An intelligent eye-protection lighting control system

Abstract

The present invention discloses an intelligent eye protection lighting control system, which relates to the technical field of intelligent lighting control. The control system comprises: a physiological analysis module, which calculates the user's pupil diameter physiological index based on pre-processed multi-source data through the Daugman algorithm and marks a fatigue warning to obtain the user's fatigue state; and performs sight point processing and area division according to the user's long-term fixed line of sight area through spatial coordinate mapping and clustering algorithm to generate focus area coordinates; an instruction generation module, which fuses the scene initial light strategy with the focus area coordinates through a weighted fusion method, combines the user's vision parameters to form light parameters, and analyzes them through the NSGA-II algorithm to obtain lighting control instructions; an execution module, which transmits and executes the lighting control instructions through a wireless communication method, monitors the user's visual physiological response in real time, and obtains feedback data; the present invention predicts environmental changes through a Markov model to avoid lighting lagging behind environmental changes, thereby improving the foresight and adaptability of lighting.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent lighting control, in particular to an intelligent eye-protection lighting control system. Background Art

[0002] With the growing demand for healthy lighting, smart lighting technology has developed rapidly. Traditional smart lighting systems rely on a single or limited set of parameters, such as ambient light intensity and infrared sensing, to control light on / off and brightness through pre-set programs. These systems have gradually evolved to include systems that can adjust lighting parameters based on simple scene modes, such as reading and resting. These technologies have, to a certain extent, met users' basic demand for intelligent lighting and are driving the transformation of the lighting industry from traditional to intelligent.

[0003] Current intelligent lighting systems generally lack personalized adaptation, making it difficult to dynamically optimize the lighting environment based on user physiological parameters such as vision characteristics, light sensitivity, and color perception. Traditional solutions rely solely on ambient lighting or simple preset modes, failing to precisely adjust the lighting based on the user's real-time visual fatigue, work scenario requirements (such as reading and screen usage), and eye health data (such as refractive index and glare sensitivity). This leads to a mismatch between lighting parameters and individual eye protection needs, which can exacerbate long-term problems such as visual fatigue and dry eyes, compromising visual comfort and health protection. Summary of the Invention

[0004] In view of the above existing problems, the present invention is proposed.

[0005] Therefore, the present invention provides an intelligent eye protection lighting control system to solve the problem that the lighting environment is difficult to adapt to personalized eye protection needs.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0007] The present invention provides an intelligent eye-protection lighting control system, which includes: a data acquisition module, which collects multi-source data in real time for preprocessing, and analyzes it through a Markov model to obtain an initial light strategy for the scene; a physiological analysis module, which calculates the user's pupil diameter physiological index and marks a fatigue warning based on the preprocessed multi-source data through the Daugman algorithm, obtains the user's fatigue state, and performs sight point processing and area division according to the user's long-term fixed sight area through spatial coordinate mapping and clustering algorithm to generate focus area coordinates; an instruction generation module, which fuses the scene initial light strategy with the focus area coordinates through a weighted fusion method, combines the user's vision parameters to form light parameters, and analyzes them through the NSGA-II algorithm to obtain lighting control instructions; an execution module, which transmits and executes the lighting control instructions through a wireless communication method, monitors the user's visual physiological response in real time, and obtains feedback data; an optimization module, which dynamically corrects the light parameters using a fuzzy logic algorithm based on the user's fatigue state to obtain optimized light parameters, and generates optimal intelligent eye-protection lighting control parameters through data fusion and an intelligent decision-making algorithm in combination with the feedback data.

[0008] As a preferred solution of the intelligent eye-protection lighting control system of the present invention, the multi-source data includes ambient light intensity data, ambient color temperature data, user pupil diameter data, user blinking frequency data, and data on the user's long-term fixed line of sight area;

[0009] The preprocessing includes standardization, noise removal, filling missing values ​​and outlier processing.

[0010] As a preferred solution of the intelligent eye protection lighting control system of the present invention, wherein: the Markov model is used to analyze and obtain the scene initial light strategy, and the specific steps are as follows:

[0011] Based on the pre-processed historical multi-source data, the Markov model is trained by the maximum likelihood estimation method to obtain a trained Markov model;

[0012] The pre-processed multi-source data is input into the Markov model. Based on the transition probability relationship between different environments and user status data, the trend of ambient light intensity changes in the current scene is predicted, and the corresponding preliminary light strategy parameters for different scenes are analyzed.

[0013] The preliminary lighting strategy parameters are sorted and screened through the hierarchical analysis method to obtain the initial lighting strategy for the scene.

[0014] As a preferred solution of the intelligent eye protection lighting control system of the present invention, wherein: the user's physiological indicators are calculated based on the pre-processed multi-source data through the Daugman algorithm and fatigue warning is marked to obtain the user's fatigue status. The specific steps are as follows:

[0015] The pre-processed multi-source data is input into the Daugman algorithm, and the pupil and iris boundaries are extracted through circle detection and bilinear interpolation, and the user's pupil diameter change rate physiological indicator is calculated;

[0016] Based on historical fatigue data and historical normal state data, the fatigue threshold is obtained by optimizing parameters using the support vector machine method.

[0017] The user's pupil diameter change rate physiological indicator and the user's blinking frequency data are compared with the fatigue threshold to obtain the user's fatigue status.

[0018] As a preferred solution of the intelligent eye protection lighting control system of the present invention, wherein: according to the user's long-term fixed line of sight area, the line of sight landing point processing and area division are performed through spatial coordinate mapping and clustering algorithm to generate the focus area coordinates. The specific steps are as follows:

[0019] The area where the user's line of sight is fixed for a long time is converted into a unified standardized coordinate system through a spatial coordinate mapping algorithm;

[0020] The DBSCAN clustering algorithm is used to cluster the areas where the user's gaze is fixed for a long time based on their spatial location similarity, thereby dividing the initial focus area and removing abnormal points and smoothing the boundaries.

[0021] Based on the position of the preliminary focus area in the standardized coordinate system, the coordinates of the focus area are obtained by analysis using the minimum enclosing rectangle method.

[0022] As a preferred solution of the intelligent eye protection lighting control system of the present invention, wherein: the scene initial light strategy and the focus area coordinates are fused by a weighted fusion method, and the light parameters are formed in combination with the user's vision parameters. The specific steps are as follows:

[0023] Based on the scene's initial light strategy data and the focus area coordinates, the principal component analysis method is used to analyze the degree of impact and correlation on the user's visual experience, and the user's visual experience impact weight is obtained. The weighted fusion method is used to fuse them and obtain the light strategy.

[0024] The light strategy is combined with the user's vision parameters, and optimized and matched through an adaptive adjustment algorithm to form light parameters suitable for the user.

