Intelligent mirror control method and system based on data processing

Abstract

The application discloses an intelligent mirror control method and system based on data processing, and belongs to the technical field of intelligent control. It aims to solve the problem that the display window of the intelligent mirror blocks the user's line of sight and cannot automatically eliminate it. First, an interactive model containing a digitized face is generated and a sampling point is set, the interactive record is screened in combination with the environmental brightness factor and the historical interactive log, the line of sight focus relationship formula is fitted and the expected line of sight focus density is calculated, and the difference between the current and expected densities is used to accurately identify the risk points of blocking; then, the projection overlap and the shortest projection distance of the application software window and the risk points of blocking are analyzed, the visual interference efficiency index and the blocking index are calculated to screen the high-risk blocking window; finally, the spatial obstacle avoidance displacement vector and the size scaling ratio are calculated according to the blocking index, and the window is automatically re-rendered or the opacity is dynamically reduced until the blocking is eliminated. The application can adapt to environmental changes, accurately identify and automatically eliminate the line of sight blocking, and realize the non-sensing optimization of the display interface.

Description

Technical Field

[0001] This invention relates to the field of intelligent control technology, specifically to an intelligent mirror control method and system based on data processing. Background Technology

[0002] Smart mirrors offer traditional grooming functions while also providing dynamic information display and multimedia interaction capabilities. However, because the digital display layer and the physical reflection layer are highly overlapping in space, when the digital interactive interface is presented as an independent window on the mirror surface, it inevitably obstructs the user's physical image. This visual obstruction disrupts the continuity of traditional mirror use, especially when users need to observe details; the abrupt interface obstruction severely breaks the visual experience.

[0003] Currently, existing mirror display interaction technologies largely rely on static interface layout management or passive user command adjustments, lacking the ability to dynamically perceive and adaptively reconstruct users' real-time visual intentions. On one hand, under complex and changing ambient lighting conditions, users' natural visual attention distribution dynamically shifts. Existing technologies struggle to establish dynamic benchmarks that adapt to environmental changes, making it difficult to accurately distinguish between natural gaze shifts and abnormal visual compensation behaviors caused by screen occlusion, easily leading to occlusion detection failure. On the other hand, even when visual obstruction is detected, existing systems lack proactive spatial obstacle avoidance mechanisms, often requiring users to manually interrupt their current task to drag or close the interface, resulting in a cumbersome interaction process that severely undermines immersion. Furthermore, existing technologies fail to comprehensively assess the physical visual characteristics of interface elements and their cumulative interference to the user's visual channel, hindering automated and seamless interface optimization and dimensionality reduction. This significantly reduces the practicality and user experience of smart mirrors in complex multi-tasking scenarios. Summary of the Invention

[0004] The purpose of this invention is to provide a smart mirror control method and system based on data processing, in order to solve the problems that, under complex and changing ambient lighting conditions, the application software window displayed by the smart mirror can easily obstruct the user's view of their own facial image, and the system cannot intelligently distinguish between the user's natural gaze shift and abnormal gaze caused by obstruction, thus failing to adaptively eliminate visual interference.

[0005] To address the aforementioned technical problems, this invention provides a data processing-based intelligent mirror control method, comprising:

[0006] When using a smart mirror, because it needs to provide mirror reflection and run various interactive applications, changes in ambient light and user movement can interfere with the focus of vision in complex bathroom or bedroom settings. Therefore, environmental parameters, user characteristics, software parameters, and historical interaction logs are collected.

[0007] Environmental parameter data includes different brightness factors, which are physical indicators that quantify the lighting conditions of the environment in which the smart mirror is located, such as the illuminance and color temperature of the bathroom ceiling light. These indicators directly determine the contrast relationship between the screen display content and the mirror reflection.

[0008] User characteristics include facial features and facial turning angles collected by a depth camera. The depth information can be used to more accurately separate the user's face from the background wall, and the facial turning angle can reflect whether the user is looking directly at the mirror to observe details or turning to the side. The user refers to the user of the smart mirror.

[0009] The software parameters include the display area of ​​the application window, edge brightness contrast factor, idle time without interaction, current spatial coordinates, and current display size. These parameters not only define the spatial occupancy of the window in the mirror coordinate system, but also reflect the window's ability to visually capture the user's attention and its current activity status.

