A method and system for detecting contrast sensitivity function

Abstract

This invention belongs to the field of visual perception technology and provides a method and system for detecting contrast sensitivity functions. It generates a target set by combining a hybrid contrast sensitivity function model of a parameter domain particle model and a stimulus domain grid model. Subjects are then tested using this target set, and the hybrid model can be updated based on the subjects' responses. The updated model then generates new target sets for further testing. This approach employs a hybrid contrast sensitivity function model and a resampling strategy. The hybrid contrast sensitivity function model simultaneously provides parameter estimation functionality in the parameter domain and records the information accumulation of each stimulus point target in the stimulus-target domain. This helps generate more effective test targets for subjects and improves the method's adaptability to individual differences. The resampling strategy improves test resolution without increasing computational load, ensuring high accuracy under various conditions.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of visual perception, and particularly relates to a contrast sensitivity function detection method and system. BACKGROUND

[0002] The contrast sensitivity function (CSF) is a basic index for measuring visual function, which represents the ability of the human visual system to detect luminance differences in different spatial frequency ranges.

[0003] CSF detection is different from traditional visual tests, which evaluate the smallest letters or symbols that a person can recognize under high contrast. CSF detection can provide a more comprehensive assessment of visual quality, especially under low-contrast conditions in the real world. CSF detection is particularly useful for detecting subtle changes in visual function caused by various eye diseases and neurological diseases such as glaucoma, cataract, diabetic retinopathy, amblyopia, and multiple sclerosis.

[0004] Traditional CSF testing methods, such as the Pelli-Robson visual acuity chart, the Westphal-Young visual acuity chart, and the computer-based sinusoidal grating test, often take a long time, require patients to cooperate under various testing conditions, and may have limited precision and repeatability. Moreover, many existing methods cannot efficiently provide complete CSF curves, which limits their clinical application value in time-constrained clinical environments or patients with poor compliance. In addition, CSF testing methods often have the defect of fixed contrast sensitivity spatial resolution, which also limits the upper limit of testing accuracy. SUMMARY

[0005] To overcome the above-mentioned defects of the prior art, the present application provides a contrast sensitivity function detection method and system, which comprises:

[0006] A hybrid contrast sensitivity function model composed of a parameter domain particle model and a stimulus domain grid model is used to generate a set of test stimuli, and the subject is tested with the generated set of test stimuli.

[0007] The subject's responses to each stimulus in the set of test stimuli are collected, and the parameter domain particle model and the stimulus domain grid model are updated based on the responses.

[0008] If the current number of tests does not reach the preset maximum number of tests, the hybrid contrast sensitivity function model is used to generate a new set of test stimuli and perform a new round of testing until the current number of tests reaches the maximum number of tests.

[0009] Specifically, the method further comprises initializing the mixed contrast sensitivity function model, and the method of initializing the mixed contrast sensitivity function model comprises:

[0010] The initialization of the parameter domain particle model is completed by determining the number of particles, determining the parameters of the particles, and determining the weights of the particles.

[0011] The first probability that the subject can clearly see a specific target and the second probability that the subject cannot see the specific target are set, and the initialization of the stimulation domain grid model is completed by determining the grid range, determining the resolution, and determining the values in the grid.

[0012] Specifically, the set of to-be-detected stimulus point targets is generated by:

[0013] Using the current stimulation domain grid model, the entropy reduction of all possible stimulus point targets is calculated, and the index of the maximum entropy reduction stimulus point target is obtained based on the calculation result.

[0014] The set of to-be-detected stimulus point targets required for the next test is generated by the current stimulation domain grid model and the index of the maximum entropy reduction stimulus point target.

[0015] Preferably, the set of to-be-detected stimulus point targets required for the next test is generated by the current stimulation domain grid model and the index of the maximum entropy reduction stimulus point target, comprising:

[0016] The probability value of the first stimulus point target obtained based on the current stimulation domain grid model and the index of the maximum entropy reduction stimulus point target is obtained.

[0017] If the probability value is the median of the first probability and the second probability, search in the set of target contrast not less than the index contrast and the set of target contrast not greater than the index contrast, obtain the second stimulus point target whose probability value satisfies not less than the preset maximum probability threshold and the third stimulus point target whose probability value satisfies not greater than the preset minimum probability threshold, determine the highest contrast by the second stimulus point target, and determine the lowest contrast by the third stimulus point target, and then generate the set of to-be-detected stimulus point targets based on the spatial frequency and target contrast of the first stimulus point target, the highest contrast, and the lowest contrast.

[0018] If the probability value is closer to the first probability than the second probability, search in the set of target contrast not less than the index contrast, obtain the second stimulus point target whose probability value satisfies not less than the preset maximum probability threshold, determine the highest contrast by the second stimulus point target, and then generate the set of to-be-detected stimulus point targets based on the spatial frequency and target contrast of the first stimulus point target and the highest contrast.

[0019] If the probability value is closer to the second probability than the first probability, search in a set of index contrasts in which the target contrast is not greater than the index contrast, obtain a third stimulus point target whose probability value satisfies a preset minimum probability threshold, determine a minimum contrast through the third stimulus point target, and generate a set of stimulus point targets to be detected based on the spatial frequency and target contrast of the first stimulus point target and the minimum contrast.

[0020] Further, the generating of the set of stimulus point targets to be detected based on the updated hybrid contrast sensitivity function model comprises:

[0021] The updated stimulus domain grid model is filtered through the maximum probability threshold and the minimum probability threshold, and a candidate stimulus point target sample set is obtained based on the filtering result;

[0022] An approximate sample set is obtained through the updated parameter domain particle model, and the step of calculating the entropy reduction of all possible stimulus point targets is performed based on the candidate stimulus point target sample set and the approximate sample set.

[0023] Preferably, the way of updating the hybrid contrast sensitivity function model comprises resampling, and the updating of the parameter domain particle model and the stimulus domain grid model based on the answer content comprises:

[0024] The parameter domain particle model is updated based on the answer content, the number of effective particles is calculated through the parameter domain particle model updated by the answer content, and when the number of effective particles is less than a preset determination threshold, the hybrid contrast sensitivity function model is updated through resampling.

[0025] Specifically, the method of updating the hybrid contrast sensitivity function model through resampling comprises:

[0026] The sample weights in the parameter domain particle model updated by the answer content are taken as probabilities, the set elements are taken as sampling objects, non-replacement sampling is performed the same number of times as the number of particles, a sample set with equal weights and a weight sum of 1 is obtained;

[0027] Each sample in the sample set is taken as a mean value, a Gaussian distribution with the same number of particles as the number of particles is constructed through a preset perturbation covariance matrix, each Gaussian distribution is sampled, and the parameter domain particle model updated by resampling is obtained based on the sampling result.

[0028] Further, the method of updating the hybrid contrast sensitivity function model through resampling comprises:

[0029] Based on the parameter domain particle model updated by the resampling method, the stimulus domain grid model after updating is directly calculated and obtained through the total probability formula.

