Color vision testing system and method

Abstract

The invention provides a color vision testing system and method, and is applied to the technical field of data processing. The method comprises the following steps: processing to-be-checked background picture information and to-be-checked background adjustment information to generate a color block selection influence factor vector; processing the medical record information of the target user to generate a user color block adjustment factor; processing the target user information based on the color block selection influence factor vector and the user color block adjustment factor, and generating preset color block information and color block selection information of the target user; processing a preset color difference recognition model based on the training sample set to generate a target color difference recognition model; processing the preset color block information and the color block selection information of the target user based on the target color difference identification model to generate target color block identification information; the target color block identification information is processed, target color difference identification information is generated, and the target color difference identification information is used for representing that the target user has any color difference information of red, orange, yellow, green, blue, blue and purple.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and in particular to a system and method for color vision testing. Background Art

[0002] Color vision detection is of key significance in many fields such as medicine, transportation, and professional qualification certification. Its accurate assessment is crucial to ensuring people's quality of life, work safety, and the normal operation of society. In the prior art, the widely used color blindness books (such as Ishihara's color blindness test chart, etc.) and pseudo-isochromatic color chart detection methods are mainly detected by observing graphic discrimination. However, this method relies heavily on the patient's cognitive and comprehension abilities, and has large errors for people with low cognitive levels (such as children, mentally retarded people, etc.). Moreover, its test results are mostly qualitative judgments (such as normal, color blindness, color weakness, etc.), lack of accurate quantitative data, and cannot carefully reflect the degree of color vision abnormality. Although color vision can be quantitatively measured by color vision mirror method, it is complicated to operate, requires professional personnel to operate and interpret the results, and is not convenient for the general public to use. The equipment is expensive and large in size, and it is difficult to be widely popularized, which limits its application in large-scale screening.

[0003] With the development of society, the requirements for the accuracy, comprehensiveness, efficiency and universality of color vision detection in various fields are increasing. For example, the transportation industry needs to ensure that drivers have accurate color vision to ensure traffic safety; the medical field needs more accurate detection to assist in disease diagnosis and treatment; some occupations in industrial production have strict requirements on color vision and require accurate screening of qualified personnel. Therefore, there is an urgent need for an innovative color vision detection technology to overcome the shortcomings of existing methods and meet diverse practical needs.

[0004] It should be noted that the information disclosed in the above background technology section is only used to enhance the understanding of the background of the present disclosure, and therefore may include information that does not constitute the prior art known to ordinary technicians in the field. Summary of the invention

[0005] The purpose of this application is to provide a system and method for color vision testing, at least to a certain extent, to overcome the problems of the prior art, by changing the detection range that was previously based on red, green or yellow and blue, and dividing it according to the color nm number, measuring seven colors of red, orange, yellow, green, cyan, blue and purple, and comprehensively diagnosing the recognition of each color by the human eye, so as to accurately judge the color vision ability of the human eye. Breaking through the limitation of only giving qualitative conclusions in the past, it can display specific objective data such as the minimum recognizable color difference within a specific color range, and define it as the minimum difference that can be recognized by the human eye alone, so as to achieve the refinement of color vision detection and accurately evaluate the degree of color vision abnormality. Using two color blocks to display, the color block color is made consistent by dragging the color bar. This method is fast and accurate, not affected by the cognitive differences of users, can effectively shorten the detection time, improve the detection efficiency, and it only takes 5 minutes to test 7 colors per person. Setting the detection background to a variety of life scenes (static or dynamic scenes such as driving, sports, roads, etc.) is more in line with the actual visual environment of life, and can find color vision abnormalities missed in hospital detection, helping users understand the color vision problems faced in life.

[0006] Other features and advantages of the present application will become apparent from the following detailed description, or may be learned in part by the practice of the present invention.

[0007] According to one aspect of the present application, a method for color vision testing is provided, including: obtaining background image information to be checked, target user information, background adjustment information to be checked, medical record information of the target user, a preset color difference recognition model and a training sample set, wherein the background image information to be checked is a dynamic life background or a static life background, and the background adjustment information to be checked is used to characterize the background information customized by the target user; processing the background image information to be checked and the background adjustment information to be checked to generate a color block selection influencing factor vector; processing the medical record information of the target user to generate a user color block adjustment factor; based on the The target user information is processed based on the color block selection influencing factor vector and the user color block adjustment factor to generate preset color block information and color block selection information of the target user; the preset color difference recognition model is processed based on the training sample set to generate a target color difference recognition model; the preset color block information and the color block selection information of the target user are processed based on the target color difference recognition model to generate target color block recognition information; the target color block recognition information is processed to generate target color difference recognition information, wherein the target recognition color difference information is used to characterize the existence of any color difference information of red, orange, yellow, green, cyan, blue and purple for the target user.

[0008] Another aspect of the present application is a device for color vision testing, characterized in that it includes: an acquisition module, used to obtain background image information to be checked, target user information, background adjustment information to be checked, medical record information of the target user, a preset color difference recognition model and a training sample set, wherein the background image information to be checked is a dynamic life background or a static life background, and the background adjustment information to be checked is used to characterize the background information customized by the target user; a processing module, used to process the background image information to be checked and the background adjustment information to be checked to generate a color block selection influencing factor vector; and to process the medical record information of the target user to generate a user color block. adjustment factor; processing the target user information based on the color block selection influencing factor vector and the user color block adjustment factor to generate preset color block information and color block selection information of the target user; processing the preset color difference recognition model based on the training sample set to generate a target color difference recognition model; processing the preset color block information and the color block selection information of the target user based on the target color difference recognition model to generate target color block recognition information; processing the target color block recognition information to generate target color difference recognition information, wherein the target recognition color difference information is used to characterize the existence of any color difference information of red, orange, yellow, green, cyan, blue and purple for the target user.

[0009] According to another aspect of the present application, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a second processor, the color vision test method described above is implemented.

[0010] A system and method for color vision test provided by the present application, by changing the detection range based on red, green or yellow and blue in the past, according to the color nm number custom division, measure the seven colors of red, orange, yellow, green, blue, purple, and comprehensively diagnose the recognition of each color by the human eye, so as to accurately judge the color vision ability of the human eye. Break through the limitation of only giving qualitative conclusions in the past, can show specific objective data such as the minimum recognizable color difference (such as 2nm) in a specific color range (such as red 635nm-700nm), and define it as the minimum difference that the human eye can recognize for a single color, realize the refinement of color vision detection, and accurately evaluate the degree of color vision abnormality. Use two color blocks to show, by dragging the color bar to make the color block color tend to be consistent, this method is fast and accurate, is not affected by the cognitive differences of users, can effectively shorten the detection time, improve the detection efficiency, and it only takes 5 minutes to test 7 colors per capita. Set the detection background to a variety of life scenes (such as static or dynamic scenes such as driving, sports, roads, etc.), which is more in line with the actual life visual environment, can find the color vision abnormality missed in the hospital detection, and help users understand the color vision problems faced in life.

[0011] It is to be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present disclosure. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] Figure 1 A flow chart showing a method for color vision testing provided by an embodiment of the present application;

[0013] Figure 2 A schematic diagram of the structure of a color vision testing device provided by an embodiment of the present application is shown;

[0014] Figure 3 A schematic diagram of an operation interface for a color vision test provided in an embodiment of the present application is shown. DETAILED DESCRIPTION

[0015] The preferred embodiments of the present invention are described below in conjunction with the accompanying 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.

[0016] Combine the following Figure 1 The method of color vision testing according to an exemplary embodiment of the present application is described below. In one embodiment, the present application also provides a system and method for color vision testing. Figure 1 The following schematically shows a flow chart of a method for color vision testing according to an embodiment of the present application. Figure 1 As shown, the method is applied to a server, comprising:

[0017] S101, obtaining background image information to be checked, target user information, background adjustment information to be checked, medical record information of the target user, a preset color difference recognition model and a training sample set.

[0018] In one implementation, the background image information to be checked is a dynamic life background or a static life background, and the background adjustment information to be checked is used to characterize the background information customized by the target user. The system provides a background image library, which contains a variety of dynamic life backgrounds (such as videos of urban traffic intersections, dynamic pictures of sunrise and sunset at the seaside, etc.) and static life scenes (such as indoor living room layout pictures, outdoor park scenery photos, etc.). The picture format supports common JPEG, PNG and other formats, and the video format supports MP4 and other formats. The resolution of these background pictures is moderate to ensure clear display on different devices while taking into account the system processing efficiency. For example, for a general computer screen, the picture resolution can be set to 1920×1080 pixels and the video resolution to 1280×720 pixels.

[0019] Users can select background images from the image library that they think are similar to color vision scenes in real life. For example, users who drive frequently can choose traffic intersection scene images. If users want to use photos or videos they have taken as backgrounds, the system provides an upload function that supports direct uploading after shooting with a mobile phone or computer camera. During the upload process, the system automatically detects the format, resolution and other information of the image or video. If it does not meet the requirements, the user is prompted to make corresponding adjustments (such as adjusting the resolution, converting the format, etc.). For example, a user took a photo of his own garden, but the resolution was too high (such as 4000×3000 pixels). The system prompts the user to reduce the resolution to an appropriate range before uploading to ensure that the system can quickly process the background image information.