[0010] Historical interaction logs include interaction records of each time the smart mirror runs under different environmental parameters. Each interaction record includes environmental parameters and user characteristics. By accumulating this data over a long period of time, it is possible to learn the user's natural gaze habits under different lighting conditions, thereby providing a dynamic historical comparison benchmark for judging "gaze abnormalities".

[0011] The digital display panel of the smart mirror is used as a two-dimensional coordinate plane to unify the line of sight in physical space with the window in digital space for calculation.

[0012] By analyzing the eye image in facial features, the relative vector between the pupil center and the corneal reflection point is extracted based on the pupil-corneal reflection method, and the gaze direction vector is calculated.

[0013] This method uses an infrared light source to form a reflection point on the cornea, and by tracking the relative position change between the pupil center and the reflection point, it can accurately calculate the three-dimensional orientation of the user's gaze, even in a dimly lit bathroom environment.

[0014] The gaze direction vector is intersected with the two-dimensional coordinate plane of the smart mirror to obtain the coordinate sequence of the gaze landing point as gaze tracking data, that is, the focal trajectory point of the user's gaze on the mirror surface.

[0015] The gaze focus density distribution is obtained by analyzing user characteristics of each interaction record in the historical interaction log. The gaze focus density distribution refers to the duration of time that the user's gaze lingers on different positions on the surface of the digital facial model within a preset time window. and number of fixations The normalized value distribution.

[0016] Calculate the gaze focus density of a user within a specific time period, focusing on a sampling point area on the surface of a digital facial model. The gaze focus density is a normalized value that combines the "gaze duration percentage" and the "gaze frequency percentage".

[0017] A single indicator cannot fully reflect the intensity of fixation. For example, frequent but short-duration rapid squinting, or prolonged fixed gaze, both represent high attention levels, thus requiring a combined calculation. Among these, visual focus density... The calculation formula is:

[0018] ;

[0019] In the formula, The total duration of the preset time window is used to normalize the dwell time and eliminate the dimensional differences caused by different statistical durations.

[0020] This is the total number of sampled frames within a preset time window, used to normalize the number of fixations, representing the probability of fixation occurring, and avoiding counting bias caused by high-frequency sampling devices.

[0021] and These are preset weighting coefficients used to adjust the relative importance of duration and frequency in the overall density calculation, adapting to the visual habits of different users. Users who require close observation can increase these coefficients. For users who browse globally, the setting can be increased. .

[0022] The gaze tracking data is processed into gaze focus coordinates on a two-dimensional coordinate plane.

[0023] The coordinates of the gaze focus are mapped onto the user's facial surface to generate a digital facial model with superimposed gaze focus density distribution, thereby constructing an interactive model that includes the digital facial model and the smart mirror model.

[0024] This interactive model places the user's virtual facial image and the application window on the screen in the same reference frame, enabling intuitive calculation of the spatial interference relationship between the two.

[0025] To perform a detailed assessment of different areas of the face, the continuous facial model needs to be discretized. A sampling density is set, and digitized facial surface mesh data is acquired. The total number of sight sampling points is determined based on the total area of ​​the surface mesh data and the sampling density.

[0026] Based on the Poisson disk distribution algorithm, a total number of sampling points are generated uniformly and iteratively on the surface grid data. This algorithm can ensure that the distance between any two sampling points is kept to a minimum, avoid the evaluation bias caused by the local clustering of sampling points, achieve uniform coverage of complex areas such as the nose and the corner of the eye, and establish a mapping relationship between sampling points and the coordinates of the gaze focus.

[0027] Since users' natural line of sight varies under different lighting conditions, direct comparison can lead to misjudgment. Therefore, it is necessary to find a historical benchmark that is most similar to the current environment.

[0028] Obtain the values ​​of various brightness factors of the smart mirror at the current moment. And the values ​​of various brightness factors under the environmental parameters in each interaction record in the historical interaction log. .

[0029] Substitute into the formula to calculate the difference coefficient of each interaction record. :

[0030] ;

[0031] In the formula, The number of terms in the brightness factor represents the total number of illumination indices and is used to define the scope of the summation calculation.