[0030] Or, the answer content is determined as the subject can see, the subject cannot see or the subject does not know, the corresponding stimulus domain grid likelihood function is determined based on the divided answer content, and the stimulus domain grid model, the stimulus domain grid likelihood function and the parameter domain particle model updated by the resampling method are used to obtain the stimulus domain grid model updated by the resampling method.

[0031] The application further provides a contrast sensitivity function detection system, which comprises:

[0032] A test module is configured to generate a set of to-be-detected stimulus point targets by using a hybrid contrast sensitivity function model composed of a parameter domain particle model and a stimulus domain grid model, and perform a round of test on the subject by using the currently generated set of to-be-detected targets.

[0033] An updating module is configured to collect the answers of the subject to each stimulus point in the target set, and update the parameter domain particle model and the stimulus domain grid model based on the answer content.

[0034] A circulation module is configured to make the test module generate a set of to-be-detected stimulus point targets again by using the hybrid contrast sensitivity function model and perform a new round of test until the current test number reaches the maximum test number, when the current test number does not reach the maximum test number.

[0035] The application further provides a computer readable storage medium, wherein the readable storage medium stores computer instructions, and the instructions are executed by a processor to realize the steps of the contrast sensitivity function detection method.

[0036] The application has at least the following beneficial effects:

[0037] The scheme provided by the application can generate a set of to-be-detected stimulus point targets more quickly by updating the hybrid contrast sensitivity function model, avoid the redundant steps in the traditional test method, reduce the time and resources required for the test, better adapt to the perception characteristics of different subjects by dynamically updating the model, reduce errors and ensure comprehensive coverage of all stimulus points to be detected, and improve the sensitivity and accuracy of the test.

[0038] Further, the scheme provided by the present application ensures that the selection of the to-be-tested stimulus point set is based on the optimal matching of probability and contrast by selecting and generating the stimulus point target, and the resampling step can gradually gather the parameter domain samples to the parameter range with higher value along with the updating of the model, thereby improving the accuracy of the contrast sensitivity function detection without increasing the sample amount, and ensuring that the visual ability of the subject is accurately reflected.

[0039] In addition, the present scheme also removes the information-saturated stimulus point target, realizes the effect of reducing the calculation amount of the entropy reduction calculation step without reducing the accuracy, and the saved calculation amount can also allow the entropy reduction calculation step to be distributed to more parameter domain samples, thereby improving the accuracy of the entropy reduction calculation step.

[0040] Therefore, the present application provides a contrast sensitivity function detection method and system, the scheme provided by the present application adopts a mixed contrast sensitivity function model and a resampling strategy, the mixed contrast sensitivity function model has the functions of parameter estimation of the parameter domain and recording the information accumulation of each stimulus point target in the stimulus point target domain, which is helpful to generate more effective test targets for the subjects, improves the adaptability of the method to individual differences, and the resampling strategy improves the resolution of the test without increasing the calculation amount, thereby ensuring that the test can maintain high accuracy under various conditions. BRIEF DESCRIPTION OF DRAWINGS

[0041] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed in the embodiment description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0042] Figure 1 A method flowchart of a contrast sensitivity function detection method provided for embodiment 1 is shown in the figure.

[0043] Figure 2 A framework flowchart of the contrast sensitivity function detection method is shown in the figure.

[0044] Figure 3 A stimulus point target example diagram composed of different spatial frequencies and sensitivities is shown in the figure.

[0045] Figure 4 A stimulus point grid and CSF curve example diagram is shown in the figure.

[0046] Figure 5 A method flowchart for generating a to-be-detected stimulus point set is shown in the figure.

[0047] Figure 6A block diagram of a module structure of a contrast sensitivity function detection system provided for the example 2.

[0048] Reference numerals

[0049] 10 - test module; 11 - calculation unit; 12 - generation unit; 20 - update module; 30 - loop module. DETAILED DESCRIPTION

[0050] The technical solutions in the embodiments of the present application will be apparently and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments in the present application, all the other embodiments obtained by a person of ordinary skill in the art without creative work belong to the scope of protection of the present application.

[0051] In the following, various embodiments of the present application will be described more fully. The present application can have various embodiments, and adjustments and changes can be made therein. However, it should be understood that there is no intention to limit various embodiments of the present application to the specific embodiments disclosed herein, but the present application should be understood to cover all adjustments, equivalents and / or alternatives falling within the spirit and scope of various embodiments of the present application.

[0052] In the following, the term "include" or "may include" used in various embodiments of the present application indicates the presence of the disclosed functions, operations or elements, and does not limit the addition of one or more functions, operations or elements. In addition, as used in various embodiments of the present application, the terms "include", "have" and their synonyms only intend to indicate the presence of a specific feature, number, step, operation, element, component or combination of the foregoing, and should not be understood as first excluding the presence or addition of one or more other features, numbers, steps, operations, elements, components or combinations of the foregoing or the possibility of one or more features, numbers, steps, operations, elements, components or combinations of the foregoing.

[0053] In various embodiments of the present application, the expression "or" or "at least one of A or / and B" includes any combination of the listed terms or all combinations thereof. For example, the expression "A or B" or "at least one of A or / and B" can include A, can include B, or can include both A and B.

[0054] The expressions used in the various embodiments of the present application, such as "first", "second", etc., can modify various constituent elements in the various embodiments, but can not limit the corresponding constituent elements. For example, the above expressions do not limit the order and / or importance of the elements. The above expressions are only for the purpose of distinguishing one element from other elements. For example, the first user device and the second user device indicate different user devices, although both are user devices. For example, without departing from the scope of the various embodiments of the present application, a first element can be referred to as a second element, and likewise, a second element can be referred to as a first element.

[0055] It should be noted that in the present application, unless otherwise explicitly specified and defined, the terms "mounting", "connecting", "fixing" and the like should be understood in a broad sense, for example, can be fixedly connected, can be detachably connected, or integrally connected; can be mechanically connected, or can be electrically connected; can be directly connected, or can be indirectly connected through an intermediate medium; can be internal communication of two elements. For those of ordinary skill in the art, the specific meaning of the above terms in the present application can be understood according to the specific circumstances.

[0056] In the present application, those of ordinary skill in the art need to understand that the terms indicating the orientation or positional relationship herein are based on the orientation or positional relationship shown in the drawings, which is only for the purpose of facilitating the description of the present application and simplifying the description, and is not intended to indicate or imply that the device or element referred to must have a particular orientation, be constructed and operated in a particular orientation, and therefore cannot be understood as a limitation on the present application.

