary function A(Q, K, V) synthesizes the feedback result Q, the key information K inside the multi-task adaptive model, and the value information V. It does not solely rely on the correct recognition rate but incorporates other important information during the testing process into the difficulty adjustment calculation, comprehensively considering the test-taker's answering situation and the internal state of the model, making the difficulty adjustment more scientific and reasonable, and has a value range of (0, 1.5), and , where , , are weights with values of 0.4, 0.3, and 0.3, , , are respectively the mapping value functions based on the comprehensive feedback result Q, the key information K inside the multi-task adaptive model, and the value information V. And the comprehensive feedback result Q, the key information K inside the multi-task adaptive model, and the value information V can be obtained by matching from the feedback-value comparison table, the key information-value comparison table, and the value-value comparison table in sequence. The feedback-value comparison table is a reference table for converting the test-taker's feedback result Q into a numerical mapping value f(Q). The feedback result Q contains various information during the visual acuity test of the test-taker, such as the number of questions answered correctly or incorrectly, the response time of answering questions, the operation behaviors of answering questions (clicking, voice input, etc.). The key information-value comparison table is a table for converting the key information K inside the multi-task adaptive model into a numerical mapping value f(K). The key information K comes from inside the model, such as the neuron activation values of specific layers of the model, the parameter correlation degree between different task branches, the weight distribution when the model processes different visual features (such as the shape and color of the visual target), etc. For example, the value obtained from this table is 0.3. The value-value comparison table is a reference table for converting the value information V into a numerical mapping value f(V). The value information V is related to the model's emphasis on different test tasks and the importance evaluation of different visual features. For example, in some cases, the recognition of the shape of the visual target in the visual acuity test may be more important than the color, or the model pays more attention to the contrast sensitivity test task in a specific test scenario. For example, the model focuses more on the visual acuity test task, and the size information of the visual target is more critical than the direction information in the visual acuity test. At this time, the matched value is 0.6.

[0064] In this embodiment, the value of γ is 1.

[0065] The beneficial effects of the above technical solution are as follows: When the correct recognition rate reaches or exceeds the preset recognition rate, the difficulty level is increased, which can timely adapt to the better vision performance of the tester, further explore the visual ability of the tester, and make the test results more discriminative and accurate. Obtain the first new set that matches the actually changed distance and the increased difficulty level from the level-distance database for retesting, fully considering the changes in the test environment and the tester's ability. When the correct recognition rate is less than the preset recognition rate, the difficulty level is decreased, and the second new set is obtained for continuous testing, which can timely adjust the test difficulty and adapt to the poorer vision performance of the tester.

[0066] The present invention provides a functional vision test method for visual health self-monitoring. For the contrast sensitivity test, it includes: Set an initial contrast, and randomly present test questions consistent with the initial contrast on the display screen based on the initial contrast, and obtain the feedback vector of the visual tester for each contrast test question. Among them, the feedback vector f = {Pj, tj}, Pj represents the judgment result of the jth contrast test question; tj represents the answering duration of the jth contrast test question. Input the feedback vector f into the multi-task adaptive model, fuse the bar grating information and user interaction information corresponding to the contrast test question to form high-dimensional features. According to all the high-dimensional features under the initial contrast, dynamically update the initial contrast and continue the test. Among them, represents the auxiliary function of all high-dimensional features U, key information K, and value information V inside the multi-task adaptive model under the initial contrast; represents the contrast after dynamic update; represents the initial contrast; represents the contrast difficulty adjustment step size.

[0067] In this embodiment, the initial contrast is the reference contrast value set at the beginning of the contrast sensitivity test, usually determined based on the population average level or the preliminary evaluation result of the tester. For example, in a medical vision test, the initial contrast may be set at a medium level (such as 50%), that is, the brightness difference between the test pattern and the background is 50%.

[0068] In this embodiment, the visual test content generated based on the initial contrast is usually bar grating patterns (such as sine gratings) with different spatial frequencies. For example, a test question may be a vertical bar grating pattern with a contrast of 50% and a spatial frequency of 3 cycles / degree. The bar grating direction (horizontal / vertical / diagonal) and spatial frequency (such as 1, 3, 6, 12 cycles / degree) are randomly selected by the test software to ensure the randomness and objectivity of the test.