[0020] A special user information entry form is designed on the software interface, and the user manually enters basic information such as name (such as Zhang San), gender (such as male), date of birth (such as May 10, 1985). At the same time, to ensure the accuracy of the information, the system strictly verifies the input format, such as the name cannot be empty and the length is limited to a certain range (such as 2-20 characters), the gender can only be selected as "male" or "female", and the date of birth must conform to the date format (such as YYYY-MM-DD). If the user enters incorrect information, the system will immediately pop up a warning box to prompt the user to modify the incorrect content until the correct information is entered. For example, when the user enters only one character in the name, the system will pop up a warning "The name input is incomplete, please re-enter". The system has the function of docking with the medical institution information system to obtain more comprehensive user information. After obtaining the user's authorization, the system connects to the medical institution's HIS system through a secure data interface to obtain the user's medical history information (such as whether there is a history of eye diseases, systemic diseases, etc.), past examination records (such as previous color vision test results, eye examination reports, etc.). For example, if a user has undergone eye surgery in a hospital, the system will obtain detailed information such as the surgery time, surgery type (such as cataract surgery), and postoperative recovery status from the hospital's HIS system. This information will be used as part of the target user information to provide richer data support for subsequent color vision tests, so as to more accurately analyze the user's color vision condition.

[0021] In the background setting interface of the software, a series of adjustment tools are provided to users, such as the brightness adjustment slider (ranging from -100 to 100, representing the percentage of brightness reduction or increase), the contrast adjustment slider (ranging from -50 to 50, indicating the degree of contrast reduction or increase), the saturation adjustment slider (ranging from -50 to 50, used to adjust the color vividness), and the cropping function button (the cropping area can be freely selected). Users can operate the selected background image according to their own visual experience and actual needs. For example, if the user feels that the background image is too bright, the brightness slider can be dragged to the left to -30 to reduce the brightness of the image; if the background color is not bright enough, the saturation can be appropriately increased (such as adjusting the saturation slider to 20). After the user adjusts the background, the system records the user's operation parameters in real time as the background adjustment information to be checked. For brightness adjustment, the system records the percentage change between the adjusted brightness value and the original brightness value; contrast adjustment records the difference between the adjusted contrast value and the original contrast; saturation adjustment records the difference between the adjusted saturation value and the original saturation; and cropping operation records the coordinate information of the cropping area. For example, the original background image has a brightness of 50%, and after the user adjusts it to 30%, the system records the brightness adjustment value as -20% (indicating a 20% decrease in brightness). These quantified adjustment information will be used to subsequently generate a color block selection influencing factor vector to meet the personalized needs of different users for backgrounds and improve the accuracy and reliability of color vision detection.

[0022] If the user has an electronic medical record file, the system provides an upload entry and supports common electronic medical record formats, such as PDF, XML, etc. After the user selects the medical record file to upload, the system automatically identifies the file format and parses it. For example, if the user uploads a medical record in PDF format, the system uses a special PDF parsing tool to extract the text information, including disease diagnosis (such as glaucoma, diabetic retinopathy, etc.), illness time (such as diabetes diagnosed in 2015), treatment process (such as drug treatment records, surgical records, etc.) and doctor's diagnosis opinions. During the parsing process, the system marks and classifies key information for subsequent processing. For users who do not have electronic medical records or whose medical record information is incomplete, the system provides the function of manually entering medical record information. Design a corresponding form on the interface to guide users to enter important information such as disease name, illness duration, and treatment status. In order to help users enter accurately, the system provides a drop-down menu of common disease names for users to choose from, and at the same time verifies the input time format. For example, a user manually enters that he or she suffers from macular degeneration, has been ill for 3 years, and is currently receiving medication. The system records this information and integrates it with other acquired medical records to ensure the integrity of the user's medical records and provide an accurate basis for generating the user's color block adjustment factor, thereby making the color vision test more in line with the user's actual health status.

[0023] The preset color difference recognition model is pre-built using a deep learning architecture, such as a model structure based on a convolutional neural network (CNN). The initial parameters of the model are obtained by pre-training on a large sample of people with normal color vision and people with abnormal color vision. For example, the model structure contains multiple convolutional layers (used to extract local features of the image, such as color edges, texture patterns, etc.), pooling layers (used to reduce data dimensions and enhance feature robustness), and fully connected layers (used to integrate features and perform classification predictions). Parameters such as the convolution kernel size and step size of the convolutional layer are set based on a large number of experiments to optimize the model's ability to extract color block image features.

[0024] The trained preset color difference recognition model is stored in the model library specified by the system. The model file format can be the common H5 format (applicable to the Keras framework) or PT format (applicable to the PyTorch framework), etc. When performing a color vision test, the system accurately loads the preset color difference recognition model from the model library into the memory according to the user's operation instructions to ensure that the model can participate in the subsequent calculation and prediction process in a timely manner. For example, when the user starts the test, the system detects that the preset color difference recognition model needs to be used, and then finds the corresponding model file from the local model library, loads it into the running memory, and initializes the parameters and status of the model to prepare for the subsequent processing of the preset color block information and the color block selection information of the target user.

[0025] The training sample set is collected from a wide range of sources, including color vision test data from professional ophthalmology institutions (such as detailed records of color vision examinations conducted by hospitals on users, including test results for people of different ages, genders, and health conditions), synthetic image data simulating different color vision abnormalities (using image processing algorithms to generate color block combination images with different degrees of color vision abnormalities, simulating various types of color vision defects), and data that is continuously expanded through actual user test feedback (new user test data is added to the training sample set after screening and annotation). For example, 1,000 sets of color vision test data from real users were obtained from a large ophthalmology hospital, including data from users with different types of color vision abnormalities (such as red-green color blindness, color weakness, etc.) and people with normal color vision; at the same time, 500 images simulating color vision abnormalities were generated using image synthesis technology, covering combinations of color blocks of different colors and different degrees of color difference, in order to enrich the diversity of training samples.

[0026] For each sample obtained, detailed annotations are made, including color information of the color block (such as the specific wavelength range of red, orange, yellow, etc.), color difference information (color difference between color blocks expressed in nanometers), color vision abnormality type (such as normal, mild abnormality, moderate abnormality, color blindness, etc.) and corresponding user feature information (such as age, gender, whether there is a history of eye disease, etc.). Before the sample data is input into the model for training, the data is preprocessed, such as normalization of image data (mapping pixel values ​​to a specific numerical range, such as between 0-1) and data enhancement operations (rotating, flipping, scaling, etc. the image to increase the number and diversity of samples) to improve the generalization ability of the model and the training effect. For example, a sample image consisting of a red block (wavelength range 630nm-700nm) and a green block (wavelength range 490nm-560nm) is annotated with a color difference of 140nm (calculated based on the wavelength range), and the type of color vision abnormality is normal (the sample comes from a population with normal color vision). After the image is normalized, it is added to the training sample set to provide accurate and rich data support for model training, ensuring that the model can learn effective features and patterns, thereby improving the accuracy of identifying color vision abnormalities.

[0027] S102, processing the to-be-checked background picture information and the to-be-checked background adjustment information to generate a color block selection influencing factor vector.

[0028] In one implementation, feature extraction is performed on the background image information to be inspected to generate color feature information, brightness feature information, and texture feature information. The background image to be inspected is converted from the RGB color space to the HSV color space, because the HSV space is more in line with the way humans perceive color, which is convenient for subsequent feature extraction. For example, for a pixel with an RGB value of (128, 128, 255), the converted HSV value is (240, 1, 1) (this is only an example, and the actual calculation will involve complex conversion formulas). Calculate the histogram of hue (H), saturation (S), and brightness (V) in the HSV space. Divide the hue range into 36 intervals (0-359 degrees, each interval is 10 degrees), and count the number of pixels in each interval. For example, if in a background image, the proportion of pixels in the blue hue interval (220-230 degrees) reaches 30%, it means that blue is one of the main colors of the image. At the same time, the mean and variance of saturation and brightness are calculated. For example, the mean saturation value is 0.5 and the variance is 0.1. These statistics can reflect the vividness of the color and the overall brightness distribution as part of the color feature information.

[0029] For a picture of 800×600 pixels, traverse the brightness value of each pixel (which can be the brightness component after RGB value conversion, or the V value in HSV space), add up the brightness values ​​of all pixels and divide by the total number of pixels to get the average brightness value. Assuming that the sum of the brightness values ​​of all pixels in the picture is 240,000, then the average brightness value L = 240,000 / (800×600) = 0.5 (here it is assumed that the brightness value has been normalized to the range of 0-1). In addition to the average brightness, the distribution of brightness is also calculated, such as the brightness histogram (dividing the brightness value range into several intervals and counting the number of pixels in each interval), the variance or standard deviation of the brightness, etc. For example, if the brightness histogram shows that there are more pixels in the low brightness range (0-0.3) and the brightness variance is small, it means that the background image is dark overall and the brightness change is not obvious. This feature will affect the subsequent color block selection strategy, because the darker background makes bright color blocks easier to select, and the distinction between dark color blocks and the background is reduced. It is necessary to consider adjustment when generating the color block selection influencing factor vector.

[0030] Convert the background image to a grayscale image, and then calculate the gray-level co-occurrence matrix (GLCM). Select appropriate distance and direction parameters, such as a distance of 1 pixel and directions of horizontal, vertical, and diagonal. For example, for the horizontal GLCM, the element (i, j) represents the number of times a pixel with a grayscale value of i and a pixel with a horizontal distance of 1 pixel and a grayscale value of j appear at the same time in the image. A series of texture feature parameters, such as contrast, energy, entropy, etc., can be calculated using GLCM. Contrast is calculated using the formula Contrast = ∑ i,j |ij| 2 P(i,j) (where P(i,j) is an element in the gray-level co-occurrence matrix) reflects the degree of change of the gray value of the pixel in the image. High contrast indicates clear texture and rich changes. Calculate energy (Energy), the formula is Energy = ∑ i,j P(i,j) 2 , a large energy value indicates that the texture distribution in the image is relatively uniform; calculate entropy (Entropy), the formula is Entrogr = -∑ i,j P(i,j)log(P(i,j)), the entropy value reflects the complexity of the image texture. The larger the entropy, the more complex the texture. For example, if the contrast of a background image is calculated to be 0.8, the energy is 0.3, and the entropy is 1.2, these texture feature information will be used together with the color and brightness features to generate the color block selection influencing factor vector. Because the complex texture background will interfere with the user's choice of color blocks, the color block display and selection strategy needs to be adjusted according to the texture features.