[0057] The terms used in the various embodiments of the present application are only for the purpose of describing specific embodiments and are not intended to limit the various embodiments of the present application. As used herein, the singular form is intended to include the plural form as well, unless the context clearly dictates otherwise. Unless otherwise defined, all terms used herein, including technical and scientific terms, have the same meaning as commonly understood by those of ordinary skill in the art to which the various embodiments of the present application belong. The terms such as those defined in a generally used dictionary will be interpreted to have the same meaning as the contextual meaning in the relevant technical field and will not be interpreted to have an idealized or overly formal meaning, unless clearly defined in the various embodiments of the present application.

[0058] Embodiment 1

[0059] The present embodiment proposes a contrast sensitivity function detection method, please refer to Figures 1-4 , the method comprises:

[0060] S100: generating a set of test stimuli by a hybrid contrast sensitivity function model composed of a parameter domain particle model and a stimulus domain grid model, and testing the subject by the generated set of test stimuli.

[0061] Please refer to Figure 3 , Figure 3 The figure shows a test stimulus set containing 9 different parameters, each fringe picture in the figure represents a test stimulus, the parameters of the fringe image are determined by spatial frequency and sensitivity, and the test stimulus can be arranged according to the parameter type and size represented by the coordinate axis.

[0062] Specifically, the hybrid contrast sensitivity function model includes a model composed of a parameter domain particle model and a stimulus domain grid model Ms t,k (i f ,i s ) combination, the hybrid contrast sensitivity function model can identify the curve distribution of the hybrid contrast sensitivity function, please refer to Figure 4 , the parameter domain particle model and the stimulus domain grid model express the same probability distribution from different angles, which can be understood from the perspective of a computer as using a composite data structure to represent the curve distribution of the hybrid contrast sensitivity function;

[0063] It should be noted that the parameter domain particle model includes a test of the subject defined as a given set of test stimuli, which can obtain the subject's answer to whether the set of test stimuli is clear, and the updated CSF curve parameter distribution of the kth test stimulus in the tth test is defined as P t,k (θ CSF )=P t,k (γ max ,f max ,β,δ), which represents the joint probability distribution composed of four parameters θ CSF ; P0(θ CSF ) represents the CSF parameter distribution at time 0, i.e. the initial parameter distribution;

[0064] Since the four-dimensional parameter represented by θ CSF is a continuous distribution, in order to estimate the distribution by computer algorithm, it needs to be expressed as a discrete distribution, the method proposed in this embodiment can use a set of particles with weights to represent the multi-dimensional continuous probability distribution, where w is the weight corresponding to the particle, N Φ is the number of particles, represents the i-th particle with weight, so as to realize the use of samples to represent the population, and the strategy of particle filtering is consistent;

[0065] The stimulation domain grid model includes a stimulation point target grid Ms updated by a defined t-th k-th stimulation point target t,k (i fs ,i c ),Ms t,k (i fs ,i c ) is a two-dimensional matrix in form, where the row index i fs is a uniformly discrete spatial frequency index, referred to as a spatial frequency index, and the column index i c is a uniformly discrete sensitivity index, referred to as a sensitivity index, so the index [i fs ,i c ] of the matrix represents an index of a discrete stimulation point target, and the index ranges as 0≤i fs ≤i fs_max , 0≤i c ≤i c_max . Given the index, the value of the corresponding index, i.e., Ms t,k (i fs ,i c ), can be obtained, which represents the probability that the stimulation corresponding to the index can be seen by the subject. Ms0(i fs ,i c ) is the stimulation grid at time 0, which can also be referred to as the initial stimulation grid.

[0066] The method proposed in the embodiment also defines functions f i_to_v (i fs )、s i_to_v (i c )、f v_to_i (f s )、s v_to_i (c),wherein the functions f i_to_v (i fs )、s i_to_v (i c ) represent functions of converting from the spatial frequency index or the sensitivity index to the spatial frequency value or the sensitivity value, and f v_to_i (f s )、s v_to_i (c) represent corresponding inverse functions, i.e., functions of converting from the stimulation point target value to the respective index.

[0067] Preferably, before performing step S100, the method proposed in the embodiment will initialize the hybrid contrast sensitivity function model in advance, and the initialization manner of the parameter domain particle model can include but is not limited to determining the number of particles, determining the parameters of the particles, and determining the weights of the particles, and the initialization manner of the stimulation domain grid model can include but is not limited to determining the range of the grid, determining the resolution, and determining the values in the grid.

[0068] The setting process of particles in the parameter domain is a process of uniformly dividing the four-dimensional parameter space. The formula for uniformly dividing the four-dimensional parameter space includes:

[0069]

[0070] wherein, represents a Cartesian product; i represents a natural number index; Δ represents a discrete increment, that is, a hyperparameter, and the smaller the Δ is, the more samples there will be, and the higher the resolution of the estimation will be, so it needs to be set according to the actual hardware adopted; the superscripts low and high respectively represent the upper and lower bounds of the corresponding parameter, which can be set according to the curve range of the human eye's mixed contrast sensitivity function, and the particle set can be determined, and N Φ is determined.

[0071] The initialization of the stimulus domain grid model can also be set according to the upper and lower bounds of the spatial frequency of the human visual ability ( and ), and the upper and lower bounds of the contrast ( low and max ), and then sampling is performed within the upper and lower bounds. Different from the parameter domain, the initialization of the stimulus domain grid needs to first take the logarithm of the value with base 10, and then uniformly divide it. In the present embodiment, the number of row indexes i f of the stimulus domain grid is defined as N fs , the number of column indexes i c is N c , and the total number of grids is N MS =N fs ·N c .

[0072] To reduce the amount of calculation and storage space, the method proposed in the present embodiment can introduce a probability saturation mechanism to avoid the problem of slow updating speed. Exemplarily, when the first probability that the subject can clearly see a specific visual target is set to 100, and the second probability that the subject cannot see the specific visual target is set to 0, once the probability exceeds 100 or 0 in the updating process, it is set to the corresponding boundary value. Since there is no information, it is uncertain whether the visual target sample of different stimulus points can be seen by the subject, at this time, it can be initialized as M fs s0(i c )=50.

[0073] After the initialization of the parameter domain and the stimulus domain is completed, the number of joint sample domains composed of the parameter domain and the stimulus domain N U =N Φ ·N MS .

[0074] S200: Collect the answers of the subject to each stimulus point in the target set, and update the parameter domain particle model and the stimulus domain grid model based on the answer content.

[0075] Specifically, in the dynamic detection process, the stimulus point target set is constantly updated according to the answers of the subjects, and whether the stimulus point target set to be detected has been generated is determined, that is, whether the stimulus point is updated, specifically to ensure that the detection process can cover all the parameter ranges to be evaluated, while avoiding repeated detection of the areas that have been fully understood, thereby improving the detection efficiency; and updating the model according to the answers of the subjects to the stimulus point target set can more accurately describe the visual function state of the subjects, and provide more accurate guidance for subsequent detection.

[0076] In the embodiment, when the subject has not answered all the stimulus points in the current target set, the subsequent steps of updating the model and generating a new target set are not performed.