[0069] In this embodiment, for example, the average response time: 3.2 seconds per question (standard deviation 1.1 seconds), the correct rate for low spatial frequency (1 - 3 cycles / degree): 92%, the correct rate for high spatial frequency (12 cycles / degree): 65%, and there is a positive correlation between the response time and the spatial frequency (r = 0.62). In this embodiment, the physical properties of the bar grating pattern in the test questions include contrast, spatial frequency, orientation, etc. For example, a bar grating information vector can be expressed as [contrast = 40%, spatial frequency = 6 cycles / degree, orientation = vertical].

[0070] In this embodiment, the user interaction information is the interaction behavior data between the tester and the system, such as response time, mouse trajectory, click position, etc. For example, the hesitation time of the user when judging the visibility of the bar grating may reflect the uncertainty of their visual perception.

[0071] In this embodiment, the high - dimensional feature is an abstract feature representation formed by fusing the bar grating information and the user interaction information. For example, through a neural network, the original data is mapped to a 128 - dimensional feature space, and each dimension represents different visual or behavioral features. The high - dimensional feature space captures richer visual perception features of the tester, providing a basis for personalized testing.

[0072] In this embodiment, the auxiliary function B(U, K, V) comprehensively considers various factors and does not simply rely on the judgment result of the tester or a single piece of information. The high - dimensional feature U fuses the bar grating information and the user interaction information, and can comprehensively reflect the tester's visual perception and operation behavior of the current contrast; the key information K and the value information V reflect the internal operation state of the model and the degree of emphasis on different factors. Considering multiple factors comprehensively makes the contrast adjustment more comprehensive, scientific, and in line with the actual test requirements, and takes the value of 1.

[0073] , where 、 、 are weights, taking the values of 0.4, 0.3, 0.3, is the mapping value function based on the high - dimensional feature U, and the high - dimensional feature U is obtained by matching from the feature - value look - up table. The feature - value look - up table is a reference table for converting the high - dimensional feature U into a numerical mapping value f(U).

[0074] The beneficial effects of the above - mentioned technical solution are: By randomizing the test questions, the memory effect and expected deviation of the tester are reduced, and the dynamic update mechanism significantly improves the accuracy of threshold estimation and the test efficiency.

[0075] The present invention provides a functional vision test method for visual health self - monitoring. For color vision testing, it includes: Capture the dragging behavior of the visual tester on the display screen for each color vision test question block, where the dragging behavior of the question block includes: the full dragging trajectory of each test question block and the individual dragging trajectories involved in the full dragging trajectory, and the individual dragging trajectories are consistent with the number of dragging times; Analyze the dragging error probability of each test question block based on the dragging behavior of the question block, and generate a color gamut distribution in combination with the dragging position deviation of each test question block; Update the color vision test level according to the color gamut distribution and continue the test.

[0076] In this embodiment, the dragging behavior of the question block is the behavior of the tester dragging the question block in the color vision test on the display screen. The full dragging trajectory is the complete path that the question block moves on the display screen when the tester drags a certain question block from the start to the end. For example, when the tester drags a red question block from the left side of the screen to the right side, the moving path from the starting point on the left to the ending point on the right is the full dragging trajectory. The individual dragging trajectory is the trajectory formed by each independent dragging action in the full dragging trajectory. For example, when the tester drags the red question block from the left to the right in three drags, the trajectory formed by each drag is the individual dragging trajectory, and the number of individual dragging trajectories is consistent with the number of dragging times. Taking a web version of the color vision test program developed based on HTML5 and JavaScript as an example, by listening to the mousedown (mouse press), mousemove (mouse move) and mouseup (mouse release) events, record the coordinate changes of the question block on the display screen, so as to obtain the full dragging trajectory and the individual dragging trajectory. For touch screen devices, similar functions can be achieved through the touchstart, touchmove and touchend events.

[0077] In this embodiment, the dragging position deviation is the difference between the position where the tester finally drags the question block and the correct position.

[0078] In this embodiment, a mapping relationship is established between the color gamut distribution and the color vision test level. Rules for adjusting the color vision test level corresponding to different color gamut distribution characteristics are preset in advance. For example, if the color gamut distribution shows that the tester makes more mistakes in distinguishing low-difficulty colors, then lower the color vision test level; if the tester performs better in distinguishing high-difficulty colors, then raise the color vision test level. Automatically judge and update the color vision test level according to the generated color gamut distribution through programming, and then retrieve the color vision test questions corresponding to the level from the test question library to continue the test on the tester.