[0031] The background adjustment information to be checked is processed to generate a feature adjustment amplitude value. If the user adjusts the brightness of the background, the system records the brightness values ​​before and after the adjustment. Assuming that the average brightness of the original background image is 0.5 (as calculated above), the user drags the brightness slider to the right to increase it by 30%, then the adjusted brightness value is 0.5×(1+0.3)=0.65. The brightness adjustment amplitude value is calculated as the difference between the adjusted brightness value and the original brightness value, that is, 0.65-0.5=0.15. This value will be used for subsequent adjustments to the color feature information and brightness feature information to reflect the influence of the user's preference for background brightness on the color block selection. For example, if the brightness adjustment amplitude value is large, it means that the user is more sensitive to brightness changes. When generating the color block selection influencing factor vector, the brightness contrast of the color block needs to be adjusted accordingly to make the color block easier to select and distinguish in the new brightness environment.

[0032] For contrast adjustment, the contrast values ​​before and after the adjustment are also recorded. If the original contrast is 0.5 and the user increases the contrast by 20%, the adjusted contrast is 0.5×(1+0.2)=0.6. The contrast adjustment amplitude is 0.6-0.5=0.1. A higher contrast adjustment amplitude indicates that the user wants to select a color block in a more vivid color contrast environment, which will affect the color contrast-related parameter settings in the color block selection influencing factor vector. For example, when generating color blocks, the contrast difference between different colors is increased to meet the user's visual needs. Assuming that the original saturation average is 0.5, the user increases the saturation by 10%, the adjusted saturation is 0.5×(1+0.1)=0.55, and the saturation adjustment amplitude is 0.55-0.5=0.05. The change in saturation affects the vividness of the color, and thus affects the distinction between the color block and the background. When generating the color block selection influencing factor vector, the saturation range of the color block color is adjusted according to the saturation adjustment amplitude to ensure that the color block can still be clearly identified and selected in the background saturation environment adjusted by the user.

[0033] Based on the feature adjustment amplitude value, the color feature information, brightness feature information and texture feature information are processed to generate a color block selection influencing factor vector. According to the color histogram of the background image and the brightness, contrast and saturation adjustment amplitude values, the weights of different tones in the color feature information are adjusted. If the blue tones in the background image account for a high proportion (such as 30%), and the user increases the contrast, since blue and yellow are contrasting colors, the distinction between blue and yellow will be more obvious when the contrast is increased, so the weights of blue and yellow tones in the color block selection influencing factor vector can be appropriately increased. For example, the weight of the blue tones is increased from the original 0.3 to 0.4, and the weight of the yellow tones is increased from 0.2 to 0.3, so as to guide the user to be more inclined to select blue and yellow related color blocks when generating the preset color block information in the future, or to pay more attention to the selection operation of blue and yellow when the user selects the color block.

[0034] Considering the impact of the saturation and brightness adjustment amplitude values ​​on color perception, the saturation and brightness related parameters in the color feature information are adjusted in an associated manner. If the user increases the saturation and the background brightness is high, in order to avoid the color being too bright and dazzling and affecting the color vision judgment, when generating the color block selection influencing factor vector, the weight of the high-saturation color in the high-brightness area can be appropriately reduced, while the weight of the low-saturation color can be increased to make the color block color selection more balanced. For example, for the red block, if the original saturation weight is 0.4, under the above background adjustment, its weight is reduced to 0.3, while the weight of the green block (assuming the original saturation weight is 0.3) is increased to 0.4, so as to guide the color block selection to be more in line with the requirements of visual comfort and color vision detection accuracy.

[0035] According to the brightness adjustment amplitude value and the average brightness of the background image, adjust the value of the user's sensitivity parameter to brightness in the color block selection influencing factor vector. If the user significantly increases the background brightness (such as the brightness adjustment amplitude value is greater than 0.2), and the background image originally has a high average brightness, it means that the user is more sensitive to bright color blocks. At this time, the brightness sensitivity weight of the bright color blocks can be increased in the factor vector, and the weight of the dark color blocks can be reduced. For example, the brightness sensitivity weight of the bright color blocks (such as white, light yellow, etc.) is increased from the original 0.5 to 0.7, and the weight of the dark color blocks (such as black, dark blue, etc.) is reduced from 0.5 to 0.3. In this way, when generating preset color block information and users select color blocks, the system will pay more attention to the selection and display of bright color blocks to adapt to the user's visual perception characteristics under high-brightness backgrounds and improve the accuracy of color vision detection. Combined with the contrast adjustment amplitude value, adjust the value of the brightness contrast parameter between the color block and the background in the factor vector. If the user increases the contrast, the acceptable range of the brightness difference between the color block and the background is appropriately expanded, that is, when generating the color block selection influencing factor vector, the brightness contrast between the color block and the background is allowed to be stronger to enhance the visibility and distinguishability of the color block in the background. For example, the original threshold of the brightness difference between the color block and the background is 0.3. After the contrast is increased, the threshold is increased to 0.4, which means that when selecting and generating color blocks, color blocks with a larger brightness difference from the background can be selected, making the color blocks easier to be recognized and selected by users against the adjusted background, thereby more accurately detecting the user's color vision ability.

[0036] According to the texture feature parameters of the background image (such as contrast, energy, entropy, etc.) and the user's adjustment of the background, adjust the value of the weight of the texture complexity on the color block selection in the factor vector. If the background image has a complex texture (such as high texture contrast and large entropy value) and the user has not made obvious adjustments to the background, it indicates that the user has a strong ability to distinguish colors in a complex texture environment. At this time, the weight of the texture complexity on the color block selection can be appropriately reduced, because the user can better select color blocks in this environment. For example, the texture complexity weight is reduced from the original 0.4 to 0.3, so that the color block selection depends more on color and brightness features. On the contrary, if the user adjusts the background, such as reducing the contrast to make the texture blurred, it implies that the user has difficulty distinguishing colors in a complex texture environment. The texture complexity weight should be increased, such as from 0.4 to 0.5, so that the system will consider the texture factor more carefully in the process of generating preset color block information and user selection, and select color blocks with less interference with the background texture to ensure the accuracy of the color vision detection results.

[0037] When calculating texture features, the texture direction (such as texture features in horizontal, vertical, diagonal, etc.) is taken into account. According to the user's adjustment of the background and the preference of the viewing direction (assuming that the user's preference information on the direction can be obtained through the user's operation behavior or other means, such as the user is more inclined to drag the color block horizontally during the operation, suggesting that the visual attention to the horizontal direction is higher), the weight of the influence of different texture directions on the color block selection in the factor vector is adjusted. For example, if the user is more sensitive to horizontal texture (such as operating more accurately under a background with obvious horizontal texture), the weight of the horizontal texture feature in the color block selection influencing factor vector can be increased, such as from the original 0.3 to 0.4, while the weight of texture features in other directions can be reduced, guiding the system to give priority to the color block layout and selection strategy that is adapted to the horizontal texture when generating color blocks and user selection, thereby improving the color vision detection effect of the user under a specific texture direction preference. It can generate a color block selection influencing factor vector based on the background image information to be checked and the background adjustment information in a more comprehensive and detailed manner. This vector can fully consider the impact of background factors and user adjustment operations on color block selection, and provide an important basis for the subsequent generation of personalized preset color block information and accurate color vision detection, making the entire color vision test system more in line with the user's actual visual perception and improving the accuracy and reliability of detection.

[0038] S103: Process the medical record information of the target user to generate a user color block adjustment factor.

[0039] In one embodiment, the medical record information of the target user is processed to generate eye disease characteristic information and physiological characteristic information. Eye disease diagnosis information, such as glaucoma, cataract, macular degeneration, retinitis pigmentosa, etc., is extracted from the user's medical record and classified and labeled according to the disease type. For example, if the medical record records that the user suffers from glaucoma, it is labeled as "glaucoma" type eye disease characteristic information. For each disease, its detailed information is further recorded, such as glaucoma users record intraocular pressure values ​​(such as intraocular pressure of 25mmHg, normal range of 10-21mmHg), visual field defects (such as nasal visual field defects of 30 degrees), etc.; cataract users record the degree of lens opacity (such as nuclear opacity, moderate opacity), etc. These detailed information helps to more accurately assess the impact of eye diseases on color vision. Record the onset time, treatment process and current disease progression status of eye diseases. Taking diabetic retinopathy as an example, the user's diabetes diagnosis time (such as 2010), the time when the retinopathy was first discovered (such as 2015), the treatment methods received (such as the number of laser photocoagulation treatments, the name of the drug used and the course of treatment, etc.), and the current stage of the disease (such as proliferative diabetic retinopathy stage III according to the international staging standard) are recorded. The progression of the disease reflects the changes in the structure and function of the eye, which is of great significance to the calculation of the color vision influencing factor, because different stages of the disease have different degrees of influence on color vision.