[0025] As a preferred solution of the intelligent eye protection lighting control system of the present invention, wherein: the lighting control instructions are obtained by analyzing through the NSGA-II algorithm, the specific steps are as follows:

[0026] Based on light parameters, a multi-objective function of lighting energy saving and visual comfort is set, and a set of lighting control instruction parameter combination populations containing multiple individuals is randomly generated and initialized;

[0027] Perform non-dominated sorting and crowding analysis on the lighting control command parameter combination population through selection, crossover, mutation, merging, updating and elite retention;

[0028] When the multi-objective functions of lighting energy saving and visual comfort converge, the optimal individual is selected from the lighting control instruction parameter combination population to obtain the lighting control instruction.

[0029] As a preferred solution of the intelligent eye protection lighting control system of the present invention, wherein: the lighting control instructions are transmitted and executed through a wireless communication method, the user's visual physiological response is monitored in real time, and feedback data is obtained. The specific steps are as follows:

[0030] Encode the lighting control instructions, convert them into JSON format for wireless communication transmission, and send them to the lighting end through the wireless communication protocol;

[0031] The lighting end receives the lighting control instructions transmitted by wireless communication, decodes them and restores them to the original lighting control instructions, and adjusts and executes the lighting parameters according to the lighting control instructions;

[0032] Monitor the user's visual physiological response data in real time and perform preprocessing to extract feedback data related to the user's visual physiological response.

[0033] As a preferred solution of the intelligent eye protection lighting control system of the present invention, wherein: the light parameters are dynamically corrected using a fuzzy logic algorithm based on the user's fatigue state to obtain optimized light parameters. The specific steps are as follows:

[0034] The user's fatigue state is fuzzified through fuzzy logic algorithm and converted into fuzzy language variables. The fuzzy reasoning method is used to perform logical reasoning based on the fuzzy rule base to obtain the fuzzy conclusion of light parameter adjustment.

[0035] The center of gravity method is used to defuzzify the fuzzy conclusions and convert them into numerical values ​​to obtain the optimized light parameters.

[0036] As a preferred solution of the intelligent eye-protection lighting control system described in the present invention, the optimized light parameters are combined with the feedback data, integrated and processed through weighted average fusion, and in-depth analysis is performed through the use of Bayesian decision making to generate the optimal intelligent eye-protection lighting control parameters.

[0037] The beneficial effects of the present invention are: by predicting environmental changes through the Markov model, it is possible to perceive the evolution trend of factors such as ambient light intensity in advance, adjust the lighting strategy in advance, and avoid lighting lagging behind environmental changes; it improves the foresight and adaptability of lighting, reduces the impact of environmental mutations on lighting effects, creates a more stable and comfortable lighting environment for users, and enhances the user experience; the fuzzy logic algorithm responds to fatigue status, and can convert the user's complex and fuzzy fatigue status into precise light parameter adjustment instructions; it can flexibly adapt the lighting environment according to different fatigue levels, effectively relieve visual fatigue, provide users with personalized and dynamic eye protection lighting solutions, and improve lighting intelligence and user comfort. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0039] Figure 1 Schematic diagram of the intelligent eye-protection lighting control system.

[0040] Figure 2 Schematic diagram generated for the initial lighting strategy for the scene.

[0041] Figure 3 Schematic diagram of user physiological analysis.

[0042] Figure 4 Schematic diagram generated for intelligent eye-protection lighting control parameters. DETAILED DESCRIPTION

[0043] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0044] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0045] Secondly, the term "one embodiment" or "embodiment" herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in various places throughout this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive of other embodiments.

[0046] Reference Figures 1 to 4 , is an embodiment of the present invention, which provides an intelligent eye-protection lighting control system, including the following steps:

[0047] Data acquisition module collects multi-source data in real time for preprocessing.

[0048] Multi-source data includes ambient light intensity data, ambient color temperature data, user pupil diameter data, user blinking frequency data, and data on the area where the user's gaze is fixed for a long time.

[0049] Specifically, the ambient light intensity data is collected by using a light intensity sensor at different time points to obtain the ambient light intensity values ​​of the working area in front of the user, the area around the screen, the area facing the light source, and the area around the user's head; the ambient color temperature data is collected by using a color temperature sensor to analyze the color temperature of the light around the user and extract the color temperature value; the user's pupil diameter data is collected by using a camera to obtain the pupil diameter parameters; the user's blinking frequency data is collected by detecting the user's eyelid opening and closing status through a continuous image frame sequence, and counting the number of blinks through a time window to obtain the user's blinking frequency per unit time; the user's line of sight fixed area data is collected by using a camera to track the user's eye movement trajectory, and the coordinates of the line of sight landing point are accumulated and recorded over time to form the user's line of sight fixed area data.

[0050] Preprocessing includes standardization, noise removal, imputation of missing values, and outlier handling.

[0051] Specifically, the mean and standard deviation of the ambient light intensity data, ambient color temperature data, user pupil diameter data, user blinking frequency data and user line of sight long-term fixed area data are calculated respectively by the Z-score normalization method. The corresponding mean is subtracted from each ambient light intensity data, ambient color temperature data, user pupil diameter data, user blinking frequency data and user line of sight long-term fixed area data point and then divided by the standard deviation to complete the Z-score normalization method standardization processing; for the ambient light intensity data, ambient color temperature data, user pupil diameter data and user blinking frequency data, a fixed-length sliding window containing the current point is selected at each time point, and the ambient light intensity data, ambient color temperature data, user pupil diameter data and user blinking frequency data in the fixed-length sliding window are sorted from small to large, and then the ambient light intensity data, ambient color temperature data, user pupil diameter data and user blinking frequency data in the middle position are taken. The data is used as the new value at the current time point to complete the noise removal using the median filtering method; for all missing positions marked as empty in the user pupil diameter data and the user blink frequency data, the Euclidean distance between the missing position and the non-missing pupil diameter data and the user blink frequency data point is analyzed, the non-missing position with the smallest distance is selected, and the non-missing position is weighted averaged as the filling point to complete the K-nearest neighbor interpolation method for filling the missing position; for the ambient light intensity data, ambient color temperature data, user pupil diameter data and user blink frequency data, the upper quartile and the lower quartile are calculated respectively, and then the interquartile range is calculated. The upper boundary of the outlier is the upper quartile plus one times the interquartile range, and the lower boundary of the outlier is the lower quartile minus one times the interquartile range. The values ​​beyond the boundary are marked as outliers and are removed from the ambient light intensity data, ambient color temperature data, user pupil diameter data and user blink frequency data to complete the outlier processing using the box plot method.

[0052] The initial light strategy of the scene is obtained through analysis of the Markov model.

[0053] Based on the preprocessed historical multi-source data, the Markov model is trained by the maximum likelihood estimation method to obtain a trained Markov model.