[0079] The beneficial effects of the above technical solution are as follows: accurately capture the behavior of dragging color blocks, provide a rich and accurate data basis for subsequent analysis of the tester's color vision, generate a color gamut distribution by analyzing the dragging error probability and position deviation, and can visually present the tester's color vision in a visual way. Dynamically update the color vision test level according to the color gamut distribution, which can better match the test difficulty with the tester's color vision ability.

[0080] The present invention provides a functional vision test method for visual health self-monitoring, which analyzes the dragging error probability of each test color block based on the dragging behavior of the color block, and combines the dragging position deviation of each test color block to generate a color gamut distribution, including: Statistically analyze the time of each unit color block passing point in each full dragging trajectory involved in each color vision test question to construct a trajectory vector , where represents the position of passing through the second unit color block in the full dragging trajectory and the residence time ; represents the position of passing through the m-th unit color block in the full dragging trajectory and the residence time ; Screen the positions with residence time greater than the preset time from the trajectory vector, regard them as the first positions, and split the full dragging trajectory according to the first positions; Determine the color difference between the color identification of the unit color block at the last position in each individual dragging trajectory and the color of the test identification, and obtain the color difference vector and position difference vector corresponding to the full dragging trajectory; When the number of the individual dragging trajectories is 1, keep the color difference vector and the position difference vector unchanged; When the number of the individual dragging trajectories is multiple, determine the dragging times and the dragging directions each time at each first position, and draw a dragging vector diagram; Determine the dragging invalidity between the last position in each individual dragging trajectory and the first position of the next individual dragging trajectory according to the dragging vector diagram, and adjust the color difference and position difference at the last position to update the color difference vector and the position difference vector; Determine the number of test color blocks in each color gamut under each color vision test question, and construct a color difference matrix based on the final color difference vector to obtain the color difference feature vector corresponding to the color gamut. At the same time, construct a position difference matrix based on the final position difference vector to obtain the position difference feature vector corresponding to the color gamut; The dragging error probability of each test color patch is obtained according to the final color difference vector, and by combining the color difference feature vector and the position difference feature vector, the error vertical value corresponding to the color gamut is obtained. Based on the error vertical values under different color gamuts, a color gamut distribution is generated.

[0081] In this embodiment, the color vision test questions are questions designed to detect the color vision ability of the test taker, usually including combinations of color patches of different colors. For example, moving the green color patch A1 from position 1 to position 2, by overriding the onTouchEvent method of the View, the coordinate values of each touch operation are obtained, and these coordinate values are recorded in chronological order to form the dragging trajectory of the color patch. At the same time, a counter is set to count the number of drags, so as to determine the number of individual dragging trajectories.

[0082] In this embodiment, the dragging position deviation is the gap between the actual placement position and the ideal accurate position when the test taker drags the test color patch to the target area. For example, ideally, the green color patch should be placed at the center position of the green area, but the test taker places it 2 centimeters away from the center, and this 2 centimeters is the dragging position deviation.

[0083] In this embodiment, the color gamut distribution is a distribution of the test taker's perception and operation of different colors, comprehensively depicted according to the dragging error probability and position deviation of the test taker for different color test color patches. It can be displayed in the form of graphs (such as heat maps, radar charts). In Matplotlib, with the color category as the abscissa and the comprehensive score of the error probability and position deviation as the ordinate, a bar chart, radar chart or heat map is drawn to display the color gamut distribution. For example, the average dragging error probability of normal color vision test takers for blue and yellow color patches is 5%, and the average position deviation is 0.5 centimeter; among color vision abnormal test takers, the average dragging error probability of red-green color blind patients for red and green color patches is 40%, and the average position deviation is 1.5 centimeters. Through the generated color gamut distribution graph, it can be intuitively seen that red-green color blind patients show obvious characteristics of high error probability and large position deviation in the red and green areas.

[0084] In this embodiment, when the color gamut distribution shows that the dragging error probabilities of the test taker for multiple colors are all low and the position deviations are small, it is judged that the color vision ability is good, and the test level can be increased to retrieve more difficult color vision test questions from the question bank, such as questions with combinations of color patches with similar colors; when the color gamut distribution shows that the test taker has a high dragging error probability and a large position deviation, especially in some key color areas (such as the red and green areas for red-green color blindness detection), it is judged that the color vision ability is poor, and the test level is reduced to provide simpler questions with large color differences in color patch combinations.

[0085] In this embodiment, the preset time is 1 second.

[0086] In this embodiment, as Figure 2 shown, there are color patches 1, 2, 3, 4, 5, and 6. At this time, it is necessary to move from color patch 1 to color patch 6 to pass the test. At this time, color patches 2 and 5 are the corresponding first positions. At this time, the individual drag trajectories are: color patch 1--color patch 2, color patch 3--color patch 4--color patch 5, color patch 6.