[0032] For the first The weighting coefficient of the brightness factor is used to determine the importance of the corresponding brightness factor in the overall illumination evaluation, and is used to make distance calculation more consistent with the characteristics of human vision.

[0033] The historical interaction logs are filtered to identify interaction records with a difference coefficient less than a preset threshold, which means filtering out historical scenes with highly similar lighting environments.

[0034] Based on the gaze focus density distribution of each interaction record, the gaze focus density of each sampling point is obtained and used as training data for regression fitting.

[0035] Data pairs consisting of gaze focus density and brightness factor at the same moment in the interaction record are extracted as sample data segments. For each sampling point, a gaze focus relationship between gaze focus density and brightness factor is fitted. Specifically:

[0036] The number of sample data segments in the interaction record for statistical sampling points The instantaneous values ​​of various brightness factors at the same time in each sample data segment are extracted as independent variables, and the line-of-sight focus density is used as the dependent variable and packaged into training samples.

[0037] Establish a linear regression model, setting the initial intercept and regression coefficients. Then... Using all independent variables in the training samples as input data, and given the parameters of intercept and regression coefficients, the model outputs the predicted gaze density, which is the gaze density that the user should have under the given illumination.

[0038] The difference between the dependent variable and the expected focal density of the line of sight in each training sample is calculated as the line of sight gap coefficient. The line of sight gap coefficients of all training samples are summed to obtain the line of sight gap index.

[0039] Using the line-of-sight gap index as the objective function for hyperparameter optimization, the intercept and regression coefficients are iteratively adjusted. The intercept and regression coefficient values ​​corresponding to the minimum line-of-sight gap index are then fixed to obtain the fitted line-of-sight focus relationship.

[0040] By obtaining the brightness factor in the environmental parameters at the current moment and substituting it into the gaze focus relationship of each sampling point, the expected gaze focus density is obtained, which in turn calculates the user's attention distribution in the current natural state when there is no application window occlusion interference.

[0041] In real-world scenarios, when an application window obscures part of a user's face, the user will instinctively and repeatedly try to look at the obscured area. This visual compensation behavior of "trying to penetrate the obstruction" will cause the actual visual density in that area to spike abnormally, far exceeding the natural prediction value, thus accurately identifying it as a risk point.

[0042] Therefore, the current line-of-sight focus density of each sampling point is obtained, and sampling points where the difference between the current line-of-sight focus density and the expected line-of-sight focus density is greater than the risk threshold are identified as occlusion risk points.

[0043] Identify the projection overlap between application software windows and occlusion risk points in the interaction model, associate application software windows with projection overlap with occlusion risk points, and thus pinpoint the specific interference source causing visual abnormalities.

[0044] Calculate the shortest projection distance between interconnected application software windows and occlusion risk points, and combine this with software parameters to calculate the visual interference efficiency index of the application software windows.

[0045] The Visual Interference Efficiency Index is used to evaluate the efficiency of visual interference caused by application software windows to points of risk of occlusion. It comprehensively evaluates window size, distance, facial posture, and the visual prominence of the window.

[0046] Visual interference efficiency index The calculation formula is:

[0047] ;

[0048] In the formula, This is the facial sensitivity weighting coefficient; it represents the importance of the facial area where the occlusion risk point is located, and is used to reflect the difference in the impact of the location of the occlusion on the visual experience.

[0049] The larger the area of ​​the application software window, the wider the obstruction range. The square relationship is used to enhance the visual obstruction effect of large windows.

[0050] The distance value is the shortest projection distance, which is the shortest straight-line distance from the edge of the window to the point of occlusion risk. The closer the distance, the stronger the interference, and the inversely proportional the decrease. Even if the window does not completely cover the point, the adjacent bright block will still distract visual attention.

[0051] This is the distance smoothing constant, a very small positive number; it prevents the denominator from becoming zero when the window is close to a risk point, thus ensuring computational stability.

[0052] The facial deflection angle represents the angle at which the user's head deviates from its position directly facing the mirror. Cosine attenuation is introduced; the larger the deflection angle, the smaller the effective obstruction of the mirror's projection, thus reducing interference. When the user turns their face to the side, the actual projected area of ​​the window in the user's field of vision will shrink.