[0054] Specifically, ambient light intensity data, ambient color temperature data, user pupil diameter data, user blinking frequency data and user line of sight long-term fixed area data are extracted in sequence from the preprocessed historical multi-source data, and each set of time-synchronized ambient light intensity data is divided into four exemplary intervals: 0 to 200 lux, 200 to 500 lux, 500 to 800 lux, and above 800 lux; the ambient color temperature data is divided into four exemplary intervals: 2500 to 3500 Kelvin, 3500 to 5000 Kelvin, 5000 to 6500 Kelvin, and above 6500 Kelvin; the user pupil diameter data is divided into four exemplary intervals: 2 to 3 mm, 3 to 4 mm, 4 to 5 mm, and above 5 mm; the interval numbers of the corresponding positions of each group in the above three intervals are combined into state labels to form a state label sequence arranged in chronological order. All state transition pairs consisting of two adjacent state labels in the state label sequence are enumerated. The frequency of each type of state transition pair is counted, and the sum of the frequencies of transitions from each state label to all other state labels is recorded. The frequency of transitions from each state label to another state label is divided by the sum of all state label transition frequencies to obtain the transition probability between state labels. The transition probability values ​​between all state labels are then organized to form a state transition probability matrix. Based on the maximum likelihood estimation method, all probability values ​​in the state transition probability matrix are multiplied by joint probabilities to evaluate the fit between the state transition probability matrix and the observed data, thereby verifying the rationality of the Markov model. By dividing the frequency of each state transition pair in the state label sequence by the total transition frequency of the corresponding starting state label, the probability value of the corresponding position in the state transition probability matrix is ​​directly assigned, so that the corresponding joint probability product in the state label sequence reaches the maximum value, thus obtaining the trained Markov model.

[0055] The preprocessed multi-source data is input into the Markov model. Based on the transition probability relationship between different environments and user status data, the changing trend of ambient light intensity in the current scene is predicted, and the corresponding preliminary light strategy parameters under different scenes are analyzed.

[0056] Specifically, the real-time collected data on ambient light intensity, ambient color temperature, user pupil diameter, blink frequency, and fixed gaze area are divided into intervals, and the interval numbers corresponding to the above five types of data at the current moment are extracted in chronological order, and spliced ​​together to form the current state label;

[0057] The light strategy parameter configuration rule table is constructed by collecting ambient light intensity data, ambient color temperature data, user pupil diameter data, user blinking frequency data, and data on the area where the user's gaze is fixed for a long time, and dividing it into multiple intervals. Each combination represents a specific user state and environmental conditions. Historical operation data and experimental test results are analyzed to establish a correlation model between different state combinations and indicators such as user visual comfort and work efficiency. A multi-objective optimization algorithm is then used to match the light intensity and color temperature parameters that can bring the best visual experience to each state combination while meeting energy-saving requirements. Finally, these verified parameter configurations are filled into the corresponding positions of the rule table to form a complete "state-lighting strategy" mapping relationship, ensuring that there is a corresponding lighting setting for each state. As new ambient light intensity data, ambient color temperature data, user pupil diameter data, user blinking frequency data, and data on the area where the user's gaze is fixed for a long time are collected and analyzed, the light strategy parameter configuration rule table will be continuously updated to adapt to different environmental changes and user needs.

[0058] The current state label is input into the Markov model, and all target state labels and transition probability values ​​corresponding to the current state label are obtained from the state transition probability matrix. Based on the state transition probability of the Markov model, a weighted analysis is performed on all possible target states. The typical lighting characteristics of each state and its transition possibility are comprehensively considered, and the expected change direction of the light intensity is finally calculated to obtain the expected trend value of the ambient light intensity change. The expected trend value is divided into four exemplary light intensity levels: 100 to 300 lux, 300 to 500 lux, 500 to 700 lux, and above 700 lux. The current ambient color temperature data interval number, the user pupil diameter data interval number, and the light intensity level are combined into a triplet index. The light strategy parameter setting value matching the corresponding triplet is searched from the light strategy parameter configuration rule table, and the corresponding preliminary light strategy parameters for different scenarios are output.

[0059] The preliminary lighting strategy parameters are sorted and screened through the hierarchical analysis method to obtain the initial lighting strategy for the scene.

[0060] Specifically, the hierarchical structure of the decision problem of the scene initial light strategy is generated through the hierarchical analysis method, and the multiple factors affecting the lighting parameters (ambient light intensity data, ambient color temperature data, user pupil diameter data, user blinking frequency data and user line of sight long-term fixed area data) are layered according to their importance in lighting control; each layer contains specific decision criteria, for example, the changes in ambient light intensity and color temperature are considered in the environmental layer, and the user pupil diameter, blinking frequency and other data are considered in the user physiological state layer. According to the above criteria, by analyzing each pair of multiple factors affecting the lighting parameters, the decision criteria are analyzed. The authors compare factors in pairs to determine the relative importance of the multiple factors affecting the lighting parameters. Scaling methods are typically used for evaluation, and the comparison results for each pair of factors affecting the lighting parameters are expressed numerically to form a judgment matrix. Through matrix operations and consistency tests, the weight priority of each factor is determined. The eigenvalues ​​and eigenvectors of the judgment matrix are solved to obtain the weight values ​​of each lighting parameter factor. Based on the weight values ​​of each lighting parameter factor, the most important lighting parameter factors are ranked, and the weight values ​​of the lighting parameter factors are used to determine the priority of the initial lighting strategy parameters for each scene. This method allows the most appropriate preliminary lighting strategy to be selected for different environments and user states.

[0061] The physiological analysis module calculates the user's physiological indicators and marks fatigue warnings based on the preprocessed multi-source data through the Daugman algorithm to obtain the user's fatigue status.

[0062] The preprocessed multi-source data is input into the Daugman algorithm, and the pupil and iris boundaries are extracted through circle detection and bilinear interpolation, and the user's pupil diameter change rate physiological indicator is calculated.

[0063] It should be noted that the expression for calculating the physiological indicator of the user's pupil diameter change rate is:

[0064] ;

[0065] Among them, R is the physiological index of pupil diameter change rate, D t1 is the user's pupil diameter data at the first moment t1, D t2 is the user's pupil diameter data at the second moment t2;

[0066] Specifically, the preprocessed multi-source data is input into the Daugman algorithm, and the candidate center and radius of the pupil and iris boundary are located in the user pupil diameter data image using a circle detection method. The integral differential operator traverses the user pupil diameter data image space. The integral differential operator traverses the user pupil diameter data image pixel points along the circular path generated by the candidate center and radius, and performs an accumulation operation on the change amplitude of the grayscale value of the user pupil diameter data image pixel on the circular path. The accumulated result is used as the edge intensity value. The integral differential operator moves along the circular path within the user pupil diameter data image space, with the center of the circle as the center and the length of the candidate radius as the distance, one by one. The grayscale value difference between adjacent user pupil diameter data image pixels on the path is extracted, and the absolute value of the difference is summed to generate the edge intensity value. Repeatedly perform the summation operation of the absolute value of grayscale value differences on the circular paths corresponding to different center and radius combinations. The larger the edge strength value, the closer the current center and radius combination is to the true pupil or iris boundary. The pupil and iris boundary geometric parameters corresponding to the maximum edge response value are identified, and accurate pupil position and size data are output;

[0067] The bilinear interpolation method performs interpolation operations on the pupil and iris boundaries, generating new pixel values ​​based on the weighted average of the grayscale values ​​of the four surrounding known pixels, improving the boundary positioning accuracy to the sub-pixel level. The interpolation results extract the closed curve coordinates of the pupil and iris boundaries, and fit the boundary curve equation through the least squares method to eliminate noise interference and smooth the geometric shape; calculate the user's pupil diameter change rate physiological indicator, and generate the user's pupil diameter change rate physiological indicator.