[0087] In this embodiment, the number of individual drag trajectories being 1 indicates that the entire drag trajectory is not split.

[0088] At this time: Color difference vector = {color difference between color patch 2 and color patch 6, color difference between color patch 5 and color patch 6, color difference between color patch 6 and color patch 6}.

[0089] Position difference vector = {position difference between color patch 2 and color patch 6, position difference between color patch 5 and color patch 6, position difference between color patch 6 and color patch 6}, where the values corresponding to the position difference between color patch 6 and color patch 6 and the color difference between color patch 6 and color patch 6 are 0.

[0090] In this embodiment, the drag vector diagram is a diagram constructed from the drag direction and number of times at each first position. It is to avoid incorrect difference analysis caused by time statistics of passing points due to screen lag resulting in ineffective dragging. Among them, the direction and length of each drag are the corresponding drag vectors for that time, and all drag vectors form the drag vector diagram.

[0091] In this embodiment, the drag vector diagram is a diagram of the remaining trajectories except for the last individual drag trajectory.

[0092] In this embodiment, drag ineffectiveness: If: the sum of the lengths of each drag direction / (length of a unit color patch × number of drags in the drag vector diagram at the last position in the individual drag trajectory) is greater than or equal to 1, at this time, the drag ineffectiveness is 1; Otherwise, the drag ineffectiveness is: the sum of the lengths of each drag direction / (length of a unit color patch × number of drags in the drag vector diagram at the last position in the individual drag trajectory).

[0093] In this embodiment, if the drag ineffectiveness is 1, at this time, reduce the color difference corresponding to the last position in the corresponding individual drag trajectory by half and reduce the position difference by half.

[0094] In this embodiment, if the drag ineffectiveness is not 1, at this time, the updated color difference = original color difference × drag ineffectiveness × (1 / 2); the updated position difference = original position difference × drag ineffectiveness × (1 / 2).

[0095] In this embodiment, the color gamut can be green, blue, and red.

[0096] In this embodiment, the test color block refers to moving the specified color block to the position where it needs to be placed, that is, the position of the final test color block.

[0097] In this embodiment, , position .

[0098] In this embodiment, the eigenvector is obtained by solving the eigenvalues of the matrix, which belongs to the prior art.

[0099] In this embodiment, the drag error probability is the ratio of the number of non-zero color differences in the final color difference vector (i.e., the number of incorrect operations) to (the total number of elements in the final color difference vector + 1). It should be noted that the matrix is an alignment matrix, and the remaining vectors are filled based on the number of elements in the vector with the most elements. For example, if there are 3 elements and vector U1 is {0.1, 0}, after filling, it becomes: {0.1, 0, 0}; if vector U1 is {0.1, 0.2}, after filling, it becomes: {0.1, 0.2, 0.2}.

[0100] In this embodiment, the error vertical value of the color gamut = , where respectively represent the average values of the features of the corresponding color difference feature vector and the position difference feature vector, represents the average value of the sum of the drag error probabilities of each test color block obtained from the final color difference vector, , represents mapped to between 0 and 1 respectively, and then the two are added together.

[0101] In this embodiment, for example, the error vertical value of the green color gamut is 2, the error vertical value of the red color gamut is 2.3, and the error vertical value of the blue color gamut is 3.1. At this time, a bar chart can be drawn to obtain the color gamut distribution.

[0102] The beneficial effects of the above technical solutions are as follows: By capturing the color block dragging behavior, detailed operation information of the tester during the color vision test can be obtained. By analyzing the drag error probability and position deviation to generate the color gamut distribution, the color vision situation of the tester can be presented in a visual way. Updating the color vision test level according to the color gamut distribution and continuing the test can make the test difficulty dynamically adapt to the color vision ability of the tester.

[0103] The present invention provides a functional vision test method for visual health self-monitoring, generating a remote visual assessment report, including: Based on all the visual acuity test questions at different difficulty levels in the visual acuity test and the fusion features of each visual acuity test question, a visual acuity threshold is obtained; Based on all the contrast sensitivity test questions at different difficulty levels in the contrast sensitivity test and the fusion features of each contrast sensitivity test question, a contrast sensitivity curve is obtained; For each vertical error value of the color gamut distribution of the color vision test questions involved at different difficulty levels in the color vision test, an average process is performed to obtain a color vision arrangement error distribution; Based on the visual acuity threshold, contrast sensitivity curve, and color vision arrangement error distribution, a remote visual assessment report is generated.