[0040] Extract basic physiological information such as the user's age (e.g., 45 years old), gender (e.g., female), height (e.g., 160cm), and weight (e.g., 60kg). Age is one of the important factors affecting color vision. With age, physiological changes such as lens aging and decreased retinal function lead to reduced color sensitivity. Gender is also associated with color vision. Although the overall difference is small, some studies have shown that there are subtle differences in color perception between men and women, so it is also taken into consideration. Height and weight information can be used to calculate body mass index (BMI). Abnormal BMI values ​​(e.g., obesity or thinness) are associated with some systemic diseases, which in turn indirectly affect color vision. For example, obesity increases the risk of diseases such as diabetes, and diabetes can cause retinopathy that affects color vision. Collect information on whether the user suffers from systemic diseases that affect color vision, such as diabetes (record blood sugar control, such as fasting blood sugar value of 7.5mmol / L, blood sugar value 2 hours after meal of 11.2mmol / L, indicating poor blood sugar control), hypertension (record blood pressure values, such as systolic blood pressure 145mmHg, diastolic blood pressure 90mmHg), cardiovascular disease, etc. At the same time, understand the user's living habits, such as smoking history (number of cigarettes smoked per day, smoking years, such as smoking 15 cigarettes a day, smoking history of 10 years), drinking (number of times per week, each drink amount, such as drinking 3 times a week, each drink amount is 100ml of liquor), eye habits (daily use of electronic devices, such as using computers and mobile phones for more than 8 hours a day), etc. These systemic diseases and lifestyle factors have a potential impact on color vision by affecting the health of blood circulation, nervous system or eye tissue, and are an important part of generating physiological characteristic information.

[0041] The characteristic information of eye diseases and physiological characteristic information are processed to generate color vision influencing factors and weight information matching the color vision influencing factors. According to the research data and clinical experience of the impact of different eye diseases on color vision, the corresponding color vision influencing factors are set for each eye disease. For example, due to damage to the optic nerve, glaucoma users have decreased color sensitivity and weakened color discrimination ability, and their color vision influencing factors are set to 0.8 (indicating a greater impact on color vision, the specific value can be obtained based on the statistical analysis of color vision test data of a large number of glaucoma users); cataract users have crystalline opacity that affects light transmission, so the color vision influencing factor is set to 0.6 (moderate impact); macular degeneration users mainly affect central vision and the ability to distinguish color details, so the color vision influencing factor is set to 0.7. For the same disease, the influencing factors can be further subdivided according to the severity of the disease. For example, the color vision influencing factor of early glaucoma is 0.6, 0.8 in the middle stage, and 1.0 in the late stage (indicating a very serious impact on color vision).

[0042] The color vision impact factor is adjusted in combination with physiological characteristics information. In terms of age factors, the adjustment value is calculated using an age-related function. For example, for users aged 40-50, the color vision impact factor increases by 0.02 for every additional year (assuming the basic impact factor is 0.5). If the user is 45 years old, the color vision impact factor increases by 0.1 (0.02×5) due to age factors. For systemic diseases, such as diabetic users, the color vision impact factor is adjusted according to the blood sugar control situation. When blood sugar control is poor (such as the above-mentioned fasting blood sugar and postprandial blood sugar values ​​exceed the normal range by a large amount), the color vision impact factor increases by an additional 0.3; for users with hypertension, if the blood pressure is unstable for a long time (such as systolic blood pressure continuously higher than 140mmHg or diastolic blood pressure continuously higher than 90mmHg), the color vision impact factor increases by 0.2. Lifestyle factors are also involved in the adjustment. For users with a long smoking history (such as a smoking history of more than 10 years) and a large daily smoking volume (such as more than 15 cigarettes), the color vision impact factor increases by 0.1; for long-term use of electronic devices every day (such as more than 8 hours), the color vision impact factor increases by 0.05. These adjusted color vision impact factors comprehensively reflect the potential impact of eye diseases and physiological characteristics on user color vision.

[0043] By analyzing a large amount of clinical data, it is found that the impact of eye diseases on color vision is relatively direct and significant, so it is given a higher weight. For example, the weight of eye diseases is set to 0.7, which means that when considering the factors affecting color vision, eye diseases account for a large proportion. Within eye diseases, weights are assigned according to factors such as the incidence of different diseases and the severity of the impact on color vision. For example, the weight of glaucoma is 0.3 (it accounts for a high proportion among eye diseases because it has a more serious and common impact on color vision), the weight of cataracts is 0.2, the weight of macular degeneration is 0.15, and the weight of diabetic retinopathy is 0.25 (considering the large number of diabetic users and the obvious impact of retinopathy on color vision). The sum of these weights is 1, which indicates the relative importance of the impact of each disease type on color vision within the eye disease factor. The weight of physiological feature information is set to 0.3, indicating that it also plays a certain role in the impact of color vision, but the weight is relatively low compared to the eye disease factor. Within the physiological characteristics, the weight of age is 0.4 (because age is a common and important physiological factor affecting color vision), the weight of gender is 0.1 (gender has a relatively small effect on color vision), the weight of systemic diseases is 0.3 (systemic diseases indirectly affect color vision through various pathways and are of certain importance), and the weight of lifestyle habits is 0.2 (lifestyle habits have an impact on color vision under long-term effects). For example, among systemic diseases, the weight of diabetes is 0.4 (because its impact on the eyes is more common and serious), the weight of hypertension is 0.3, the weight of cardiovascular disease is 0.2, and the weight of other diseases is 0.1, etc. The sum of these weights is 1, which reflects the relative importance of the impact of each sub-factor within the physiological characteristic factor on color vision. Through such weight distribution, the degree of influence of eye diseases and physiological characteristics on color vision can be more reasonably considered, and the contribution ratio of each factor can be accurately reflected in the subsequent calculation of the color block adjustment value.

[0044] The color vision influencing factors and the weight information matching the color vision influencing factors are processed to generate color block adjustment values. Using the given calculation formula Calculate the color patch adjustment value, where n represents the number of major factor categories affecting color vision, m represents the number of subdivided factors under each major factor category, p represents the number of special factors related to eye diseases, i represents the index of the major factor category affecting color vision, j represents the index of the subdivided factors under each major factor category, and F ij represents the color vision impact factor of the jth subdivision factor in the i-th main factor, which is used to quantify the impact of the subdivision factor on color vision. ij represents the weight of the jth sub-factor in the i-th main factor, indicating the relative importance of the sub-factor in the influence of the main factor on color vision. k represents the index of special factors related to eye diseases. G k represents the impact factor of the kth special factor, H krepresents the weight of the kth special factor, I t represents the influencing factor of the tth lifestyle factor, J t represents the weight of the tth lifestyle factor, and K represents the individual difference coefficient, which is used to adjust the impact of differences between individuals in physiological functions, gene expression, etc. on color vision.

[0045] Assume that the main factors affecting color vision are eye diseases (n=1) and physiological characteristics (n=2) (this is just a simplified example, and there are more categories in reality). For the eye disease category (i=1), subdivided factors such as disease type (j=1 to 4, corresponding to glaucoma, cataract, macular degeneration, and diabetic retinopathy, respectively), assuming that the color vision influencing factor F for glaucoma is 11 =0.8, weight W 11 =0.3; Cataract F 12 =0.6,W 12 =0.2; Macular degeneration F 13 =0.7,W 13 =0.15; diabetic retinopathy F 14 =0.75,W 14 =0.25. Special factors related to eye diseases (such as family history, assuming k = 1), if there is a family history, G1 = 0.5, H1 = 0.8 (indicating that family history has a greater impact on color vision and has a higher weight). For the physiological characteristics category (i = 2), subdivided factors such as age (j = 1), systemic diseases (j = 2), etc., assuming that the age factor color vision influence factor F 21 =0.1 (calculated based on the above age), weight W 21 =0.4; Diabetes factor in systemic diseases F 22 =0.3 (the influencing factor adjusted for poor blood sugar control), weight W 22 =0.3. Lifestyle factors (assuming t=1 to 3, corresponding to smoking, drinking, and eye habits), smoking I1=0.1, J1=0.2; drinking I2=0.05 (assuming that the amount of drinking is relatively small, the impact is relatively small), J2=0.05; eye habits I3=0.05, J3=0.2, individual difference coefficient K=1.1 (assuming that the user has good overall physical fitness, the impact on color vision is relatively small, and the coefficient is slightly greater than 1). The calculation process is as follows:

[0046]

[0047] V = (0.6525 + 0.13) + 0.4 + (0.035 × 1.1) = 1.221. The calculated color block adjustment value V = 1.221 will be used to generate the user color block adjustment factor in the future. It combines multiple factors such as eye disease characteristic information, physiological characteristic information and their respective weights, special factors, and individual difference coefficients, reflecting the quantitative demand for adjusting the color block selection according to the user's medical history and physiological characteristics, so as to more accurately adapt to individual differences in color vision testing and improve the accuracy of test results.

[0048] The color block adjustment value is processed to generate a user color block adjustment factor. The user color block adjustment factor is generated according to the calculated color block adjustment value. A simple method is to normalize the color block adjustment value and map it to a specific range (such as 0-1 or -1-1) to determine the user color block adjustment factor. Assuming that the 0-1 range is used for mapping, if the calculated color block adjustment value V = 1.221, through the normalization formula (such as adjustment factor = V / (V max -V min ), assuming V max =2, V min =0) to calculate the adjustment factor of 0.6105. This adjustment factor will be used to subsequently adjust the preset color block information and the color block selection information of the target user. For example, in the color range selection of the preset color block, if the wavelength range of the original red color block is 630nm-700nm, it is adjusted to 635nm-695nm according to the adjustment factor to adapt to the color vision conditions of users with specific eye diseases and physiological characteristics, making the test more targeted and accurate. At the same time, when the user selects a color block, the adjustment factor can also affect the display properties of the color block (such as brightness, contrast, saturation, etc.) or the difficulty of selection (such as adjusting the sensitivity of dragging the slider, increasing or decreasing the selection steps, etc.), ensuring that the test process can accurately reflect the user's actual color vision ability and reduce test errors caused by individual differences.