[0068] Based on historical fatigue data and historical normal state data, the fatigue threshold is obtained by parameter optimization using the support vector machine method.

[0069] Specifically, the support vector machine method uses the user's pupil diameter change rate and blink frequency data from historical fatigue data and historical normal state data as feature inputs. The historical fatigue data and historical normal state data contain the mean of the user's pupil diameter change rate and the variance of the user's blink frequency data. The support vector machine method uses a radial basis kernel function to map the feature vectors of the historical fatigue data and historical normal state data into a high-dimensional space. During the training process, a grid search method is used to traverse candidate combinations of kernel function parameters and penalty coefficients. Cross-validation is performed on each candidate combination to evaluate classification accuracy. The cross-validation method divides the historical fatigue data and historical normal state data into training and validation sets. The training set determines the support vectors by solving the Lagrangian dual problem and generates a hyperplane with the maximum classification margin. The hyperplane equation parameters are derived from a linear combination of the support vectors and the label values. The validation set tests the generalization performance of the hyperplane. The classification decision function is generated based on the optimal kernel function parameters and penalty coefficient. The optimal parameters correspond to the candidate combination with the highest classification accuracy. After the optimal parameters are locked, the fatigue threshold is determined by the geometric distance between the classification hyperplane and the origin.

[0070] The user's pupil diameter change rate physiological indicator and the user's blinking frequency data are compared with the fatigue threshold to obtain the user's fatigue status.

[0071] Specifically, the value of the user's pupil diameter change rate physiological indicator is compared with the fatigue threshold value. When the value of the user's pupil diameter change rate physiological indicator exceeds the fatigue threshold value, the user's fatigue state is marked as fatigue. When the value of the user's pupil diameter change rate physiological indicator is lower than the fatigue threshold value, the user's fatigue state is marked as non-fatigue. The value of the user's blinking frequency data collected in real time is compared with the fatigue threshold value. When the value of the user's blinking frequency data exceeds the fatigue threshold value, the user's fatigue state is marked as non-fatigue. When the value of the user's blinking frequency data is lower than the fatigue threshold value, the user's fatigue state is marked as fatigue. When the comparison result of the user's pupil diameter change rate physiological indicator is marked as fatigue or the comparison result of the user's blinking frequency data is marked as fatigue, the user's fatigue state is determined to be fatigue; when the comparison result of the user's pupil diameter change rate physiological indicator is marked as non-fatigue and the comparison result of the user's blinking frequency data is marked as non-fatigue, the user's fatigue state is determined to be non-fatigue.

[0072] Based on the area where the user's gaze is fixed for a long time, the spatial coordinate mapping and clustering algorithm are used to process the gaze point and divide the area to generate the focus area coordinates.

[0073] The area where the user's line of sight is fixed for a long time is converted into a unified standardized coordinate system through a spatial coordinate mapping algorithm.

[0074] Specifically, the original coordinate data in the data of the area where the user's line of sight is fixed for a long time is input into the spatial coordinate mapping algorithm. The original coordinate data includes the horizontal pixel position and the vertical pixel position. The spatial coordinate mapping algorithm determines the horizontal maximum value and the vertical maximum value of the standardized coordinate system based on the screen resolution parameter. The screen resolution parameter includes the total number of horizontal pixels and the total number of vertical pixels. The original coordinate data is normalized by the screen resolution parameter, and the horizontal and vertical pixel positions are converted into a standardized coordinate system in the range of 0 to 1. The standardized horizontal coordinate and the standardized vertical coordinate are combined into a coordinate value in the standardized coordinate system. The horizontal coordinate range of the standardized coordinate system is limited to 0 to 1, and the vertical coordinate range is limited to 0 to 1. Each original coordinate point in the data of the area where the user's line of sight is fixed for a long time performs horizontal and vertical normalization conversion operations to generate a standardized coordinate system.

[0075] The DBSCAN clustering algorithm is used to cluster the areas where the user's gaze is fixed for a long time based on their spatial location similarity, to divide the initial focus area and remove abnormal points and smooth boundaries.

[0076] Specifically, the coordinate set of the user's long-term fixed line of sight data in the standardized coordinate system is input into the DBSCAN clustering algorithm. The DBSCAN clustering algorithm defines the neighborhood radius parameter and the minimum neighborhood sample number parameter. Based on the Euclidean distance formula, each focus area coordinate point in the focus area coordinate set is traversed, and the number of neighboring coordinate points within the neighborhood radius of each focus area coordinate point is counted. The focus area coordinate points whose number of neighboring focus area coordinate points exceeds the minimum neighborhood sample number parameter are marked as core points. The core points and other coordinate points in the neighborhood are merged into the same cluster. The focus area coordinate points that cannot be classified into any cluster are marked as noise points. The noise points are removed from the focus area coordinate set, and the remaining coordinate points form the preliminary focus area coordinate points.

[0077] The set of preliminary focus area coordinate points is input into the convex hull generation algorithm. The convex hull generation algorithm selects the preliminary focus area coordinate point with the smallest y coordinate value from the preliminary focus area coordinate point set as the initial extreme point, measures the extreme angles of other coordinate points in the coordinate point set relative to the initial extreme point, and sorts the preliminary focus area coordinate points in ascending order of extreme angles. The convex hull generation algorithm connects adjacent coordinate points in the order of the sorted preliminary focus area coordinate points to form candidate edges. The candidate edges and the polygon edges formed by the connected coordinate points are checked for convexity, and preliminary focus area coordinate points that cause the polygon to be concave are deleted. The convex hull generation algorithm traverses all sorted preliminary focus area coordinate points, retains only the preliminary focus area coordinate points that constitute a convex polygon, and uses the closed polygon formed by the final connection as the boundary polygon. The vertex coordinate set of the boundary polygon is combined with the preliminary focus area coordinate points to demarcate the preliminary focus area.

[0078] Based on the position of the preliminary focus area in the standardized coordinate system, the coordinates of the focus area are obtained by analysis using the minimum enclosing rectangle method.