[0053] The contrast factor is the edge brightness of the application software window, which is the difference in brightness between the window edge and the background environment. The higher the contrast, the more dazzling the window becomes, and the stronger the interference with the viewer's gaze, exhibiting a positive enhancement effect.

[0054] Calculate the visual interference efficiency index of each application window that is associated with the occlusion risk point, and substitute it into the formula to calculate the occlusion index between each application window and the occlusion risk point:

[0055] ;

[0056] In the formula, The occlusion index serves as a comprehensive risk rating for the current window causing visual obstruction; it triggers the obstacle avoidance threshold and determines the intensity of subsequent dimensionality reduction processing.

[0057] The hyperbolic tangent activation function acts as a nonlinear mapping function, compressing the linear combination of input values ​​into the (0,1) interval to prevent excessively large values ​​from overflowing, thus making the exponential output smooth and standardized.

[0058] It represents the interference sensitivity gain, serving as an amplification factor for physical interference; it is used to adjust the weighting of physical interference in the final decision.

[0059] The difference in visual focus density represents the difference between the current actual visual density and the predicted natural visual density. It reflects the degree of abnormal visual focus caused by occlusion, indicating that the occlusion is more severe.

[0060] This is the density difference normalization coefficient, used as a scaling factor for the density difference value; it is used to adjust the density difference value to a scale that is consistent with the normalization coefficient. Having the same order of magnitude ensures the mathematical validity of adding the two together.

[0061] As a negative exponential decay function, it achieves the effect of forgetting decay over time.

[0062] The forgetting decay rate is a constant used to control the decay speed; it determines how quickly the user's "tolerance" for the window being obscured increases over time.

[0063] This refers to the idle time of an application window without interaction, which is the time during which the user does not interact with the application window.

[0064] In the formula, the longer the window is idle, the more it indicates that the user does not currently need the window, and the more intolerable its occlusion becomes. This is mitigated by a negative exponential term. Values ​​that approach their maximum value are prioritized for cleaning or reduction.

[0065] The line-of-sight focus density difference refers to the difference between the current line-of-sight focus density and the expected line-of-sight focus density.

[0066] Calculate the final occlusion index of the application software window. Combine the physical visual interference of the window, the degree of abnormal concentration of the user's gaze, and introduce a time-based forgetting decay mechanism.

[0067] Application windows with an occlusion index greater than a preset occlusion threshold are selected as high-risk occlusion windows.

[0068] Obtain the two-dimensional planar coordinates of the occlusion risk points associated with the high-risk occlusion window.

[0069] The vector from the center point coordinates of the high-risk obstruction window to the two-dimensional plane coordinates of the obstruction risk point is used as the risk direction vector, and the opposite direction of the risk direction vector is used as the obstacle avoidance direction, that is, the control window moves away from the area that the user is highly concerned about in the line of sight.

[0070] The occlusion index is multiplied by a preset step size coefficient to calculate the displacement scalar; the more severe the occlusion, the farther the movement distance. The obstacle avoidance direction is multiplied by the displacement scalar to obtain the spatial obstacle avoidance displacement vector.

[0071] The current spatial coordinates of high-risk occlusion windows are updated based on the spatial obstacle avoidance displacement vector. It is then determined whether the updated current spatial coordinates exceed the display boundary of the smart mirror display to prevent the window from being moved off the screen and becoming inoperable.

[0072] If the window does not exceed the display boundary, re-render the high-risk occlusion window according to the updated current spatial coordinates.

[0073] If the window exceeds the display boundary, the distance beyond the display boundary is multiplied by a preset scaling factor to analyze the size scaling ratio. The current display size is updated according to the size scaling ratio, and the high-risk occlusion window is re-rendered. That is, when it is impossible to move the window to avoid obstacles, the window volume is reduced to make room for visual space.

[0074] The occlusion index of the high-risk occlusion window after rendering is recalculated. If the occlusion index is still greater than the preset safety threshold, the opacity of the high-risk occlusion window is dynamically reduced according to the occlusion index until the occlusion index is no greater than the preset safety threshold. As a last resort for spatial obstacle avoidance failure, the window is made semi-transparent, allowing the user's line of sight to penetrate the window and see the facial details behind it, achieving seamless optimization.