[0049] S104: Process the target user information based on the color block selection influencing factor vector and the user color block adjustment factor to generate preset color block information and color block selection information of the target user.

[0050] In one implementation, the target user information is processed to generate the color vision information to be detected. Basic information such as age (such as 35 years old) and gender (such as male) are extracted from the target user information. According to the age factor, referring to the research data related to color vision development and age, the color vision of 35-year-olds is basically mature, but with the increase of age, a slight trend of color vision change begins to appear. There are certain differences between men and women in the sensitivity of certain colors (such as men are slightly less sensitive to green than women), and this information is recorded. For example, the age information is quantified as an age-related coefficient (such as 35 years old The corresponding coefficient is 0.95, indicating the relative state of color vision relative to normal adults, and the coefficient range of different ages can be determined based on a large amount of experimental data), and the gender information is marked as "male" for subsequent preliminary adjustment of the color vision information to be detected. If the user has a history of eye disease or systemic disease, the impact of these medical histories on color vision is analyzed in detail. For example, if the user has suffered from mild retinopathy, the type of lesion, the time of illness (such as illness 2 years ago), treatment (such as receiving laser treatment, vision has recovered to a certain extent after treatment) and other information are recorded. According to medical research, retinal lesions affect the ability to distinguish red and green, especially in the visual field corresponding to the lesion area. These medical history information is converted into descriptive information on the impact on color vision, such as "there is a mild disorder in red-green discrimination, and the affected range is about 20% of the visual field (estimated based on the degree and range of the lesion)", and integrated with the basic information to form part of the color vision information to be tested, so as to adjust the color block selection range and test strategy in a targeted manner in the subsequent steps.

[0051] The color vision information to be detected is processed based on the color block selection influence factor vector to generate a color selection range value. Assume that the blue hue weight in the color block selection influence factor vector is high (such as 0.4, higher than the weights of other colors), and the blue tone accounts for a large proportion in the background image (such as 40%), and the brightness is high (such as the average brightness value is 0.7). Based on this information, when generating the color selection range value, for the blue-related color range (such as blue, cyan, purple, etc.), the selection range is appropriately expanded. For example, for blue, the original standard wavelength range is 450nm-490nm. According to the influence factor, it is adjusted to 445nm-495nm to increase the probability of blue colors appearing in the test and adapt to the background and user's visual preferences. For colors that contrast strongly with blue (such as yellow), their ranges are also appropriately adjusted, such as from 560nm-590nm to 555nm-595nm, to enhance the color contrast effect, so that users can select color blocks and make color judgments under this background. Considering the high background brightness (such as the above average brightness value of 0.7), when generating the color selection range value, the selection range for bright areas (such as light blue, light green, light yellow, etc.) is appropriately narrowed to avoid the low distinction between bright colors affecting the test effect; and the range for dark areas (such as dark blue, dark green, purple, etc.) can be appropriately expanded. For example, the wavelength range of light blue was originally 470nm-490nm, which was adjusted to 475nm-485nm; the dark blue range was adjusted from 450nm-470nm to 445nm-475nm. At the same time, according to the saturation of the background (such as the average saturation value of 0.6, which is relatively high), the selection range of high-saturation colors (such as bright red, green, blue, etc.) is appropriately fine-tuned, such as red from 630nm-700nm to 632nm-698nm, to avoid visual fatigue or difficulty in distinguishing due to overly bright colors, and to ensure that the color selection range of the color block under the current background can not only reflect the color difference, but also facilitate user operation and accurate judgment of color vision.

[0052] The color selection range value is processed based on the user color block adjustment factor to generate preset color block information. If the user color block adjustment factor shows that the user suffers from a certain eye disease (such as glaucoma, the color vision impact factor is 0.8 and the weight is 0.3), the preset color block information is adjusted according to the characteristics of the disease's impact on color vision. Users with glaucoma have difficulty distinguishing low-contrast colors, so the number and types of low-contrast color combinations are increased in the preset color blocks. For example, increase the number of color block pairs of gray and similar colors (such as light gray and white, dark gray and black, etc.), as well as color block pairs with similar colors but different saturations (such as light red and dark red), while reducing the proportion of high-contrast color combinations (such as red and green) to more accurately detect the user's color vision condition under the influence of the disease. For color blocks related to the color range affected by glaucoma (such as colors on the blue-yellow axis, because glaucoma affects the transmission of color information by the optic nerve, especially the color perception on this color axis), the wavelength range selection is further refined. For example, blue is subdivided from 450nm-490nm into two sub-ranges of 450nm-470nm and 470nm-490nm, and different color blocks are set for each sub-range, so as to more accurately evaluate the user's color vision discrimination ability in this color area.

[0053] Considering the user's physiological characteristics (such as the color vision impact factor increases by 0.1 due to age at 45 years old) and living habits (such as the color vision impact factor increases by 0.05 due to long-term use of electronic devices every day), the preset color block information is optimized. Older age leads to a decrease in sensitivity to color, especially the weakening of the perception of blue and green. Therefore, different brightness and saturation combinations of these two colors are appropriately added to the preset color blocks, such as dark blue, light blue, dark green, light green and other color blocks with various changes, to improve the ability to detect subtle changes in user color vision. For users who use electronic devices for a long time, there is eye fatigue and temporary decrease in color discrimination ability. Some soothing color combinations are added to the preset color blocks (such as light blue and light green blocks to simulate the color matching in the natural environment). At the same time, the brightness and contrast of the overall color block are appropriately reduced to avoid excessive stimulation of the user's eyes, so that the test process is more in line with the user's actual visual conditions and improve the accuracy of the test results.

[0054] Process the preset color block information to generate the color block selection information for the target user. Design a suitable operation interface based on the user's age and cognitive ability (such as young age, 8 years old, limited cognitive ability). For child users, increase the size of the color block (such as increasing the original color block with a side length of 30 pixels to 50 pixels) to make the color block easier to click and operate; reduce the number of color blocks that need to be selected in each test (such as from 7 to 4) to reduce the difficulty of operation; at the same time, simplify the operation process, such as only providing a simple dragging color block function, and not setting complex color adjustment buttons. In terms of color selection order, first present color block pairs with large color differences and easy to distinguish (such as red and blue). As the test progresses, gradually increase the color block pairs with smaller color differences to gradually guide child users to conduct color vision tests and improve their participation and test accuracy.

[0055] During the process of the user selecting a color block, the user's operation behavior is monitored in real time. For example, the speed at which the user drags the color block (such as the number of pixels dragged per second), the pause time (such as staying on a color block for more than 2 seconds is considered a pause), the accuracy of the selection (such as the color difference range with the preset correct color block) and other information are recorded. If the user spends a long time on selecting a color block of a certain color (such as the time to select a red color block exceeds 5 seconds) and the operation is frequent (such as adjusting the position of the red color block many times), it means that the color is close to the user's color vision limit or the user has difficulty in distinguishing the color. Based on this feedback information, the system dynamically adjusts the selection range and difficulty of subsequent color blocks. For example, for this user, the selection range of red-related colors (such as orange, purple and other colors similar to red) is narrowed, and the color block pairs with more obvious color differences are increased. At the same time, the difficulty of operation is reduced (such as increasing the sensitivity of dragging the slider to make the color adjustment more obvious) to more accurately test the user's color vision ability and avoid inaccurate test results due to difficult operation. At the same time, the system can provide real-time voice or text prompts to encourage users to continue the operation or give operational guidance, such as "You did a good job, keep trying to adjust the color blocks to make the colors on both sides look more similar", thereby improving user experience and test results.

[0056] S105, processing the preset color difference recognition model based on the training sample set to generate a target color difference recognition model.

[0057] In one implementation, the training sample set is processed to generate a training set and a verification set. The training sample set is divided into a training set and a verification set by random segmentation. For example, the training sample set has a total of 10,000 samples, which are segmented in a ratio of 80:20, i.e., 8,000 samples are used for the training set and 2,000 samples are used for the verification set. During the segmentation process, it is ensured that the distribution ratios of various types of samples (such as different types of color vision abnormalities, different age groups, different genders, etc.) in the training set and the verification set are roughly the same to ensure that the training and verification of the model are representative. For example, in the original sample set, the red-green color blindness user samples account for 10%, and in the divided training set and verification set, the red-green color blindness user samples account for nearly 10%. If the sample has obvious stratified characteristics (such as being divided into three layers of mild, moderate, and severe according to the severity of color vision abnormality), it can be divided by stratified sampling. First, separate the samples by layer, and then randomly sample each layer to form a training set and a verification set. This ensures that each layer has enough samples to participate in the training and validation process, improving the model's ability to identify different degrees of color vision abnormalities. For example, there are 3,000 samples for the mild color vision abnormality layer, 4,000 samples for the moderate color vision abnormality layer, and 3,000 samples for the severe color vision abnormality layer. When dividing the training set and validation set, samples are drawn from each layer in proportion, such as 60% (1,800) of the mild color vision abnormality layer is used for the training set, 40% (1,200) is used for the validation set, and so on, to ensure that the training set and validation set can fully reflect the distribution of various characteristics of the samples.