[0079] Specifically, the minimum bounding rectangle method traverses the horizontal preliminary focus area coordinate values ​​of all preliminary focus area coordinate points in the preliminary focus area coordinate point set, records the minimum value of the left boundary horizontal coordinate value as the left boundary horizontal preliminary focus area coordinate value of the focus area, and records the maximum value of the right boundary horizontal preliminary focus area coordinate value as the right boundary horizontal preliminary focus area coordinate value of the focus area;

[0080] The minimum bounding rectangle method traverses the longitudinal boundary coordinate values ​​of all coordinate points in the initial focus area coordinate point set, records the minimum longitudinal boundary coordinate value as the lower longitudinal boundary coordinate value of the focus area, and records the maximum longitudinal boundary coordinate value as the upper longitudinal boundary coordinate value of the focus area. The left boundary horizontal initial focus area coordinate value, the right boundary horizontal initial focus area coordinate value, the lower boundary vertical coordinate value, and the upper boundary vertical coordinate value are combined to form the four vertex coordinates of the focus area: the left boundary horizontal initial focus area coordinate value and the upper boundary vertical initial focus area coordinate value form the upper left vertex coordinate, the right boundary horizontal initial focus area coordinate value and the upper boundary vertical initial focus area coordinate value form the upper right vertex coordinate, the right boundary horizontal initial focus area coordinate value and the lower boundary vertical initial focus area coordinate value form the lower right vertex coordinate, and the left boundary horizontal initial focus area coordinate value and the lower boundary vertical initial focus area coordinate value form the lower left vertex coordinate. The horizontal and vertical coordinate values ​​of the four vertex coordinates are expressed as numerical values ​​in the standardized coordinate system to obtain the focus area coordinates.

[0081] The instruction generation module fuses the scene's initial light strategy with the focus area coordinates through a weighted fusion method, and combines the user's vision parameters to form light parameters.

[0082] Based on the scene's initial light strategy data and the focus area coordinates, the principal component analysis method is used to analyze the degree of influence and correlation on the user's visual experience, and the weight of the user's visual experience influence is obtained. The weighted fusion method is used to fuse them and obtain the light strategy.

[0083] Specifically, the scene's initial light strategy data and the focus area coordinate data are input into the principal component analysis method. The principal component analysis method performs standardization on the illumination intensity parameters and color temperature parameters of the scene's initial light strategy data and the horizontal and vertical position parameters of the focus area coordinate data to eliminate dimensional differences. The principal component analysis method calculates the covariance values ​​between each pair of the standardized illumination intensity parameters, color temperature parameters, horizontal and vertical position parameters. All pairwise covariance values ​​of the four variables are arranged into a 4×4 symmetric matrix, where the element in the i-th row and j-th column of the symmetric matrix is ​​the covariance value between the i-th parameter and the j-th parameter, resulting in a covariance matrix.

[0084] The eigenvalue decomposition of the covariance matrix extracts the principal component directions and variance contributions. The principal component direction corresponds to the eigenvector with the largest variance contribution. The absolute value of the element of the eigenvector with the largest variance contribution represents the weight of the impact of the light intensity parameter, color temperature parameter, horizontal position parameter, and vertical position parameter on the user's visual experience. The principal component analysis method normalizes the influence weights into weight coefficients whose sum is 1. The light intensity parameters and color temperature parameters of the scene's initial light strategy data, and the horizontal position parameters and vertical position parameters of the focus area coordinate data are weighted summed according to the weight coefficients. The weighted summation result is used as the fused light strategy parameters. The light intensity value and color temperature value in the light strategy parameters are combined to form the light strategy.

[0085] The light strategy is combined with the user's vision parameters, and optimized and matched through an adaptive adjustment algorithm to form light parameters suitable for the user.

[0086] Specifically, the user's vision parameters are obtained by using the vision test report data generated by the medical testing equipment, including the specific values ​​of the vision grade parameters measured by the optometer, the light sensitivity parameters recorded by the light sensitivity tester, and the color vision difference parameters calibrated by the color vision test card;

[0087] The light intensity parameter, color temperature parameter in the light strategy and the vision level parameter, light sensitivity parameter and color vision difference parameter in the user vision parameter are input into the adaptive adjustment algorithm. The adaptive adjustment algorithm performs normalization processing on the user vision parameter and converts the vision level parameter, light sensitivity parameter and color vision difference parameter into values ​​ranging from 0 to 1. The adaptive adjustment algorithm adjusts the light intensity parameter and color temperature parameter according to the normalized user vision parameter value. The light intensity parameter and the vision level parameter value perform reverse linear adjustment, and the color temperature parameter and the light sensitivity parameter value perform forward linear adjustment. The adaptive adjustment algorithm adjusts the adjusted light intensity parameter and the color temperature parameter according to the normalized user vision parameter value. The light intensity parameter, color temperature parameter, and color vision difference parameter in the user's vision parameter are input into the color temperature compensation formula. The color temperature compensation formula performs addition and subtraction offset operations on the color temperature parameter according to the color vision difference parameter value. The adaptive adjustment algorithm performs a range limitation operation on the finally adjusted light intensity parameter and color temperature parameter. The light intensity parameter is limited to between the minimum and maximum values ​​allowed by the user's vision parameter, with an exemplary minimum value of 3000K warm yellow light and a maximum value of 6000K cold white light. The color temperature parameter is limited to the color temperature range defined by the user's comfort level. The light intensity parameter and color temperature parameter after range limitation are combined into light parameters suitable for the user.

[0088] The lighting control instructions are obtained through analysis using the NSGA-II algorithm.

[0089] Based on light parameters, a multi-objective function of lighting energy saving and visual comfort is set, and a set of lighting control instruction parameter combination populations containing multiple individuals is randomly generated and initialized.

[0090] Specifically, the allowable ranges of the light intensity parameter and the color temperature parameter in the light parameters are used as constraints. The lighting energy-saving objective function is set as the ratio of the lamp energy consumption value to the rated power. The visual comfort objective function is the sum of the absolute deviation between the actual light intensity parameter and the light intensity parameter in the light parameters, and the absolute deviation between the actual color temperature parameter and the color temperature parameter in the light parameters. A random number generator generates random light intensity parameter values ​​within the allowable range of the light intensity parameter, generates random color temperature parameter values ​​within the allowable range of the color temperature parameter, and generates random light switching time parameter values ​​within the allowable range of the light switching time. The three random parameter values ​​are combined into a lighting control instruction parameter combination individual. The random parameter value generation operation is repeated to generate a preset number of lighting control instruction parameter combination individuals. All individuals constitute an initialized lighting control instruction parameter combination population.

[0091] The non-dominated sorting and crowding analysis of the lighting control command parameter combination population are performed through selection, crossover, mutation, merging, updating and elite retention.

[0092] Specifically, each individual in the lighting control instruction parameter combination population is input into the non-dominated sorting method. The non-dominated sorting method compares the individual values ​​on the lighting energy-saving objective function and the visual comfort objective function to determine the dominance relationship between individuals. The non-dominated sorting method divides individuals without dominated relationships into the first non-dominated level. The remaining individuals are compared again for dominance and divided into the second non-dominated level. This is repeated until all individuals are assigned to the non-dominated level. The crowding analysis method performs an objective function spatial distribution density assessment on individuals within the same non-dominated level, measuring the sum of the numerical differences between each individual and its adjacent individuals in the lighting energy-saving objective function and the visual comfort objective function as the crowding value.