[0077] S300: If the current test number does not reach the preset maximum test number, the mixed contrast sensitivity function model is used to generate the stimulus point target set to be detected again and perform a new round of test, until the current test number reaches the maximum test number.

[0078] It should be noted that, in order to control the detection time and avoid excessive fatigue of the subjects, the method proposed in the embodiment needs to set a maximum detection number, and the detection process is ensured to be completed within a reasonable time range through the judgment of whether the maximum detection number is reached, while ensuring the reliability and effectiveness of the detection results.

[0079] Specifically, please refer to Figure 5 , the step of generating the stimulus point target set to be detected in the step S100 includes:

[0080] S110: Calculate the entropy reduction of all possible stimulus point targets using the current stimulus domain grid model, and obtain the index of the maximum entropy reduction stimulus point target based on the calculation result.

[0081] S120: Generate the stimulus point target set to be detected for the next test using the current stimulus domain grid model and the index of the maximum entropy reduction stimulus point target.

[0082] In the embodiment, the step S120 can obtain the probability value of the first stimulus point target based on the current stimulus domain grid model and the index of the maximum entropy reduction stimulus point target, and generate the stimulus point target set to be detected according to the judgment result of the probability value; specifically, the step S120 can obtain the index I opt of the maximum entropy reduction stimulus point target based on the step S110, and the probability value of the first stimulus point target based on the current stimulus domain grid model and the index I fs_opt of the maximum entropy reduction stimulus point target. c_opt[and the stimulus domain raster model Ms updated after t tests] t (i fs i c (This refers to obtaining the set of stimulus targets indexes required for the next test.) Each stimulus target in this set has an index of I. fs_opt That is, the spatial frequency of each stimulus target in the stimulus target set is consistent, but the sensitivity changes.

[0083] Specifically, [i] can be defined fs_opt i c ≥i c_opt [] represents the spatial frequency index of a given stimulus point target, and derives all sensitivity indices i that satisfy the given conditions. c_max ≥i c ≥i c_opt The constructed index set, in this embodiment, when Ms t (i fs_opt i c_opt The value is 50, the median of the first and second probabilities, meaning that when no information is obtained about the visibility of the target at that stimulus point, the target contrast is not less than the set [i] of index contrast. fs_opt i c ≥i c_opt The set of [i] whose target contrast is not greater than the index contrast. fs_opt i c ≤i c_opt Search within these two sets to find the set that satisfies Ms. t (i fs_opt i c_upper )≥T Hmax The index, i.e., the supremum i c_upper Find what satisfies Ms. t (i fs_opt i c_lower )≤T Hmin The index, i.e., the infimum i c_lower Obtain the set of test stimulus target indexes I, consisting of three indices. t+1 ={ fs_opt i c_lower +1>, fs_opt i c_opt >, fs_opt i c_upper -1>};

[0084] When Ms t (i fs_opt i c_opt When the value is greater than 50, the target has acquired visible information, therefore it falls within the set [i] where the target contrast is not less than the index contrast. fs_opt ​​​i c ≥i c_opt Search within [the area] to obtain the upper bound i c_upper This allows us to obtain a set of three indexes for the stimulus targets to be tested, I. t+1 ={ fs_opt i c_opt >, fs_opt ,(i c_upper -i c_opt ) / 2>, fs_opt i c_upper -1>};

[0085] When Ms t (i fs_opt i c_opt When the value is less than 50, the target has acquired visible information, therefore it falls within the set [i] where the target contrast is not greater than the index contrast. fs_opt i c ≤i c_opt Search within [i] to obtain the infimum i c_lower This allows us to obtain a set of three indexes for the stimulus targets to be tested, I. t+1 ={ fs_opt i c_lower +1>, fs_opt ,(i c_opt -i c_lower ) / 2>, fs_opt i c_opt >}.

[0086] In this embodiment, step S200 can filter the updated stimulus domain grid model using preset maximum probability thresholds and minimum probability thresholds, and obtain a candidate stimulus point target sample set based on the filtering results; it can also obtain an approximate sample set using the updated parameter domain particle model, and continue to execute the step of calculating the entropy reduction of all possible stimulus point targets as described in step S110 based on the candidate stimulus point target sample set and the approximate sample set, so as to continue generating a stimulus point target set to be detected.

[0087] Specifically, step S200 can be based on the stimulus domain raster model Ms updated after t test responses. t (i fs i c ), thus obtaining the index of the candidate stimulus point target sample set. This removes stimulus targets with high information content and retains stimulus targets with low information content, thereby reducing the computational cost of subsequent entropy reduction calculations and allocating the saved computational cost to more parameter domain samples. It is clear whether the stimulus targets with high information content can be seen by the subjects, while it is unclear whether the retained stimulus targets with low information content can be seen by the subjects.​​​​​​

[0088] If Ms t (i fs i c The value of ) is within the preset maximum probability threshold T Hmax and minimum probability threshold T Hmin If the value is between [value] and [value], it indicates that the stimulus point has low visual information content and can be selected as a candidate stimulus sample point; if Ms [value] t (i fs i c The value of ) is not in T Hmax and T Hmin If the sample is between these values, it is not considered a candidate stimulus target. After obtaining the candidate stimulus target sample set, the number of samples in the candidate stimulus target sample set can be recorded as N. Η .

[0089] Accordingly, step S200 can also be based on the updated parameter domain particle model after t tests. Obtain an approximate sample set In this embodiment, the approximate sample set Θ t Used to approximate Φ t This facilitates the subsequent calculation of entropy in the entropy reduction step using Monte Carlo integration, significantly reducing the computational load.

[0090] It should be noted that the method proposed in this embodiment uses N Θ For the sample size, The samples in the sample are used as the sampling objects, and sampling without replacement is performed using sample weights as probabilities. In this embodiment, the number of samples is... Where N UΘ N represents the joint number of samples in the parameter domain and the stimulus point target domain used for entropy reduction calculation. UΘ As a hyperparameter, the information content of the parameter domain distribution increases and the distribution becomes more concentrated with the number of updates. Therefore, by setting the number of samples, N can be made more flexible. Θ As the number of updates increases, the resolution of the parameter domain increases, which in turn improves the accuracy of the estimation.

[0091] In this embodiment, the method of updating the hybrid contrast sensitivity function model in step S200 includes resampling. Step S200 can update the parameter domain particle model based on the answer content, calculate the effective particle number through the parameter domain particle model updated by the answer content, and update the hybrid contrast sensitivity function model by resampling when the effective particle number is less than the preset judgment threshold.

[0092] Specifically, step S200 can be achieved by updating the parameter domain particle model based on the response from the k-th stimulus point target after t tests. determine whether resampling is needed for the parameter-domain model at present;

[0093] In this embodiment, the threshold T eff and calculate the effective particle number When e t,k <T eff , it is determined that resampling is needed. Resampling can remove low-weight particles as non-efficient data, and the particles with higher weights and the particles near them are duplicated several times, so as to occupy a larger proportion in the sample set, thereby ensuring the invariability of the particle number, improving the representativeness and effectiveness of the sample, better approximating the real distribution, reducing the sample degradation phenomenon, and improving the estimation accuracy.