[0058] Get any number of data features in the training set, and generate a sampling ratio based on the number of each data feature in the training set. Select data features related to color difference recognition from the training set, such as color information of color blocks (including color name, RGB value, HSV value, etc.), color difference value (color difference between color blocks in nanometers), user's age, gender, whether there is a history of eye disease, etc. For example, for a training sample containing a color block image and user information, extract the RGB value of the red color block (such as 255, 0, 0), the RGB value of the green color block (such as 0, 255, 0) and the color difference value between them (assuming it is 150nm), and record the user's age (such as 30 years old), gender (male) and history of eye disease (none) as data features. Count the number of each type of data features in the training set. Assume that there are 5 different values ​​of the age feature in the training set (such as 20-29 years old, 30-39 years old, 40-49 years old, 50-59 years old, and over 60 years old), and count the number of samples in each age group, such as 1500 samples for 20-29 years old, 2000 samples for 30-39 years old, etc. The sampling ratio is calculated according to the preset sampling strategy. If equal-proportional sampling is adopted, the sampling ratio of each age group is the ratio of the number of samples in this age group to the total number of samples. Taking the age group of 30-39 years old as an example, the sampling ratio = 2000 / 8000 = 0.25, which means that during the sampling process, the samples in this age group have a 25% probability of being selected as sampling features.

[0059] The training set is sampled based on the sampling ratio to generate a preset number of sampling features. According to the calculated sampling ratio, the training set is sampled with replacement. For example, for the above-mentioned age group of 30-39 years old, according to the sampling ratio of 0.25, each time a sample is randomly selected from the training set, the probability of the sample in this age group being selected is 0.25. Repeat the sampling process until the preset number of sampling features is generated. Assuming that the preset number of sampling features is 1000, through multiple random samplings, 1000 sampling features are finally obtained. These sampling features include different age groups, different color combinations, different user characteristics and other situations, which can better represent the overall feature distribution of the training set. During the sampling process, some combination and adjustment operations can be performed on the sampling features to increase the diversity of the samples. For example, for the sampling features of two adjacent age groups (such as 30-39 years old and 40-49 years old), some of their features can be randomly combined, such as combining the color features of the samples in the 30-39 age group with the user features of the samples in the 40-49 age group to generate new sampling features. This can simulate the color vision of users of different age groups in different color environments, enrich the diversity of sampling features, and help improve the generalization ability of the model.

[0060] Based on any data feature and each sampling feature, multiple data groups are generated, wherein each data group contains a preset number of data samples, and at least one data sample includes identification information. Based on the data samples in the multiple data groups, the preset color difference recognition model is trained to generate a trained color difference recognition model. Any data feature (such as a sample of a specific color combination and user information) is selected from the training set, and it is combined with each sampling feature to generate multiple data groups. Each data group contains a preset number of data samples, for example, each data group is preset to contain 5 data samples. Assume that a sample of a red-green color block combination and a female user with no history of eye disease is taken as a benchmark sample, and it is combined with 1000 sampling features in sequence to generate 1000 data groups. In addition to the benchmark sample, each data group also contains 4 different samples obtained by sampling, and these samples are diverse in color, user characteristics, etc. When constructing the data group, ensure that at least one data sample includes identification information, and the identification information is used to characterize whether the sample is a risk factor affecting the user's color difference recognition. For example, samples with severe color vision abnormalities (such as color blindness) are marked as high-risk factors (marked as 1), and samples with normal or mild color vision abnormalities are marked as low-risk factors (marked as 0). In each data set, samples containing identification information are reasonably distributed. For example, in a data set of 5 samples, 1-2 samples containing identification information are set, and a certain proportion of high-risk and low-risk samples are ensured, so that the model can learn the characteristic differences of samples with different risk levels during the training process, and improve the ability to judge the risk of color difference recognition.

[0061] The data samples in the generated multiple data sets are input into the preset color difference recognition model for training. The model adopts a deep learning architecture (such as a convolutional neural network). The color block image data in the data sample is preprocessed (such as normalization, cropping, etc.) and then input into the input layer of the network. The user information and other feature data are encoded and also input into the corresponding layer of the network. The model continuously adjusts the network parameters through the back propagation algorithm to minimize the error between the prediction result and the sample identification information. For example, the model predicts the degree of color vision abnormality of the sample (such as normal, mild abnormality, moderate abnormality, color blindness, etc.) based on the input color block image and user characteristics, and compares it with the actual identification information of the sample to calculate the error, and then adjusts the convolution kernel weight, fully connected layer weight and other parameters in the network according to the error. After multiple iterative training (such as setting the number of training rounds to 100), the accuracy of the model in color difference recognition is gradually improved, and a trained color difference recognition model is generated.

[0062] The trained color difference recognition model is processed based on the validation set to generate a validation result. If the data sample containing identification information in the validation result is a risk factor that represents the risk factor affecting the user's color difference recognition, the trained color difference recognition model is used as the target color difference recognition model. The trained color difference recognition model is processed using the validation set. After the sample data in the validation set is subjected to the same preprocessing operation as the training set, it is input into the trained model to obtain the prediction results of the model for the validation set samples, including the prediction value of the degree of color vision abnormality, the color difference prediction value, etc. For example, for a sample of a blue-yellow color block combination in the validation set, the model predicts that its degree of color vision abnormality is normal, and the color difference prediction value is 30nm (assuming that the wavelength of the blue color block is 470nm and the wavelength of the yellow color block is 580nm, the calculated color difference is 110nm, and the model prediction error is 80nm). Evaluate the model performance based on the validation results. Calculate the evaluation indicators such as the accuracy, recall rate, and F1 value of the model on the validation set to judge the model's recognition ability for different types of color vision abnormalities. For example, if the recall rate of the model for color blind samples in the validation set is low (such as only 60%, indicating that only 60% of the samples that are actually color blind are correctly identified by the model), it means that the model is insufficient in identifying color blind samples. At the same time, check the data samples containing identification information in the verification results. If there are many high-risk samples (representing risk factors that affect user color difference recognition) that are misclassified as normal by the model, it means that the reliability of the model needs to be improved. If the model performs well in various evaluation indicators and has a high recognition accuracy rate for high-risk samples (such as more than 90%), the trained color difference recognition model will be used as the target color difference recognition model for subsequent color vision tests. Otherwise, the model needs to be further optimized (such as adjusting the network structure, increasing training data, optimizing hyperparameters, etc.), and then retrained and verified until the requirements are met.

[0063] S106 , processing the preset color block information and the color block selection information of the target user based on the target color difference recognition model to generate target color block recognition information.

[0064] In one implementation, the preset color block information and the color block selection information of the target user are processed based on the target color difference recognition model to generate the predicted value of the color difference between each color block pair, the color vision abnormality classification information, and the confidence score matching the color vision abnormality classification information. Figure 3 As shown, Figure 3It is an integrated operation interface diagram of the color vision test system, covering various information such as test question setting, background selection, test execution area and result judgment criteria. Specifically, the preset color block information (such as the wavelength range of the red color block is 630nm-700nm, the wavelength range of the green color block is 490nm-560nm, etc.) and the color block selection information of the target user (such as the user adjusts the red color block to a wavelength of 650nm and the green color block to a wavelength of 520nm) are input into the target color difference recognition model. The model calculates the predicted value of the color difference between each color block pair. For example, for the above red and green color block pairs, the model calculates the color difference prediction value as 130nm (obtained by calculating the adjusted color block wavelength difference). The model uses a variety of calculation methods, such as color space distance calculation (such as calculating the Euclidean distance of two colors in the CIELAB color space as the color difference prediction value), or a color difference prediction model obtained by learning a large amount of sample data (such as learning the relationship between different color block combinations and actual color differences through a neural network, thereby predicting the color difference between new color block pairs).

[0065] Color vision abnormality classification information is generated based on the color difference prediction value and the correspondence between different color difference ranges and color vision abnormality types of the model. For example, if the model sets a color difference greater than 100nm and less than 150nm to correspond to mild color vision abnormality (for the red-green color axis), for the 130nm color difference of the above red-green color block pair, the model classifies it as mild color vision abnormality (red-green color axis). The model is trained on a large number of samples with color vision abnormality types, and learns the mapping relationship between different color difference ranges and the degree of color vision abnormality (such as normal, mild abnormality, moderate abnormality, color blindness, etc.), so that the color difference prediction value can be classified according to the new color block. At the same time, the model can also consider other factors, such as the user's operation trajectory and pause time during the selection of the color block, to comprehensively judge the color vision abnormality classification and improve the accuracy of the classification.

[0066] While generating the color vision abnormality classification information, the model assigns a confidence score to each classification result, indicating the model's confidence in the classification result. For example, during the training process, the model learns that for certain specific color block combinations and user operation modes, the reliability of its classification results is high, and the corresponding confidence score will be high. For the case where the red-green color block pair is classified as mild color vision abnormality (red-green color vision axis), the model calculates a confidence score of 0.75 based on its internal calculation, indicating that the model has high confidence in the classification result, but there is still a certain degree of uncertainty. The confidence score can be obtained by probability calculation methods, such as calculating the posterior probability of the classification result based on Bayes' theorem as the confidence score, or by determining it through the accuracy statistics of the model during the training process (for example, for a certain type of classification result, if the model has a high accuracy on the training set, the confidence score for this type of result will also be correspondingly high in the actual prediction).