[0104] In this embodiment, the fusion feature is a feature set formed by integrating various features of the test content (such as visual target features) and the tester's operation behaviors (such as answering time, click position, etc.). For example, the fusion feature of a visual acuity test question may include information such as the visual target size (5′ visual angle), color (black), the reaction time (2 seconds) of the tester to click the correct answer, and the deviation (0.3 cm) between the click position and the correct answer position.

[0105] In this embodiment, the color vision arrangement error distribution is the result obtained by averaging the vertical error values of each color gamut in the color gamut distribution, which is used to overall reflect the error distribution of the tester in the color vision test.

[0106] In this embodiment, in the contrast sensitivity test software, according to the preset different difficulty levels, the test questions containing different combinations of contrast and spatial frequency are sequentially displayed. Using the Matplotlib library of Python, according to the calculated spatial frequency and contrast data, a contrast sensitivity curve is plotted.

[0107] First, data such as the visual acuity threshold, contrast sensitivity curve, and color vision arrangement error distribution are integrated. Then, using a report generation template (which can be a template based on formats such as Word, PDF, etc.), the data is filled into the template in the form of charts (such as a bar chart of the visual acuity threshold, a line chart of the contrast sensitivity curve, a radar chart of the color vision arrangement error distribution, etc.) and text descriptions. For example, in the text description part, the results of each test are interpreted, the visual function status of the tester is analyzed, evaluations such as whether there are vision problems or color vision abnormalities are given for reference, and suggestions for further examinations or corrections are provided according to the situation. Finally, the generated report is sent to the tester or relevant professionals (such as doctors, optometrists, etc.) through the network (such as email, online platform, etc.).

[0108] The beneficial effects of the above technical solution are as follows: By obtaining the above visual test results, it is possible to systematically understand the error situations and characteristics of the testee in the aforementioned visual test, generate a remote visual assessment report, and present complex visual test data to relevant personnel in a concise and intuitive form. This not only facilitates the testee to understand their own visual health status, but also provides a preliminary basis for professionals to quickly evaluate.

[0109] Obviously, those skilled in the art can make various modifications and variations to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalent technologies, the present invention is also intended to include these modifications and variations.

Claims

1. A functional vision test method for visual health self-monitoring, characterized in that, Including: Loading the parameters of the multi-task adaptive model and the self-supervised pre-training editor into the smart device of the visual tester, and presenting test questions according to requirements for visual acuity test, contrast sensitivity test, and color vision test respectively after the loading is completed; Capturing the operation behaviors of the visual tester based on each test question by the smart device; Inputting the test content and operation behaviors of each test question into the self-supervised pre-training editor to extract fused features; Based on all the fused features under each type of test and combined with the difficulty level of each test question based on the current distance between the display screen and the visual tester, obtaining the visual acuity threshold, contrast sensitivity curve, and color vision arrangement error distribution of the visual tester, and generating a remote visual assessment report.

2. The functional vision test method for visual health self-monitoring according to claim 1, characterized in that, Also including: Collecting the test results of each loaded smart device, and using the federated learning mechanism to locally train the collected data and then update the parameters of the multi-task adaptive model and the self-supervised pre-training editor.

3. The functional vision test method for visual health self-monitoring according to claim 1, characterized in that, For the visual acuity test, including: When the visual tester triggers the visual acuity test operation on the display screen, outputting a presentation set at a difficulty level matching the current distance based on the display screen, where the presentation set includes several visual acuity test questions at the initial recognition difficulty level, and the visual acuity test questions are related to the size of the visual target and the movement speed of the visual target; Determining the feedback results of the visual tester for each visual acuity test question depending on the presentation set, where the feedback results are related to the operation behaviors.

4. The functional vision test method for visual health self-monitoring according to claim 3, wherein After determining the feedback results of the visual tester for each visual acuity test question, it further includes: adjusting the visual target according to the feedback results and the actually changed distance, specifically including: Obtaining the correct recognition rate according to the feedback results at the difficulty level; If the correct recognition rate is greater than or equal to the preset recognition rate, increasing the difficulty level according to the feedback results; ； Among them, represents the increased difficulty level; represents the current difficulty level; represents the visual acuity difficulty adjustment step size; represents based on the feedback result , the auxiliary function of the key information K and value information V inside the multi-task adaptive model; represents the preset recognition rate; represents the correct recognition rate; represents the ceiling symbol; Obtaining a first new set matching the actually changed distance from the level-distance database, and conducting a visual acuity test at the next difficulty level for the visual tester, where the first new set includes several test questions related to the size of the visual target and the movement speed of the visual target at the updated level; If the correct recognition rate is less than the preset recognition rate, decreasing the difficulty level according to the feedback results, and obtaining a second new set matching the actually changed distance from the level-distance database to continue the visual acuity test.