[0094] Based on the parameter-domain particle model obtain the resampled parameter-domain particle model After that, the values of all samples will change, and all weights are the same

[0095] Specifically, the method for updating the hybrid contrast sensitivity function model by the resampling manner includes:

[0096] The sample weights in the parameter-domain particle model updated by the answer content are taken as probabilities, the set elements are taken as sampling objects, N

[0097] Each sample in the sample set is taken as a mean value, a same number of Gaussian distributions as the number of particles is constructed by using a preset disturbance covariance matrix, each Gaussian distribution is sampled, and the parameter-domain particle model updated by the resampling manner is obtained based on the sampling result;

[0098] In addition, based on the stimulus-domain grid model and the parameter-domain particle model updated by the resampling manner, the stimulus-domain grid model updated by the resampling manner is directly calculated and obtained by using the total probability formula;

[0099] Or, the answer content is determined to be seen by the subject, not seen by the subject or unknown to the subject, the corresponding stimulus-domain grid likelihood function is determined based on the divided answer content, and the stimulus-domain grid model updated by the resampling manner is obtained by using the stimulus-domain grid model, the stimulus-domain grid likelihood function and the parameter-domain particle model updated by the resampling manner.

[0100] Resampling can make the parameter-domain particle model The sample weights in the parameter-domain particle model updated by the answer content are taken as probabilities, the set elements are taken as sampling objects, N Φ times of non-replacement sampling is performed, and a sample set with equal weights and a weight sum of 1 is obtained Then, a disturbance covariance Each sample in is taken as the mean, Σ d is constructed as a covariance matrix, N Φ Gaussian distributions, the i-th Gaussian distribution is The above N Φ Gaussian distributions are sampled to obtain the resampled results By adding Gaussian perturbations to the resampled particles, the resolution of the parameter domain can be improved.

[0101] It should be noted that in the actual resampling process, sampling a large number of Gaussian distributions is inefficient. The method proposed in the embodiment can sample the Gaussian distribution Ν((0, 0, 0, 0), Σ d ) N Φ times before the resampling program runs to obtain a perturbation set with N Φ samples. During the running of the resampling program, the corresponding sample in and the corresponding perturbation in the perturbation set are added to improve the calculation efficiency.

[0102] Although the method proposed in the embodiment can directly calculate the updated stimulus domain grid model according to the current completed parameter domain particle model, this calculation method involves the total probability formula and requires integration in four dimensions of the parameter domain, which is computationally intensive. Therefore, preferably, the method proposed in the embodiment can also update the stimulus domain grid model according to the subject's answer, that is, according to the answer content r t,k of the k-th stimulus point index t,k-1 of the t-th test and the current stimulus domain grid model Ms fs (i c , i t,k ), a new stimulus domain grid model Ms fs (i c , i t,k ) is updated and generated. The method for updating the stimulus domain grid model specifically includes:

[0103] After the subject is given the stimulus point index I fs = [i c , i v ], if the answer content is determined to be visible to the subject, then on the same spatial frequency, the probability of other stimulus point indexes with lower sensitivity than the stimulus point index being visible will all increase. At this time, the updated stimulus domain grid model

[0104] If the answer content is determined as unable to be seen by the subject or unknown to the subject, on the same frequency space, the probability of other stimulus point targets being unable to be seen is increased, and when the answer content is determined as unable to be seen by the subject, an updated stimulus domain grid model can be obtained accordingly

[0105] When the answer content is determined as unknown to the subject, an updated stimulus domain grid model can be obtained accordingly

[0106] wherein V v (i c ), V i (i c ), and V u (i c ) are the stimulus domain grid likelihood functions of the stimulus point targets under the conditions of being able to see, unable to see, and unknown, respectively, and specifically,

[0107]

[0108]

[0109] wherein d v represents the update rate, d v is smaller, the update is slower, and the anti-disturbance ability is higher; and σ V represents the update influence domain, σ V is smaller, the range of the measured value influence is larger, that is, the sensitivity is higher; it should be noted that in the actual update process, the value range of Ms t,k (i fs , i c ) is [0, 100], so it is necessary to truncate the data according to the value range, force the probability value to be saturated, so as to avoid the problems of numerical overflow and loss of meaning of the calculation result.

[0110] Therefore, the embodiment proposes an innovative contrast sensitivity function detection method, by adding the resampling step, the parameter domain sample gradually gathers in the parameter range with higher value along with the model update, thereby improving the precision of the contrast sensitivity function detection without increasing the sample amount, by adopting the mixed contrast sensitivity function model, the model has the parameter estimation function of the parameter domain and the function of recording the information accumulation of each stimulus point target of the stimulus point target domain, by removing the information saturated stimulus point target, the calculation amount of the entropy reduction calculation step can be reduced without reducing the precision, in addition, the calculation amount saved by the method proposed in the embodiment can also allocate the entropy reduction calculation step to more parameter domain samples, thereby improving the precision of the entropy reduction calculation step.

[0111] Embodiment 2

[0112] The embodiment proposes a contrast sensitivity function detection system for implementing the contrast sensitivity function detection method proposed in embodiment 1, please refer to Figure 6 , which comprises:

[0113] A test module 10 is configured to generate a set of to-be-detected stimulus points by a hybrid contrast sensitivity function model composed of a parameter domain particle model and a stimulus domain grid model, and perform a round of test on the subject by the currently generated set of to-be-detected stimuli;

[0114] An update module 20 is configured to collect the answers of the subject to each stimulus point in the set of stimuli, and update the parameter domain particle model and the stimulus domain grid model based on the answer content;

[0115] A cycle module 30 is configured to make the test module 10 generate a set of to-be-detected stimulus points again by the hybrid contrast sensitivity function model and perform a new round of test until the current test number reaches the maximum test number, when the current test number does not reach the preset maximum test number.