[0067] The predicted value of the color difference between each color block pair, the color vision abnormality classification information, and the confidence score matching the color vision abnormality classification information are processed to generate user operation features. The operation time of the target user when selecting each color block pair is recorded, such as the time from starting to adjust the color block to completing the selection. Assuming that the user spent 8 seconds to select the blue-yellow color block pair and 12 seconds to select the red-green color block pair, the longer operation time suggests that the user has difficulty in distinguishing the colors of the color block pair or is not proficient in the operation. At the same time, the operation difficulty is analyzed. The operation difficulty can be measured in many ways, such as the size of the initial color difference of the color block (the smaller the color difference, the greater the operation difficulty), the number of user adjustments (the more adjustments, the greater the difficulty), etc. For example, for the purple-blue color block pair with a small initial color difference, the user needs to adjust it many times to make the colors on both sides look consistent, which indicates that the operation difficulty of this color block pair is high. Based on the operation time and difficulty, an operation difficulty score is generated for each color block pair. For example, a color block pair with an operation time of more than 10 seconds and an adjustment number of more than 5 times is rated as high difficulty (such as score 3), an operation time of 5-10 seconds and an adjustment number of 3-5 times is rated as medium difficulty (such as score 2), and an operation time of less than 5 seconds and an adjustment number of less than 3 times is rated as low difficulty (such as score 1).

[0068] Track the dragging trajectory of the user in the process of adjusting the color block. For example, when the user adjusts the red color block, the trajectory moves slowly from left to right, and there are multiple pauses and small adjustments in the middle position. This indicates that the user has a certain degree of uncertainty in the color perception of red, or is cautious in finding a suitable color match. By analyzing the length, direction, and pause points of the dragging trajectory, we can understand the user's color perception characteristics and operation habits. At the same time, observe the user's selection order and adjustment frequency of different colors to determine whether the user has a color preference. For example, if the user always selects a blue-related color block for adjustment first, and the adjustment time on the blue color block is relatively short, it implies that the user has a higher recognition of blue, or is more inclined to start color matching from blue. Combining these dragging trajectories and color preference information, generate an operation feature vector for each color block pair, such as [operation difficulty score, dragging trajectory length, number of pause points, color preference coefficient (such as a quantitative value of the preference for a certain color)], which is used to generate target color block recognition information later, providing more information about color vision status from the perspective of user operation behavior.

[0069] The user operation characteristics are processed based on the physiological characteristic information to generate the target color block identification information. Considering the age factor in the physiological characteristic information of the target user (such as the user age is 55 years old), the user operation characteristics are adjusted according to the relationship between age and color vision changes. With age, color vision sensitivity decreases, especially the ability to distinguish blue and green. If the user shows a high degree of difficulty in the operation of the blue-green color block pair (such as the operation difficulty score is 3), and the age factor indicates that the color vision has declined in this color area, then when generating the target color block identification information, the evaluation weight of the degree of color vision abnormality of the color block pair is appropriately increased. For example, originally based on the model prediction and operation characteristic analysis, the blue-green color block pair was judged as mild color vision abnormality (red-green color vision axis) with a confidence score of 0.7, but considering the age factor, it was adjusted to moderate color vision abnormality (blue-green color vision axis), and the confidence score was adjusted to 0.8, indicating that after considering the influence of age, it is more inclined to believe that the user has a more obvious color vision problem in this color area, and the confidence in this judgment is enhanced.

[0070] If the user suffers from a systemic disease (such as diabetes), the operation characteristics are analyzed according to the potential impact of the disease on color vision. Diabetes affects retinal microvessels, which in turn affects color vision. Assuming that the user shows some abnormalities in the operation of the red-orange color block pair (such as long operation time and complex dragging trajectory), combined with the history of diabetes, further analyze whether these operation characteristics are consistent with the color vision changes related to the disease. If so, when generating the target color block identification information, it is clearly pointed out that the color vision abnormality of the color block pair is related to diabetes, and relevant annotations are added to the color vision abnormality classification information (such as "red-orange color block pair color vision abnormality, affected by diabetes"). At the same time, the confidence score is adjusted according to the severity of the disease and the research data on the impact on color vision. For example, if the diabetes condition is not well controlled and has a greater impact on color vision, the confidence score for the color vision abnormality classification of the color block pair is increased from 0.7 to 0.9, emphasizing the reliability of the judgment so that it can be taken seriously in subsequent diagnosis or evaluation, and provide doctors or professionals with more comprehensive color vision status information, and make comprehensive judgments and decisions based on the user's physiological characteristics and disease history.

[0071] S107, processing the target color block identification information to generate target color difference identification information.

[0072] In one embodiment, the target color block identification information is processed to generate the average color difference prediction value of the color block pair and the majority category information in the color vision abnormality classification information. Assume that the target color block identification information contains the color difference prediction values ​​of multiple color block pairs, such as the color difference prediction value of the red-green color block pair is 120nm, the color difference prediction value of the blue-yellow color block pair is 100nm, the color difference prediction value of the orange-purple color block pair is 80nm, etc. Calculate the average value of the color difference prediction values ​​of these color block pairs, that is, (120+100+80) / 3=100nm, and use it as the average color difference prediction value of the color block pair. This average color difference prediction value can reflect the user's average perception of color differences in the overall color block selection process, and provide a basic reference for subsequent preliminary color difference prediction. For example, if the average color difference prediction value is large, it means that the user's ability to distinguish colors is relatively weak, and it is necessary to further consider whether there is color vision abnormality in subsequent analysis.

[0073] For color vision abnormality classification information, assume that the classification results of multiple color block pairs are: red-green color block pair is mild color vision abnormality (red-green color vision axis), blue-yellow color block pair is normal, and orange-purple color block pair is mild color vision abnormality (blue-purple color vision axis). Among these classification results, mild color vision abnormality appears the most times (2 times), so it is determined that the majority category in the color vision abnormality classification information is mild color vision abnormality. By determining the majority category, the main tendency of the user's color vision abnormality can be quickly understood, providing a qualitative reference for preliminary color difference prediction. For example, if the majority category is mild color vision abnormality, then when generating preliminary color difference prediction information, the features and parameters related to mild color vision abnormality can be focused on, and the specific color difference of the user in this color vision state can be further analyzed.

[0074] The average color difference prediction value of the color block pair and the majority category information in the color vision abnormality classification information are processed to generate preliminary color difference prediction information. According to the calculated average color difference prediction value of the color block pair (such as 100nm mentioned above) and the determined majority category (mild color vision abnormality), the preliminary color difference prediction information is generated and adjusted. If the average color difference prediction value is close to the typical color difference range corresponding to mild color vision abnormality (assuming that the typical color difference range of mild color vision abnormality on the red-green color vision axis is 80-120nm), and the majority category is mild color vision abnormality, then it is preliminarily judged that the user is at a mild abnormal level in overall color vision, and the main affected color range is concentrated in the red-green area. At this time, the preliminary color difference prediction information can be set to have a certain degree of color difference in the red-green area, and the degree of color difference is close to the average color difference prediction value (such as about 100nm). At the same time, considering the classification of other color block pairs (such as the blue-yellow color block pair is normal, and the orange-purple color block pair is mild color vision abnormality (blue-purple vision axis)), it is also necessary to pay attention to the color vision of the blue-purple area, and appropriately mention the potential color difference problem in the preliminary color difference prediction information, but the weight is lower than that of the red-green area.

[0075] With reference to the theoretical knowledge and clinical experience related to color vision, the preliminary color difference prediction information is further optimized. For example, according to the color vision opposition theory, red-green and blue-yellow are the two main color perception axes. When color vision abnormality occurs on one axis, it will have a certain impact on the other axis. If it is determined that the user has mild color vision abnormality on the red-green axis, then in the preliminary color difference prediction information, it is possible to consider appropriately adjusting the color difference prediction range on the blue-yellow axis. Even if the blue-yellow color block pair is classified as normal, its color difference monitoring range can be appropriately expanded (such as from the normal 30-50nm to 40-60nm) to more comprehensively evaluate the user's color vision condition. In addition, combined with the influence of different age groups, genders, eye disease history and other factors on color vision in clinical experience, if the user is an elderly person with a history of mild cataracts, the color vision sensitivity of this group of people in the blue-purple area will decrease. Therefore, in the preliminary color difference prediction information, further emphasis is placed on the color vision of the blue-purple area, and the color difference prediction parameters of this area are appropriately adjusted to make it more in line with the actual situation of the user.

[0076] The preliminary color difference prediction information is processed based on the user operation characteristics to generate the target color difference recognition information. The target recognition color difference information is used to characterize the existence of any color difference information of red, orange, yellow, green, cyan, blue, and purple for the target user. The influence of the operation time and operation difficulty information in the user operation characteristics on the preliminary color difference prediction information is analyzed. Assuming that the user has a long operation time (such as more than 10 seconds) and a high operation difficulty (such as more than 5 adjustments) when selecting the red-green color block pair, it indicates that the user has great difficulty in distinguishing the color of the color block pair. Combined with the preliminary color difference prediction information (such as it has been determined that there is a certain degree of color difference in the red-green area), the evaluation of the degree of color vision abnormality in the area is further improved. For example, the color difference prediction value of the red-green area is adjusted from about 100nm to about 110nm, and the confidence of the color vision abnormality classification in the area is increased (such as from 0.7 to 0.8), indicating that the user's color vision problem in this area is more obvious. For color block pairs with shorter operation time and lower difficulty (such as blue-yellow color block pairs), if their preliminary color difference prediction information appears normal and the user operation characteristics also support this judgment, then the original color difference prediction and classification results can be maintained, but the monitoring frequency or weight of this area can be reduced, and more attention will be paid to the color area corresponding to the color block pairs that are difficult to operate.