[0075] The present invention also provides a data processing-based intelligent mirror control system, comprising:

[0076] The data acquisition module is used to collect environmental parameters, user characteristics, software parameters, and historical interaction logs when the smart mirror is in use, and to generate an interaction model that includes the digital face and the smart mirror.

[0077] The intelligent analysis module is used to set sampling points on the digital facial surface, analyze historical interaction logs to filter interaction records, and fit the gaze focus relationship for the sampling points based on the interaction records, thereby setting occlusion risk points.

[0078] The window management module is used to identify and associate application software windows with overlapping projections and occlusion risk points in the interaction model, analyze the shortest projection distance between them, calculate the visual interference efficiency index and occlusion index of each application software window, and filter high-risk occlusion windows.

[0079] The optimization processing module is used to calculate the spatial obstacle avoidance displacement vector and size scaling ratio of high-risk occlusion windows based on the occlusion index, update the software parameters based on the spatial obstacle avoidance displacement vector and size scaling ratio, and control the smart mirror display to re-render the high-risk occlusion windows until the occlusion index is lower than the preset safety threshold.

[0080] Compared with the prior art, the beneficial effects achieved by the present invention are:

[0081] Existing technologies struggle to address gaze distribution shifts caused by dynamic changes in ambient lighting. This solution normalizes and quantifies gaze focus density by integrating gaze duration and fixation frequency. It also utilizes differences in ambient lighting indicators from historical interaction logs to dynamically predict the natural gaze baseline under the current environment, effectively filtering out normal environmental gaze shifts and providing a reliable data foundation for accurate occlusion identification.

[0082] Existing technologies struggle to distinguish between natural gaze shifts and abnormal gaze patterns caused by window obstruction. This solution calculates the difference between the current actual gaze focus density and the predicted natural gaze density, enabling accurate identification of abnormal gaze concentration areas caused by the user's visual compensation behavior of "trying to see behind the obstruction." These areas are then used as occlusion risk points, significantly reducing the misjudgment rate of occlusion identification.

[0083] Existing solutions do not comprehensively assess the cumulative interference to users caused by the physical and visual characteristics of interface elements and their duration of presence. This solution calculates the visual interference efficiency index by combining factors such as window display area, minimum projection distance, facial deflection angle, and edge brightness contrast, and calculates the occlusion index by integrating the degree of abnormal gaze concentration and the time penalty factor of idle time without interaction, thus achieving a comprehensive and accurate quantification of the intensity of physical occlusion and the user's psychological resistance.

[0084] Existing systems often rely on passive prompts or manual user adjustments when encountering obstructions, disrupting the immersive interactive experience. This solution automatically calculates obstacle avoidance direction and displacement scalar based on the occlusion index, prioritizing spatial translation of high-risk windows; automatically triggering size scaling when touching display boundaries; and further dynamically reducing window opacity when spatial obstacle avoidance is insufficient. The entire process dynamically eliminates line-of-sight obstructions without user intervention, achieving seamless reconstruction and automated optimization of the smart mirror display interface. Attached Figure Description

[0085] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings:

[0086] Figure 1 This is a flowchart illustrating an intelligent mirror control method based on data processing according to the present invention.

[0087] Figure 2 This is a schematic diagram of the digital facial model and the distribution of gaze sampling points according to the present invention;

[0088] Figure 3 This is a schematic diagram comparing the thermal distribution of facial gaze focus density according to the present invention. Detailed Implementation

[0089] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0090] Please see Figure 1 This invention provides a smart mirror control method based on data processing, comprising:

[0091] When using a smart mirror, because it needs to provide mirror reflection and run various interactive applications, changes in ambient light and user movement can interfere with the focus of vision in complex bathroom or bedroom settings. Therefore, environmental parameters, user characteristics, software parameters, and historical interaction logs are collected.

[0092] Environmental parameter data includes different brightness factors, which are physical indicators that quantify the lighting conditions of the environment in which the smart mirror is located, such as the illuminance (lux) and color temperature (Kelvin) of the bathroom ceiling light. These indicators directly determine the contrast relationship between the screen display content and the mirror reflection image.