[0116] Preferably, the system proposed in the embodiment will initialize the hybrid contrast sensitivity function model in advance, the initialization mode of the parameter domain particle model can include but is not limited to determining the number of particles, determining the parameters of the particles and determining the weights of the particles, and the initialization mode of the stimulus domain grid model can include but is not limited to determining the grid range, determining the resolution and determining the values in the grid;

[0117] The setting process of the particles in the parameter domain is a process of uniformly dividing the four-dimensional parameter space, and the formula for uniformly dividing the four-dimensional parameter space includes:

[0118]

[0119] Wherein, represents the Cartesian product; i represents the natural number index; Δ represents the discrete increment, that is, the hyperparameter, the smaller the Δ is, the more the samples will be, and the higher the estimated resolution will be, so it needs to be set according to the actual hardware used; The superscripts low and high represent the upper and lower bounds of the corresponding parameters, which can be set according to the curve range of the mixed contrast sensitivity function of the human eye, and N Φ The weight of each particle is initialized as

[0120] The initialization of the stimulus domain grid model can also be performed according to the upper and lower bounds of the spatial frequency ( and ), the upper and lower bounds of the contrast (c low and c max), and then sampling within the upper and lower bounds. Different from the parameter domain, the initialization of the stimulation domain grid needs to first take the logarithm of the value with base 10, and then uniformly divide. In the embodiment, the row index i of the stimulation domain grid is specifically defined as i = 0, 1, 2, …, N - 1, the column index i of the stimulation domain grid is specifically defined as i = 0, 1, 2, …, N - 1, the number of the row index i is N, the number of the column index i is N, and the total number of the grid is N = N N. f fs c c MS fs c ;

[0121] To reduce the amount of calculation and storage space, the system proposed in the embodiment can introduce a probability saturation mechanism to avoid the problem of slow updating speed. Exemplarily, when the first probability that a subject can clearly see a specific visual target is set to 100, and the second probability that the subject cannot see the specific visual target is set to 0, once the probability exceeds 100 or 0 in the updating process, it is set to the corresponding boundary value. Since there is no information, it is uncertain whether the visual target sample of the different stimulation points can be seen by the subject, at this time, it can be initialized as Ms0(i fs ,i c ) = 50.

[0122] After the initialization of the parameter domain and the stimulation domain is completed, the number of the joint sample domain composed of the parameter domain and the stimulation domain is N = N N. U Φ MS

[0123] Specifically, the test module 10 includes:

[0124] The calculation unit 11 is configured to calculate the entropy reduction of all possible stimulation point visual targets by using the current stimulation domain grid model, and obtain the index of the maximum entropy reduction stimulation point visual target based on the calculation result.

[0125] The generation unit 12 is configured to generate the set of to-be-detected stimulation point visual targets required for the next test by using the current stimulation domain grid model and the index of the maximum entropy reduction stimulation point visual target.

[0126] In the embodiment, the generation unit 12 can obtain the probability value of the first stimulation point visual target based on the current stimulation domain grid model and the index of the maximum entropy reduction stimulation point visual target, and generate the set of to-be-detected stimulation point visual targets according to the judgment result of the probability value. Specifically, the generation unit 12 can obtain the maximum entropy reduction stimulation point visual target index I = [i opt ,i fs_opt ,i c_opt ] obtained by the calculation unit 11 and the stimulation domain grid model Ms t (i fs ,i c ​​​​​​​​​), to obtain the index set of the next test stimulus point target Each index of the stimulus point target in the set is I fs_opt That is, the spatial frequency of each stimulus point target in the stimulus point target set is consistent, and the sensitivity changes.

[0127] Specifically, [i fs_opt , i c ] can be defined as the index set of the spatial frequency of the given stimulus point target, and all the sensitivity indexes i c_opt that satisfy the condition are obtained c_max ≥i c ≥i c_opt The index set is composed of, in the embodiment, when Ms t (i fs_opt , i c_opt ) is the median value of the first probability and the second probability, i.e., the visibility of the stimulus point target does not obtain any information, in the set [i fs_opt , i c ≥i c_opt ] of the target contrast not less than the index contrast and the set [i fs_opt , i c ≤i c_opt ] of the target contrast not greater than the index contrast, search to find the index that satisfies Ms t (i fs_opt , i c_upper ) ≥ T Hmax , i.e., the supremum i c_upper , find the index that satisfies Ms t (i fs_opt , i c_lower ) ≤ T Hmin , i.e., the infimum i c_lower , obtain the index set I t+1 ={<i fs_opt , i c_lower +1>, <i fs_opt , i c_opt >, <i fs_opt , i c_upper -1>} of the test stimulus point target composed of 3 indexes.

[0128] When the value of Ms t (i fs_opt , i c_opt ) is greater than 50, the target obtains information that can be seen, and therefore, in the set [i fs_opt , i c ≥i c_opt ] of the target contrast not less than the index contrast, search to obtain the supremum i c_upperThis allows us to obtain a set of three indexes for the stimulus targets to be tested, I. t+1 ={ fs_opt i c_opt >, fs_opt ,(i c_upper -i c_opt ) / 2>, fs_opt i c_upper -1>};

[0129] When Ms t (i fs_opt i c_opt When the value is less than 50, the target has acquired visible information, therefore it falls within the set [i] where the target contrast is not greater than the index contrast. fs_opt i c ≤i c_opt Search within [i] to obtain the infimum i c_lower This allows us to obtain a set of three indexes for the stimulus targets to be tested, I. t+1 ={ fs_opt i c_lower +1>, fs_opt ,(i c_opt -i c_lower ) / 2>, fs_opt i c_opt >}.

[0130] In this embodiment, the update module 20 can filter the updated stimulus domain grid model using preset maximum probability thresholds and minimum probability thresholds, and obtain a candidate stimulus point target sample set based on the filtering results; it can also obtain an approximate sample set using the updated parameter domain particle model, and continue to execute the step of calculating the entropy reduction of all possible stimulus point targets as described in step S110 based on the candidate stimulus point target sample set and the approximate sample set, so as to continue to generate a stimulus point target set to be detected.

[0131] Specifically, the update module 20 can update the stimulus domain raster model Ms based on the responses to t tests. t (i fs i c ), thus obtaining the index of the candidate stimulus point target sample set. This removes stimulus targets with high information content and retains stimulus targets with low information content, thereby reducing the computational cost of subsequent entropy reduction calculations and allocating the saved computational cost to more parameter domain samples. It is clear whether the stimulus targets with high information content can be seen by the subjects, while it is unclear whether the retained stimulus targets with low information content can be seen by the subjects.

[0132] If Ms t (i fs ​​​​​​i c The value of ) is within the preset maximum probability threshold T Hmax and minimum probability threshold T Hmin If the value is between [value] and [value], it indicates that the stimulus point has low visual information content and can be selected as a candidate stimulus sample point; if Ms [value] t (i fs i c The value of ) is not in T Hmax and T Hmin If the sample is between these values, it is not considered a candidate stimulus target. After obtaining the candidate stimulus target sample set, the number of samples in the candidate stimulus target sample set can be recorded as N. Η .

[0133] Accordingly, the update module 20 can also update the parameter domain particle model based on the results of t tests. Obtain an approximate sample set In this embodiment, the approximate sample set Θ t Used to approximate Φ t This facilitates the subsequent calculation of entropy in the entropy reduction step using Monte Carlo integration, significantly reducing the computational load.

[0134] It should be noted that the system proposed in this embodiment is based on N. Θ For the sample size, The samples in the sample are used as the sampling objects, and sampling without replacement is performed using sample weights as probabilities. In this embodiment, the number of samples is... Where N UΘ N represents the joint number of samples in the parameter domain and the stimulus point target domain used for entropy reduction calculation. UΘ As a hyperparameter, the information content of the parameter domain distribution increases and the distribution becomes more concentrated with the number of updates. Therefore, by setting the number of samples, N can be made more flexible. Θ As the number of updates increases, the resolution of the parameter domain increases, which in turn improves the accuracy of the estimation.