[0093] The selection operation selects and retains high-level and high-crowding individuals according to the non-dominated level priority and crowding value. The crossover operation performs arithmetic averaging on the light intensity parameter values, color temperature parameter values, and lamp switching time parameter values ​​of two individuals to generate a new individual. The mutation operation applies random perturbations to the light intensity parameter values, color temperature parameter values, and lamp switching time parameter values ​​of the individual to generate a new individual. The merge operation combines the retained original individuals with the newly generated, crossed individuals and mutated individuals into a temporary population. The elite retention operation extracts the top 10% of individuals of the non-dominated level from the original population and directly replaces the same number of lowest-level individuals in the temporary population. The update operation re-executes non-dominated sorting and crowding analysis on the temporary population to screen out individuals with the same number as the original population as the new generation of lighting control instruction parameter combination population.

[0094] When the multi-objective functions of lighting energy saving and visual comfort converge, the optimal individual is selected from the lighting control instruction parameter combination population to obtain the lighting control instruction.

[0095] Specifically, by analyzing the absolute values ​​of the numerical change rates of the lighting energy-saving objective function and the visual comfort objective function recorded in the historical lighting control instruction parameter combination population during the iteration process, the 95% quantiles of the two sets of values ​​are taken as initial candidate values. After verification and adjustment of the test data set, the final dual-objective convergence criterion threshold range is determined. When the numerical change rates of the lighting energy-saving objective function and the visual comfort objective function of the final lighting control instruction parameter combination population are both less than the dual-objective convergence criterion threshold, it is determined to have converged;

[0096] Each individual in the final lighting control instruction parameter combination population is traversed, and its non-dominated rank is checked to see if it is the first non-dominated rank. All individuals with the first non-dominated rank are retained to form a set of candidate optimal individuals. This set of candidate optimal individuals is input into the crowding analysis method, and the crowding value of each candidate optimal individual in the space of the lighting energy-saving objective function and the visual comfort objective function is calculated. The candidate optimal individuals are sorted from highest to lowest by crowding value. The candidate optimal individual with the largest crowding value is selected as the optimal individual. The degree of overlap between the optimal individual's light intensity parameter value and the allowable range of the light intensity parameter in the light parameters, and the degree of overlap between the color temperature parameter value and the allowable range of the color temperature parameter in the light parameters, are verified. The optimal individual is considered valid when the overlap exceeds 90%. The light intensity parameter value, color temperature parameter value, and lamp switching time parameter value of the optimal individual are combined to form the lighting control instruction.

[0097] The execution module transmits and executes lighting control instructions through wireless communication methods, monitors the user's visual physiological reactions in real time, and obtains feedback data.

[0098] The lighting control instructions are encoded and converted into JSON format for wireless communication transmission, and sent to the lighting end through the wireless communication protocol.

[0099] Specifically, the light intensity parameter values, color temperature parameter values, and lamp on / off time parameter values ​​in the lighting control command are arranged in a key-value pair format. The light intensity parameter values ​​are rounded to two decimal places, the color temperature parameter values ​​are rounded to integers, and the lamp on / off time parameter values ​​are converted to "HH:MM" format strings. The key-value pairs are encapsulated into JSON-formatted packets. The wireless communication protocol defines a packet structure consisting of a target device address field, a command type field, a JSON-formatted data field, and a checksum field. The checksum field is generated using the CRC32 algorithm. The wireless communication protocol converts the complete packet into a binary byte stream and establishes a socket connection with the lighting end via the TCP / IP protocol. The byte stream is then fragmented and encapsulated into TCP data frames. Each TCP data frame is appended with a header identifier and a checksum. The frame is then continuously transmitted to the lighting end at 500ms intervals. After transmission, the protocol waits for an acknowledgment frame from the lighting end. If no acknowledgment frame is received, a 3-second retransmission mechanism is initiated, with a maximum retransmission limit of 3.

[0100] The lighting end receives the lighting control instructions transmitted by wireless communication, decodes them and restores them to the original lighting control instructions, and adjusts and executes the lighting parameters according to the lighting control instructions.

[0101] Specifically, the lighting end receives binary byte stream data packets transmitted by wireless communication via the TCP / IP protocol. The binary byte stream data packets are verified for integrity based on the frame header identifier and the frame footer check sequence, which is generated using the CRC32 algorithm. After the integrity verification passes, the binary byte stream data packet parses the target device address field and the command type field, matches the current lighting end address and the command type, extracts the JSON format data field, and converts it into minutes.

[0102] The light intensity parameter value is compared with the allowable range of light intensity in the light parameters, and the color temperature parameter value is compared with the allowable range of color temperature in the light parameters. If the value exceeds the allowable range, the boundary value of the allowable range is used as a substitute. The lighting end generates a pulse width modulation signal to adjust the lamp drive current, converts the light intensity parameter value into a duty cycle value output to generate a color temperature control voltage signal, converts the color temperature parameter value into a voltage amplitude output, and inputs the lamp switching time parameter value into the timer unit. The timer unit is configured in countdown mode, and triggers the relay switching action after the countdown ends. After the lighting end executes the parameter adjustment, it returns a confirmation frame containing the execution result code to the sending end via the TCP / IP protocol. The execution result code includes the actual value of the light intensity, the actual value of the color temperature, and the remaining seconds of the switching time.

[0103] Monitor the user's visual physiological response data in real time and perform preprocessing to extract feedback data related to the user's visual physiological response.

[0104] Specifically, the user's visual physiological response is monitored in real time to extract the pupil and iris boundaries, the user's pupil diameter change rate physiological indicator is calculated, the user's blink frequency data is recorded using a near-infrared spectral sensor at a sampling frequency of 30 times per second, and the coordinate data of the user's long-term fixed gaze area is obtained using an eye tracking device at a refresh rate of 120Hz. The original user pupil diameter change rate physiological indicator is subjected to sliding average filtering, with the window width set to 10 sampling points to remove high-frequency noise; the original user blink frequency data is subjected to eye closing action feature recognition processing, and the eye closing action is identified based on the feature of eyelid coverage exceeding 80% and duration greater than 100ms; the original user's long-term fixed gaze area coordinate data is subjected to outlier removal processing, and coordinate points that deviate from the mean by three standard deviations are removed. The mean and variance per minute are extracted from the preprocessed user pupil diameter change rate physiological indicator as feedback data, the number of eye closures per minute and the average duration are extracted from the preprocessed user blink frequency data as feedback data, and the cluster center coordinates and distribution radius are extracted from the preprocessed user's long-term fixed gaze area coordinate data as feedback data.

[0105] The optimization module dynamically modifies the light parameters using a fuzzy logic algorithm based on the user's fatigue state to obtain the optimized light parameters.

[0106] The user's fatigue state is fuzzified through fuzzy logic algorithm and converted into fuzzy language variables. Combined with the fuzzy rule base, logical reasoning is performed through fuzzy reasoning method to obtain fuzzy conclusions for light parameter adjustment.