[0093] User characteristics include facial features and facial turning angles collected by a depth camera. The depth information can be used to more accurately separate the user's face from the background wall, and the facial turning angle can reflect whether the user is looking directly at the mirror to observe details or turning to the side. The user refers to the user of the smart mirror.

[0094] The software parameters include the display area of ​​the application window, edge brightness contrast factor, idle time without interaction, current spatial coordinates, and current display size. These parameters not only define the spatial occupancy of the window in the mirror coordinate system, but also reflect the window's ability to visually capture the user's attention and its current activity status.

[0095] Historical interaction logs include interaction records of each time the smart mirror runs under different environmental parameters. Each interaction record includes environmental parameters and user characteristics. By accumulating this data over a long period of time, it is possible to learn the user's natural gaze habits under different lighting conditions, thereby providing a dynamic historical comparison benchmark for judging "gaze abnormalities".

[0096] The digital display panel of the smart mirror is used as a two-dimensional coordinate plane to unify the line of sight in physical space with the window in digital space for calculation.

[0097] By analyzing the eye image in facial features, the relative vector between the pupil center and the corneal reflection point is extracted based on the pupil-corneal reflection method, and the gaze direction vector is calculated.

[0098] This method uses an infrared light source to form a reflective point (Purkin spot) on the cornea, and by tracking the relative position change between the pupil center and the reflective point, it can accurately calculate the three-dimensional orientation of the user's gaze, even in a dimly lit bathroom environment.

[0099] The gaze direction vector is intersected with the two-dimensional coordinate plane of the smart mirror to obtain the coordinate sequence of the gaze landing point as gaze tracking data, that is, the focal trajectory point of the user's gaze on the mirror surface.

[0100] The gaze focus density distribution is obtained by analyzing user characteristics of each interaction record in the historical interaction log. The gaze focus density distribution refers to the duration of time that the user's gaze lingers on different positions on the surface of the digital facial model within a preset time window. and number of fixations The normalized value distribution.

[0101] For example, when users check their facial makeup in the mirror, their gaze lingers on the eye or lip area for a longer period and scans it repeatedly. This distribution can accurately depict this attentional habit.

[0102] Calculate the gaze focus density of a user within a specific time period, focusing on a sampling point area on the surface of a digital facial model. The gaze focus density is a normalized value that combines the "gaze duration percentage" and the "gaze frequency percentage".

[0103] A single indicator cannot fully reflect the intensity of fixation. For example, frequent but short-duration rapid squinting, or prolonged fixed gaze, both represent high attention levels, thus requiring a combined calculation. Among these, visual focus density... The calculation formula is:

[0104] ;

[0105] In the formula, The total duration of the preset time window is used to normalize the dwell time and eliminate the dimensional differences caused by different statistical durations (such as unifying the observation period of different durations to the 0-1 interval).

[0106] This is the total number of sampled frames within a preset time window, used to normalize the number of fixations, representing the probability of fixation occurring, and avoiding counting bias caused by high-frequency sampling devices.

[0107] and These are preset weighting coefficients used to adjust the relative importance of duration and frequency in the overall density calculation, adapting to the visual habits of different users. Users who require close observation can increase these coefficients. For users who browse globally, the setting can be increased. .

[0108] The gaze tracking data is processed into gaze focus coordinates on a two-dimensional coordinate plane.

[0109] The coordinates of the gaze focus are mapped onto the user's facial surface to generate a digital facial model with superimposed gaze focus density distribution, thereby constructing an interactive model that includes the digital facial model and the smart mirror model.

[0110] This interactive model places the user's virtual facial image and the application window on the screen in the same reference frame, enabling intuitive calculation of the spatial interference relationship between the two.

[0111] To perform a detailed assessment of different areas of the face, the continuous facial model needs to be discretized. A sampling density is set, and digitized facial surface mesh data is acquired. The total number of sight sampling points is determined based on the total area of ​​the surface mesh data and the sampling density.