[0135] In this embodiment, the update module 20 updates the hybrid contrast sensitivity function model by means of resampling. The update module 20 can update the parameter domain particle model based on the answer content, calculate the effective particle number through the parameter domain particle model updated by the answer content, and update the hybrid contrast sensitivity function model by resampling when the effective particle number is less than the preset judgment threshold.

[0136] Specifically, the update module 20 can update the parameter domain particle model based on the response from the k-th stimulus point target after t tests. Determine whether resampling of the parameter domain model is currently necessary;

[0137] In this embodiment, the threshold T can be set eff and the effective particle number is calculated When e t,k <T eff , it is judged that resampling is needed, the resampling can remove low-weight particles as non-efficient data, and the particles with higher weights and the particles near them are copied several times, so as to occupy a larger proportion in the sample set, thereby ensuring the invariability of the particle number, improving the representativeness and effectiveness of the sample, better approximating the real distribution, reducing the sample degradation phenomenon, and improving the estimation accuracy.

[0138] Based on the parameter domain particle model After resampling, the parameter domain particle model is obtained After that, the values of all samples will change, and all weights are the same

[0139] Specifically, the method for updating the hybrid contrast sensitivity function model by the resampling manner includes:

[0140] The sample weights in the parameter domain particle model updated by the answer content are taken as probabilities, the set elements are taken as sampling objects, N

[0141] Each sample in the sample set is taken as a mean value, a same number of Gaussian distributions as the number of particles is constructed by using a preset disturbance covariance matrix, and each Gaussian distribution is sampled, and the parameter domain particle model updated by the resampling manner is obtained based on the sampling result;

[0142] In addition, based on the stimulus domain grid model and the parameter domain particle model updated by the resampling manner, the stimulus domain grid model updated by the resampling manner is directly calculated and obtained by using the total probability formula;

[0143] Or, the answer content is determined to be seen by the subject, not seen by the subject or unknown to the subject, the corresponding stimulus domain grid likelihood function is determined based on the divided answer content, and the stimulus domain grid model updated by the resampling manner is obtained by using the stimulus domain grid model, the stimulus domain grid likelihood function and the parameter domain particle model updated by the resampling manner.

[0144] The resampling can make the parameter domain particle model The sample weights in the parameter domain particle model updated by the answer content are taken as probabilities, the set elements are taken as sampling objects, N Φ times of non-replacement sampling is performed, and a sample set with equal weights and a weight sum of 1 is obtained Then, a disturbance covariance is set Each sample in the sample set is taken as a mean value, and a same number of Gaussian distributions as the number of particles is constructed by using the disturbance covariance dN Φ gaussian distributions, the i-th gaussian distribution is sampling the N Φ gaussian distributions to obtain a resampled result By adding a Gaussian disturbance to the resampled particles, the resolution of the parameter domain can be improved.

[0145] It should be noted that in the actual process of resampling, the efficiency of sampling a large number of Gaussian distributions is low. The system proposed in the embodiment can sample the Gaussian distribution N d ((0, 0, 0, 0), Σ Φ ) N Φ times before the resampling program runs, to obtain a disturbance set with N t,k samples, and the corresponding sample in and the corresponding disturbance in the disturbance set are added during the sampling of the resampling program, thereby improving the calculation efficiency.

[0146] Although the system can directly calculate the updated stimulus domain grid model according to the current completed parameter domain particle model, this calculation method involves the total probability formula and requires integration in four dimensions of the parameter domain, which is computationally intensive. Therefore, preferably, the system proposed in the embodiment can also update the stimulus domain grid model for the subject's answer, that is, according to the k-th stimulus point index corresponding to the answer content r t,k of the t-th test and the current stimulus domain grid model M t,k-1 (i fs , i c ), a new stimulus domain grid model M t,k (i fs , i c ) is updated and generated, and the way to update the stimulus domain grid model specifically includes:

[0147] After the subject is given the stimulus point index I t,k =[i fs , i c ], if the answer content is determined to be seen by the subject, then on the same spatial frequency, the probability of being seen by other stimulus point indices with lower sensitivity than the stimulus point index will increase, and at this time the updated stimulus domain grid model

[0148] If the answer content is determined to be unable to be seen by the subject or the subject does not know, then on the same frequency space, the probability of being unable to be seen by other stimulus point indices with higher sensitivity than the stimulus point index will increase, wherein when the answer content is determined to be unable to be seen by the subject, the updated stimulus domain grid model

[0149] When the answer content is determined to be unknown by the subject, an updated stimulation domain grid model can be obtained accordingly

[0150] wherein, V v (i c ), V i (i c ), V u (i c ) are stimulation domain grid likelihood functions of the stimulation point visual target under the conditions of seeing, not seeing and not knowing, respectively, and specifically,

[0151]

[0152] wherein, d v represents the updating rate, the smaller d v , the slower the updating, and the higher the anti-disturbance ability; σ V represents the updating influence domain, the smaller σ V , the greater the range of the measured value influence, that is, the greater the sensitivity; it should be noted that in the actual updating process, the value range of Ms t,k (i fs , i c ) is [0, 100], so it is necessary to truncate the data according to the value range, force the probability value to be saturated, so as to avoid the problem of numerical overflow and the loss of meaning of the calculation result.

[0153] Embodiment 3

[0154] The embodiment also proposes a computer readable storage medium, which stores computer instructions, and the instructions are executed by a processor to realize the steps of the contrast sensitivity function detection method proposed in the above embodiment 1.

[0155] Note that the computer-readable medium can include the transitory, non-transitory, removable, and non-removable mediums used to store information and accessed by computers. The computer-readable medium includes, but is not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disc read-only memory (CD-ROM), digital versatile discs (DVDs) or other optical storage, magnetic cassettes, magnetic tapes, magnetic disk storage, quantum memory, graphene-based memory medium, or other magnetic storage devices, or any other non-transitory medium that can be used to store the information and accessed by computers. According to the definition herein, the computer-readable medium does not include the transitory medium, such as modulated data signals and carrier waves.

[0156] To sum up, the present application provides a contrast sensitivity function detection method and system, the scheme provided by the present application adopts a mixed contrast sensitivity function model and a resampling strategy, the mixed contrast sensitivity function model has the functions of parameter estimation in the parameter domain and information accumulation of each stimulus target in the stimulus target domain, which helps to generate more effective test targets for the subjects and improves the adaptability of the method to individual differences, and the resampling strategy improves the resolution of the test without increasing the calculation amount, ensuring that the test can maintain high accuracy under various conditions.