[0107] Specifically, the user's fatigue state is input into the fuzzy logic algorithm, and the fuzzy logic algorithm defines the fuzzy linguistic variables of the user's fatigue state as "fatigue" and "non-fatigue", corresponding to the triangular membership function. For example, the vertices of the triangular membership function are set to 0.8 times, 1.0 times, and 1.2 times the fatigue threshold respectively;

[0108] By collecting user visual physiological response data under different combinations of light intensity and color temperature, the user visual physiological response data covers 50 groups, and the Pearson coefficient is used to screen user visual physiological response characteristics with statistical correlation in the percentage of light parameter adjustment, and an initial rule base is formed by combining logical relationships; the rule consequent parameters are adjusted through the training set (70%), and the classification accuracy is verified through the test set (30%). The iterative optimization is achieved to an accuracy rate exceeding 85%, and finally 8-12 fuzzy rules are generated, such as the exemplary "high pupil index + low blink variance + medium distribution radius → light +25%, color temperature -15%". The fuzzy inference method adopts the Mamdani inference model, and a minimum-maximum operation is performed on the antecedent membership and consequent membership of each fuzzy rule. After aggregating the output membership functions of all fuzzy rules, the centroid method is used to defuzzify and generate a fuzzy conclusion for light parameter adjustment. Exemplary fuzzy conclusions for light parameter adjustment include light intensity adjustment direction (increase / decrease / maintain) and adjustment percentage (10%-30%), color temperature adjustment direction (increase / decrease / maintain) and adjustment percentage (5%-20%).

[0109] The center of gravity method is used to defuzzify the fuzzy conclusions and convert them into numerical values ​​to obtain the optimized light parameters.

[0110] Specifically, the light intensity adjustment direction (increase / decrease / maintain) and the adjustment amount percentage (10%-30%), and the color temperature adjustment direction (increase / decrease / maintain) and the adjustment amount percentage (5%-20%) of the fuzzy conclusion of the light parameter adjustment are input into the center of gravity method, the light intensity adjustment amount percentage membership function is discretely sampled in an exemplary definition domain [-30%, 30%] with a step size of 0.1%, and the color temperature adjustment amount percentage membership function is discretely sampled in an exemplary definition domain [-20%, 20%] with a step size of 0.1%. The horizontal coordinate value of each discrete point is multiplied by the corresponding vertical coordinate membership value, the sum of all product results of the light intensity adjustment amount percentage is divided by the total membership, the sum of all product results of the color temperature adjustment amount percentage is divided by the total membership, the weighted average of the light intensity adjustment amount percentage is rounded to one decimal place, and the weighted average of the color temperature adjustment amount percentage is rounded to one decimal place. The light intensity value of the current light parameter is multiplied by 1 + the light intensity adjustment percentage / 100 to obtain the optimized light intensity value. The color temperature value of the current light parameter is multiplied by 1 + the color temperature adjustment percentage / 100 to obtain the optimized color temperature value. The optimized light intensity value and the optimized color temperature value are combined to form the optimized light parameter.

[0111] The feedback data is combined with data fusion and intelligent decision-making algorithms to generate the optimal intelligent eye-protection lighting control parameters.

[0112] The optimized light parameters are combined with the feedback data, integrated through weighted average fusion, and in-depth analysis is performed through Bayesian decision making to generate the optimal intelligent eye-protection lighting control parameters.

[0113] Specifically, the light intensity value and color temperature value in the optimized light parameters are input into a weighted average fusion method with the mean value of the physiological indicator of the user's pupil diameter change rate in the feedback data and the variance of the user's blinking frequency data. The weighted average fusion method sets the light intensity value weight coefficient according to the historical fluctuation range of the mean value of the physiological indicator of the user's pupil diameter change rate in the feedback data, and sets the color temperature value weight coefficient according to the historical fluctuation range of the variance of the user's blinking frequency data. The Bayesian decision method establishes a priori probability distribution of the light intensity value and the color temperature value, and the fused light intensity value and the fused color temperature value are input into the Bayesian decision method to analyze the posterior probability distribution. The Bayesian decision method traverses all possible combinations of light intensity values ​​and color temperature values, and selects the light intensity value and color temperature value with the largest posterior probability as the intelligent eye protection lighting control parameter to verify whether the light intensity value is within the allowable range of the user's vision parameters and whether the color temperature value is within the color temperature range defined by the user's comfort level. Finally, the light intensity value and color temperature value combination that pass the verification is used as the intelligent eye protection lighting control parameter.

[0114] In summary, the present invention uses: the Markov model to predict environmental changes, which can perceive the evolution trend of factors such as ambient light intensity in advance, adjust the lighting strategy in advance, and avoid lighting lagging behind environmental changes; improve the foresight and adaptability of lighting, reduce the impact of environmental mutations on lighting effects, create a more stable and comfortable lighting environment for users, and enhance user experience; the fuzzy logic algorithm responds to fatigue status, and can convert the user's complex and fuzzy fatigue status into precise light parameter adjustment instructions; it can flexibly adapt the lighting environment according to different fatigue levels, effectively relieve visual fatigue, provide users with personalized and dynamic eye protection lighting solutions, and improve lighting intelligence and user comfort.

[0115] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. An intelligent eye-protection lighting control system, characterized by: include, The data acquisition module collects multi-source data in real time for pre-processing and analyzes it through the Markov model to obtain the initial light strategy of the scene; Specifically, the system collects ambient light intensity data, ambient color temperature data, user pupil diameter data, user blink frequency data, and data on the area where the user's gaze is fixed for a long time in real time; performs standardization, noise removal, and fills in missing values ​​and outlier processing; sequentially extracts ambient light intensity data, ambient color temperature data, user pupil diameter data, user blink frequency data, and data on the area where the user's gaze is fixed for a long time from the preprocessed historical multi-source data, and trains a Markov model through maximum likelihood estimation; inputs the real-time collected ambient light intensity, ambient color temperature, user pupil diameter, blink frequency, and area where the user's gaze is fixed into the Markov model, and predicts the trend of light changes based on the state transition probability of the Markov model; Screen the preliminary light strategy parameters through the hierarchical analysis method and output the initial light strategy of the scene; The physiological analysis module uses the Daugman algorithm based on preprocessed multi-source data to calculate the user's pupil diameter physiological indicator and mark fatigue warnings to determine the user's fatigue status. Based on the user's long-term fixed gaze area, the module uses spatial coordinate mapping and clustering algorithms to process the gaze point and divide the area to generate the focus area coordinates. Specifically, based on the pre-processed user pupil diameter data, circle detection and bilinear interpolation are performed using the Daugman algorithm to extract the pupil boundary, and the user's pupil diameter change rate physiological indicator is calculated; Based on historical fatigue data and historical normal state data, a fatigue threshold is generated by optimizing parameters using a support vector machine method; the user's pupil diameter change rate physiological indicator and the user's blink frequency data are compared with the fatigue threshold: if the user's pupil diameter change rate physiological indicator is greater than the fatigue threshold or the user's blink frequency data is less than the fatigue threshold, the user is marked as fatigued; Otherwise, it is marked as non-fatigue; output the user's fatigue status according to the above comparison result; The instruction generation module fuses the scene initial light strategy and the focus area coordinates through a weighted fusion method, combines the user's vision parameters to form light parameters, and analyzes them through the NSGA-II algorithm to obtain lighting control instructions. Specifically, based on the scene initial light strategy data and the focus area coordinates, the principal component analysis method is used to analyze the degree of influence and correlation of the user's visual experience, and the weight of the user's visual experience influence is obtained. The weighted fusion method is used to fuse them to obtain the light strategy; the light strategy is combined with the user's vision parameters. The user's vision parameters are obtained by the specific values ​​of the vision test report data generated by the medical testing equipment, including the vision grade parameters measured by the optometer, the light sensitivity parameters recorded by the light sensitivity tester, and the color vision difference parameters calibrated by the color vision test card. The user's vision parameters are optimized and matched through the adaptive adjustment algorithm to form light parameters; based on the light parameters, a multi-objective function of lighting energy saving and visual comfort is set, and the lighting control instructions are generated through the NSGA-II algorithm; The execution module transmits and executes lighting control instructions via wireless communication methods, monitors the user's visual physiological reactions in real time, and obtains feedback data; The optimization module uses fuzzy logic algorithms to dynamically correct light parameters based on the user's fatigue status to obtain optimized light parameters. It also combines feedback data through data fusion and intelligent decision-making algorithms to generate optimal intelligent eye-protection lighting control parameters.