5. The functional vision test method for visual health self-monitoring according to claim 1, characterized in that, For the contrast sensitivity test, including: Setting an initial contrast, and randomly presenting test questions with the same initial contrast on the display screen based on the initial contrast, and obtaining the feedback vector of the visual tester for each contrast test question, where the feedback vector f = {Pj, tj}, Pj represents the judgment result of the jth contrast test question; tj represents the answering duration of the jth contrast test question; Inputting the feedback vector f into the multi-task adaptive model, fusing the bar grating information and user interaction information corresponding to the contrast test question to form a high-dimensional feature; Dynamically update the initial contrast based on all the high-dimensional features at the initial contrast and continue the test; ； Among them, represents all high-dimensional features under the initial contrast , the auxiliary functions of the key information K and value information V inside the multi-task adaptive model; represents the contrast after dynamic update; represents the initial contrast; represents the contrast difficulty adjustment step size.

6. The functional vision test method for visual health self-monitoring according to claim 1, characterized in that, For color vision testing, it includes: Capture the behavior of the visual tester dragging color patches for each color vision test question on the display screen, where the color patch dragging behavior includes: the entire dragging trajectory of each test color patch and the individual dragging trajectories involved in the entire dragging trajectory, and the individual dragging trajectories are consistent with the number of dragging times; Analyze the dragging error probability of each test color patch based on the color patch dragging behavior, and generate a gamut distribution in combination with the dragging position deviation of each test color patch; Update the color vision test level according to the gamut distribution and continue the test.

7. The functional vision test method for visual health self-monitoring according to claim 6, characterized in that, Analyze the dragging error probability of each test color patch based on the color patch dragging behavior, and generate a gamut distribution in combination with the dragging position deviation of each test color patch, including: Statistically analyze the time of each unit color block passing through points in each full - length drag trajectory involved in each color vision test question, and construct a trajectory vector , where represents the position of the second unit color block passed through in the full - length drag trajectory and the residence time ; represents the position of the m - th unit color block passed through in the full - length drag trajectory and the residence time ; Screen the positions in the trajectory vector with a residence time greater than the preset time as the first positions, and split the entire dragging trajectory based on the first positions; Determine the color difference between the color identification of the unit color patch at the last position in each individual dragging trajectory and the color of the test identification, and obtain the color difference vector and position difference vector corresponding to the entire dragging trajectory; When the number of individual dragging trajectories is 1, keep the color difference vector and position difference vector unchanged; When the number of individual dragging trajectories is multiple, determine the number of dragging times and the dragging direction each time at each first position, and draw a dragging vector diagram; Determine the dragging invalidity between the last position in each individual dragging trajectory and the first position of the next individual dragging trajectory according to the dragging vector diagram, and adjust the color difference and position difference of the last position to update the color difference vector and position difference vector; Determine the number of test color patches in each gamut under each color vision test question, and construct a color difference matrix based on the final color difference vector to obtain the color difference feature vector corresponding to the gamut. At the same time, construct a position difference matrix based on the final position difference vector to obtain the position difference feature vector corresponding to the gamut; Obtain the dragging error probability of each test color patch according to the color difference vector, and combine the color difference feature vector and position difference feature vector to obtain the error vertical value corresponding to the gamut; Generate a gamut distribution based on the error vertical values under different gamuts.

8. The functional vision test method for visual health self-monitoring according to claim 1, characterized in that Generate a remote visual assessment report, including: Obtain the visual acuity threshold based on all the visual acuity test questions at different difficulty levels of the visual acuity test and the fusion features of each visual acuity test question; Obtain the contrast sensitivity curve based on all the contrast sensitivity test questions at different difficulty levels of the contrast sensitivity test and the fusion features of each contrast sensitivity test question; Average each gamut error vertical value in the gamut distribution of the color vision test questions involved at different difficulty levels of the color vision test to obtain the color vision arrangement error distribution; Generate a remote visual assessment report based on the visual acuity threshold, contrast sensitivity curve, and color vision arrangement error distribution.

Images