[0157] The above only describes the preferred embodiments of the present application and is not intended to limit the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A method of detecting a contrast sensitivity function, characterized by, The method comprises: A set of target stimulus points is generated using a hybrid contrast sensitivity function model consisting of a parameter-domain particle model and a stimulus-domain grid model. The subjects are then tested using this generated target set. The parameter-domain particle model includes a weighted particle set. ,in, The weights of the corresponding particles, The number of particles, Indicates the first A weighted particle, This represents the extreme value of the contrast sensitivity function. This represents the frequency corresponding to the extreme value of sensitivity. This represents the bandwidth of the contrast sensitivity function. Represents low-frequency cutoff sensitivity. Indicates that it has been defined The first A stimulus point target; the stimulus domain grid model includes defined... sequence Stimulus target grid updated per stimulus target ,in, For a two-dimensional matrix, row index Spatial frequency index, column index For sensitivity index; collecting the answers of the subject to each stimulus point in the set of targets, updating the parametric particle model based on the answer content, calculating the effective particle number by the parametric particle model updated by the answer content, and updating the hybrid contrast sensitivity function model by resampling when the effective particle number is less than a preset determination threshold; if the current test number does not reach the preset maximum test number, generating a set of target stimuli to be detected again by the hybrid contrast sensitivity function model and performing a new round of test until the current test number reaches the maximum test number.

2. The contrast sensitivity function detection method according to claim 1, further comprising initializing the hybrid contrast sensitivity function model, and the method of initializing the hybrid contrast sensitivity function model comprises: initializing the parametric particle model by determining the number of particles, the parameters of the particles, and the weights of the particles; setting a first probability that the subject can clearly see a specific target and a second probability that the subject cannot see the specific target, and initializing the stimulus domain grid model by determining the grid range, the resolution, and the values in the grid.

3. The method of claim 2, wherein the contrast sensitivity function is determined by: The generated set of target stimuli to be detected comprises: calculating the entropy reduction of all possible target stimuli using the current stimulus domain grid model, and obtaining the index of the maximum entropy reduction target stimulus based on the calculation result; generating the set of target stimuli to be detected required for the next test by the current stimulus domain grid model and the index of the maximum entropy reduction target stimulus.

4. The method of claim 3, wherein the contrast sensitivity function is determined by: The generated set of target stimuli to be detected required for the next test by the current stimulus domain grid model and the index of the maximum entropy reduction target stimulus comprises: obtaining the probability value of the first target stimulus based on the current stimulus domain grid model and the index of the maximum entropy reduction target stimulus; if the probability value is the median of the first probability and the second probability, searching in the set of target contrasts not less than the index contrast and the set of target contrasts not greater than the index contrast to obtain a second target stimulus whose probability value satisfies not less than a preset maximum probability threshold and a third target stimulus whose probability value satisfies not greater than a preset minimum probability threshold, determining the highest contrast by the second target stimulus, determining the lowest contrast by the third target stimulus, and then generating the set of target stimuli to be detected based on the spatial frequency and target contrast of the first target stimulus, the highest contrast, and the lowest contrast; if the probability value is closer to the first probability than to the second probability, searching in the set of target contrasts not less than the index contrast to obtain a second target stimulus whose probability value satisfies not less than a preset maximum probability threshold, determining the highest contrast by the second target stimulus, and then generating the set of target stimuli to be detected based on the spatial frequency and target contrast of the first target stimulus and the highest contrast. If the probability value is closer to the second probability than the first probability, search in a set of index contrasts that are not greater than the target contrast, obtain a third stimulus point target whose probability value satisfies a preset minimum probability threshold, determine the lowest contrast through the third stimulus point target, and then generate a set of stimulus point targets to be detected based on the spatial frequency and target contrast of the first stimulus point target, the lowest contrast, and the like.

5. The method of claim 4, wherein the contrast sensitivity function is determined by: The method for generating the set of stimulus point targets to be detected again through the updated hybrid contrast sensitivity function model comprises the following steps: Filter the updated stimulus domain grid model through the maximum probability threshold and the minimum probability threshold, and obtain a candidate stimulus point target sample set based on the filtering result; Obtain an approximate sample set through the updated parameter domain particle model, and perform the step of calculating the entropy reduction of all possible stimulus point targets based on the candidate stimulus point target sample set and the approximate sample set.

6. The method of claim 1, wherein, The method for updating the hybrid contrast sensitivity function model through resampling comprises the following steps: Take the sample weight in the parameter domain particle model updated by the answer content as a probability, take the set element as a sampling object, perform non-replacement sampling of the same number of times as the number of particles, and obtain a sample set with equal weights and a weight sum of 1; Take each sample in the sample set as a mean value, construct a Gaussian distribution with the same number of particles as the number of particles through a preset perturbation covariance matrix, sample each Gaussian distribution, and obtain the parameter domain particle model updated by resampling based on the sampling result.

7. The method of detecting a contrast sensitivity function according to claim 6, wherein, The method for updating the hybrid contrast sensitivity function model through resampling further comprises the following steps: Directly calculate and obtain the updated stimulus domain grid model through the total probability formula based on the parameter domain particle model updated by resampling; Or, determine the answer content as being seen by the subject, not seen by the subject, or unknown to the subject, determine the corresponding stimulus domain grid likelihood function based on the divided answer content, and obtain the updated stimulus domain grid model through the stimulus domain grid model and the stimulus domain grid likelihood function.

8. A contrast sensitivity function detection system characterized by, The system comprises: The testing module is used to generate a set of target stimuli to be detected using a hybrid contrast sensitivity function model consisting of a parameter domain particle model and a stimulus domain grid model, and to conduct a test on the subject using the currently generated set of target stimuli; the parameter domain particle model includes a weighted particle set. ,in, The weights of the corresponding particles, The number of particles, Indicates the first A weighted particle, This represents the extreme value of the contrast sensitivity function. This represents the frequency corresponding to the extreme value of sensitivity. This represents the bandwidth of the contrast sensitivity function. Represents low-frequency cutoff sensitivity. Indicates that it has been defined The first A stimulus point target; the stimulus domain grid model includes defined... sequence Stimulus target grid updated per stimulus target ,in, For a two-dimensional matrix, row index Spatial frequency index, column index For sensitivity index; An updating module configured to collect the answers of the subject to each stimulus point in the target set, update the parameter domain particle model based on the answer content, calculate the effective particle number through the parameter domain particle model updated by the answer content, and update the hybrid contrast sensitivity function model through resampling when the effective particle number is less than a preset determination threshold; A loop module configured to make the test module generate a set of stimulus point targets to be detected again through the hybrid contrast sensitivity function model and perform a new round of testing when the current number of tests does not reach a preset maximum number of tests, until the current number of tests reaches the maximum number of tests.

9. A computer-readable storage medium, characterized in that, A computer readable storage medium having stored thereon computer instructions, which, when executed by a processor, implement the steps of the contrast sensitivity function detection method according to any one of claims 1-7.

Images