2. The intelligent eye-protection lighting control system according to claim 1, characterized in that: The multi-source data includes ambient light intensity data, ambient color temperature data, user pupil diameter data, user blinking frequency data, and data on the area where the user's line of sight is fixed for a long time; The preprocessing includes standardization, noise removal, filling missing values ​​and outlier processing.

3. The intelligent eye-protection lighting control system according to claim 2, wherein: The Markov model is used to analyze the scene initial light strategy. The specific steps are as follows: Based on the pre-processed historical multi-source data, the Markov model is trained by the maximum likelihood estimation method to obtain a trained Markov model; The pre-processed multi-source data is input into the Markov model. Based on the transition probability relationship between different environmental and user status data, the changing trend of ambient light intensity in the current scene is predicted, and the corresponding preliminary light strategy parameters for different scenes are analyzed. The different environmental and user status data include ambient light intensity data, ambient color temperature data, user pupil diameter data, user blink frequency data, and data on the area where the user's gaze is fixed for a long time. The preliminary lighting strategy parameters are sorted and screened through the hierarchical analysis method to obtain the initial lighting strategy for the scene.

4. The intelligent eye-protection lighting control system according to claim 1, wherein: The method uses the Daugman algorithm based on the pre-processed multi-source data to calculate the user's physiological indicators and mark the fatigue warning to obtain the user's fatigue status. The specific steps are as follows: The pre-processed multi-source data is input into the Daugman algorithm, and the pupil and iris boundaries are extracted through circle detection and bilinear interpolation, and the user's pupil diameter change rate physiological indicator is calculated; Based on historical fatigue data and historical normal state data, the fatigue threshold is obtained by optimizing parameters using the support vector machine method. The user's pupil diameter change rate physiological indicator and the user's blinking frequency data are compared with the fatigue threshold to obtain the user's fatigue status.

5. The intelligent eye-protection lighting control system according to claim 1, wherein: According to the user's long-term fixed line of sight area, the line of sight landing point processing and area division are performed through spatial coordinate mapping and clustering algorithm to generate the focus area coordinates. The specific steps are as follows: The area where the user's line of sight is fixed for a long time is converted into a unified standardized coordinate system through a spatial coordinate mapping algorithm; The DBSCAN clustering algorithm is used to cluster the areas where the user's gaze is fixed for a long time based on their spatial location similarity, thereby dividing the initial focus area and removing abnormal points and smoothing the boundaries. Based on the position of the preliminary focus area in the standardized coordinate system, the coordinates of the focus area are obtained by analysis using the minimum enclosing rectangle method.

6. The intelligent eye-protection lighting control system according to claim 5, characterized in that: The scene initial light strategy and the focus area coordinates are fused through a weighted fusion method, and the light parameters are formed by combining the user's vision parameters. The specific steps are as follows: Based on the scene's initial light strategy data and the focus area coordinates, the principal component analysis method is used to analyze the degree of impact and correlation on the user's visual experience, and the user's visual experience impact weight is obtained. The weighted fusion method is used to fuse them and obtain the light strategy. The light strategy is combined with the user's vision parameters, and optimized and matched through an adaptive adjustment algorithm to form light parameters.

7. The intelligent eye-protection lighting control system according to claim 6, characterized in that: The lighting control instructions are obtained by analyzing the NSGA-II algorithm. The specific steps are as follows: Based on light parameters, a multi-objective function of lighting energy saving and visual comfort is set, and a set of lighting control instruction parameter combination populations containing multiple individuals is randomly generated and initialized; Perform non-dominated sorting and crowding analysis on the lighting control command parameter combination population through selection, crossover, mutation, merging, updating and elite retention; When the multi-objective functions of lighting energy saving and visual comfort converge, the optimal individual is selected from the lighting control instruction parameter combination population to obtain the lighting control instruction.

8. The intelligent eye-protection lighting control system according to claim 7, characterized in that: The lighting control instructions are transmitted and executed through wireless communication methods, and the user's visual physiological response is monitored in real time to obtain feedback data. The specific steps are as follows: Encode the lighting control instructions, convert them into JSON format for wireless communication transmission, and send them to the lighting end through the wireless communication protocol; The lighting end receives the lighting control instructions transmitted by wireless communication, decodes them and restores them to the original lighting control instructions, and adjusts and executes the lighting parameters according to the lighting control instructions; Monitor the user's visual physiological response data in real time and perform preprocessing to extract feedback data related to the user's visual physiological response.

9. The intelligent eye-protection lighting control system according to claim 1, wherein: The fuzzy logic algorithm is used to dynamically modify the light parameters based on the user's fatigue state to obtain the optimized light parameters. The specific steps are as follows: The user's fatigue state is fuzzified through fuzzy logic algorithm and converted into fuzzy language variables. The fuzzy reasoning method is used to perform logical reasoning based on the fuzzy rule base to obtain the fuzzy conclusion of light parameter adjustment. The center of gravity method is used to defuzzify the fuzzy conclusions and convert them into numerical values ​​to obtain the optimized light parameters.

10. The intelligent eye-protection lighting control system according to claim 1, wherein: The optimized light parameters are combined with the feedback data, integrated through weighted average fusion, and in-depth analysis is performed through Bayesian decision making to generate the optimal intelligent eye-protection lighting control parameters.

Images