[0077] Observe the dragging trajectory and color preference information of the user in the process of adjusting the color block, and further refine the target color difference recognition information. If the user's trajectory is complex and unstable when adjusting the red color block (such as multiple dragging back and forth, with many pause points), this indicates that the user has a large uncertainty in the color perception of red. Combined with the preliminary color difference prediction information, the color vision abnormality classification in the red area is further refined into a specific type of color vision abnormality (such as red weakness), and clearly pointed out in the target color difference recognition information. At the same time, according to the user's color preference (such as the user always selects the blue-related color block for adjustment first and the adjustment time is short), if the color difference prediction of the blue-yellow area in the preliminary color difference prediction information is close to the upper limit of the normal range, considering the user's better recognition of blue, the color difference prediction value of the blue-yellow area can be appropriately adjusted downward (such as from 40-60nm to 35-55nm) to make it more in line with the user's actual color vision performance. By integrating these user operation feature information, the final target color difference recognition information is generated. This information can more accurately reflect the user's color vision status in different color areas, including specific color difference values, color vision abnormality types and corresponding confidence levels, providing more valuable reference for application scenarios such as clinical diagnosis and professional qualification assessment.

[0078] The server changes the previous detection range that was mainly red, green or yellow and blue, and divides it according to the color nm number. It measures seven colors: red, orange, yellow, green, cyan, blue and purple, and comprehensively diagnoses the human eye's recognition of each color, so as to accurately judge the human eye's color vision ability. Breaking through the limitation of only giving qualitative conclusions in the past, it can display specific objective data such as the minimum recognizable color difference (such as 2nm) within a specific color range (such as red 635nm-700nm), and define it as the minimum difference that the human eye can recognize for a single color, so as to achieve the refinement of color vision detection and accurately evaluate the degree of color vision abnormality. Two color blocks are displayed, and the color blocks are made to be consistent by dragging the color bar. This method is fast and accurate, not affected by user cognitive differences, and can effectively shorten the detection time and improve the detection efficiency. It only takes 5 minutes per person to test 7 colors. Setting the detection background to a variety of life scenes (static or dynamic scenes such as driving, sports, roads, etc.) is more in line with the actual visual environment of life, and can find color vision abnormalities missed in hospital detection, helping users understand the color vision problems they face in life.

[0079] In one embodiment, if Figure 2 As shown, the present application also provides a device for color vision testing, comprising:

[0080] The acquisition module 201 is used to acquire the background picture information to be checked, the target user information, the background adjustment information to be checked, the medical record information of the target user, the preset color difference recognition model and the training sample set, wherein the background picture information to be checked is a dynamic life background or a static life background, and the background adjustment information to be checked is used to represent the background information customized by the target user;

[0081] The processing module 202 is used to process the background image information to be checked and the background adjustment information to be checked to generate a color block selection influencing factor vector; process the medical record information of the target user to generate a user color block adjustment factor; process the target user information based on the color block selection influencing factor vector and the user color block adjustment factor to generate preset color block information and color block selection information of the target user; process the preset color difference recognition model based on the training sample set to generate a target color difference recognition model; process the preset color block information and the color block selection information of the target user based on the target color difference recognition model to generate target color block recognition information; process the target color block recognition information to generate target color difference recognition information, wherein the target recognition color difference information is used to characterize the existence of any color difference information of red, orange, yellow, green, cyan, blue and purple for the target user.

[0082] Each embodiment in the present application is described in a related manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the method, electronic device, electronic device, and readable storage medium embodiment for evaluating color vision test, since they are basically similar to the above-mentioned method embodiment for color vision test, the description is relatively simple, and the relevant parts can be referred to the partial description of the above-mentioned method embodiment for color vision test.

Claims

1. A method for color vision testing, characterized in that: include: Obtaining background image information to be checked, target user information, background adjustment information to be checked, medical record information of the target user, a preset color difference recognition model and a training sample set, wherein the background image information to be checked is a dynamic life background or a static life background, and the background adjustment information to be checked is used to characterize the background information customized by the target user; Processing the to-be-checked background image information and the to-be-checked background adjustment information to generate a color block selection influencing factor vector; Processing the medical record information of the target user to generate a user color block adjustment factor; Processing the target user information based on the color block selection influencing factor vector and the user color block adjustment factor to generate preset color block information and color block selection information of the target user; Processing the preset color difference recognition model based on the training sample set to generate a target color difference recognition model; Processing the preset color block information and the color block selection information of the target user based on the target color difference recognition model to generate target color block recognition information; The target color block identification information is processed to generate target color difference identification information, wherein the target identification color difference information is used to characterize any color difference information of red, orange, yellow, green, cyan, blue, and purple that exists for the target user.

2. The method according to claim 1, characterized in that Processing the background image information to be checked and the background adjustment information to be checked to generate a color block selection influencing factor vector includes: Performing feature extraction processing on the background image information to be checked to generate color feature information, brightness feature information and texture feature information; Processing the background adjustment information to be checked to generate a characteristic adjustment amplitude value; The color feature information, the brightness feature information and the texture feature information are processed based on the feature adjustment amplitude value to generate a color block selection influencing factor vector.

3. The method according to claim 1, characterized in that Processing the medical record information of the target user to generate a user color block adjustment factor includes: Processing the medical record information of the target user to generate eye disease characteristic information and physiological characteristic information; Processing the eye disease characteristic information and the physiological characteristic information to generate a color vision influencing factor and weight information matching the color vision influencing factor; Processing the color vision influencing factor and the weight information matching the color vision influencing factor to generate a color block adjustment value; processing the color block adjustment value to generate a user color block adjustment factor; The method includes a calculation formula for obtaining a color block adjustment value, and the calculation formula is: Where n represents the number of major factor categories affecting color vision, m represents the number of subdivided factors under each major factor category, p represents the number of special factors related to eye diseases, i represents the index of the major factor category affecting color vision, j represents the index of the subdivided factors under each major factor category, and F ij represents the color vision influencing factor of the jth subdivision factor in the i-th main factor, W ij represents the weight of the jth subdivision factor in the i-th main factor, k represents the special factor index related to eye diseases, G k represents the impact factor of the kth special factor, H k represents the weight of the kth special factor, I t represents the influencing factor of the tth lifestyle factor, J t represents the weight of the tth lifestyle factor, and K represents the individual difference coefficient.

4. The method according to claim 3, characterized in that The target user information is processed based on the color block selection influencing factor vector and the user color block adjustment factor to generate preset color block information and color block selection information of the target user, including: Processing the target user information to generate color vision information to be detected; Processing the color vision information to be detected based on the color block selection influencing factor vector to generate a color selection range value; Processing the color selection range value based on the user color block adjustment factor to generate preset color block information; The preset color block information is processed to generate color block selection information of a target user.

5. The method according to claim 1, characterized in that The preset color difference recognition model is processed based on the training sample set to generate a target color difference recognition model, including: Process the training sample set to generate a training set and a validation set; Get any number of data features in the training set; Generating a sampling ratio based on the number of each data feature in the training set; Sampling the training set based on the sampling ratio to generate a preset number of sampling features; Processing based on any data feature and each sampling feature to generate multiple data groups, wherein each data group includes a preset number of data samples, and at least one data sample includes identification information; Training the preset color difference recognition model based on data samples in the multiple data groups to generate a trained color difference recognition model; Processing the trained color difference recognition model based on the verification set to generate a verification result; If the data sample containing identification information in the verification result represents a risk factor that affects the user's color difference recognition, the trained color difference recognition model is used as a target color difference recognition model.

6. The method according to claim 3, characterized in that The preset color block information and the color block selection information of the target user are processed based on the target color difference recognition model to generate target color block recognition information, including: Processing the preset color block information and the color block selection information of the target user based on the target color difference recognition model to generate a predicted value of the color difference between each color block pair, color vision abnormality classification information, and a confidence score matching the color vision abnormality classification information; Processing the predicted value of the color difference between each color block pair, the color vision abnormality classification information, and the confidence score matching the color vision abnormality classification information to generate a user operation feature; The user operation feature is processed based on the physiological feature information to generate target color block recognition information.

7. The method according to claim 6, characterized in that Processing the target color block identification information to generate target color difference identification information includes: Processing the target color block recognition information to generate an average color difference prediction value of the color block pair and majority category information in the color vision abnormality classification information; Processing the average color difference prediction value of the color block pair and the majority category information in the color vision abnormality classification information to generate preliminary color difference prediction information; The preliminary color difference prediction information is processed based on the user operation characteristics to generate target color difference identification information.

8. A device for color vision testing, characterized in that: The device comprises: An acquisition module is used to acquire background image information to be checked, target user information, background adjustment information to be checked, medical record information of the target user, a preset color difference recognition model and a training sample set, wherein the background image information to be checked is a dynamic life background or a static life background, and the background adjustment information to be checked is used to represent the background information customized by the target user; A processing module is used to process the background image information to be checked and the background adjustment information to be checked to generate a color block selection influencing factor vector; process the medical record information of the target user to generate a user color block adjustment factor; process the target user information based on the color block selection influencing factor vector and the user color block adjustment factor to generate preset color block information and color block selection information of the target user; process the preset color difference recognition model based on the training sample set to generate a target color difference recognition model; process the preset color block information and the color block selection information of the target user based on the target color difference recognition model to generate target color block recognition information; process the target color block recognition information to generate target color difference recognition information, wherein the target recognition color difference information is used to characterize the existence of any color difference information of red, orange, yellow, green, cyan, blue and purple for the target user.

9. An electronic device, characterized in that: include: a first processor; and a memory for storing executable instructions of the first processor; Wherein, the first processor is configured to perform the color vision testing method according to any one of claims 1 to 7 by executing the executable instructions.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by the second processor, the method for color vision testing according to any one of claims 1 to 7 is implemented.

Images