[0112] Please see Figure 2 The facial mesh represents the curved topology of the digital facial model in wireframe form, and the sampling points are evenly distributed on the facial surface in the form of a dot matrix, with higher sampling density in detailed areas such as the eyes and nose.

[0113] The magnified section in the upper right corner shows the principle of pupil-corneal reflex gaze tracking, which is to calculate the direction of gaze by tracking the relative position change between the center of the pupil and the corneal reflex point (Purkin spot).

[0114] The gaze focus mapping represents the coordinate mapping relationship between the three-dimensional gaze point and the two-dimensional facial surface.

[0115] Based on the Poisson disk distribution algorithm, a total number of sampling points are generated uniformly and iteratively on the surface grid data. This algorithm can ensure that the distance between any two sampling points is kept to a minimum, avoid the evaluation bias caused by the local clustering of sampling points, achieve uniform coverage of complex areas such as the nose and the corner of the eye, and establish a mapping relationship between sampling points and the coordinates of the gaze focus.

[0116] Since users' natural line of sight varies under different lighting conditions, direct comparison can lead to misjudgment. Therefore, it is necessary to find a historical benchmark that is most similar to the current environment.

[0117] Obtain the values ​​of various brightness factors of the smart mirror at the current moment. And the values ​​of various brightness factors under the environmental parameters in each interaction record in the historical interaction log. .

[0118] Substitute into the formula to calculate the difference coefficient of each interaction record. :

[0119] ;

[0120] In the formula, This represents the number of terms in the luminance factor, signifying the total number of illuminance parameters (e.g., including illuminance and color temperature). ), used to define the range of summation calculations.

[0121] For the first The weighting coefficient of the luminance factor is used to determine the importance of the corresponding luminance factor in the overall illumination evaluation, and is used to make distance calculation more consistent with the characteristics of human vision (such as giving higher weight to color temperature and illuminance).

[0122] The historical interaction logs are filtered to identify interaction records with a difference coefficient less than a preset threshold, which means filtering out historical scenes with highly similar lighting environments.

[0123] Based on the gaze focus density distribution of each interaction record, the gaze focus density of each sampling point is obtained and used as training data for regression fitting.

[0124] Data pairs consisting of gaze focus density and brightness factor at the same moment in the interaction record are extracted as sample data segments. For each sampling point, a gaze focus relationship between gaze focus density and brightness factor is fitted. Specifically:

[0125] The number of sample data segments in the interaction record for statistical sampling points The instantaneous values ​​of various brightness factors at the same time in each sample data segment are extracted as independent variables, and the line-of-sight focus density is used as the dependent variable and packaged into training samples.

[0126] Establish a linear regression model, setting the initial intercept and regression coefficients. Then... Using all independent variables in the training samples as input data, and given the parameters of intercept and regression coefficients, the model outputs the predicted gaze density, which is the gaze density that the user should have under the given illumination.

[0127] The difference between the dependent variable and the expected focal density of the line of sight in each training sample is calculated as the line of sight gap coefficient. The line of sight gap coefficients of all training samples are summed to obtain the line of sight gap index.

[0128] Using the line-of-sight gap index as the objective function for hyperparameter optimization, the intercept and regression coefficients are iteratively adjusted. The intercept and regression coefficient values ​​corresponding to the minimum line-of-sight gap index are then fixed to obtain the fitted line-of-sight focus relationship.

[0129] By obtaining the brightness factor in the environmental parameters at the current moment and substituting it into the gaze focus relationship of each sampling point, the expected gaze focus density is obtained, which in turn calculates the user's attention distribution in the current natural state when there is no application window occlusion interference.

[0130] Please see Figure 3 The left side shows the expected density distribution in its natural state. The focus of vision is mainly concentrated on the eyes and lips, presenting a natural distribution that conforms to human eye observation habits. The density is lower in areas such as the cheeks.

[0131] The right side shows the actual density distribution after occlusion. The area on the right cheek that is occluded by the window shows a clear high-density abnormal cluster, with a significantly darker color. This reflects the user's visual compensation behavior of repeatedly trying to observe the area due to occlusion. This abnormal area is the occlusion risk point identified by the system.

[0132] In real-world scenarios, when an application window obscures a part of a user's face (such as the corner of the mouth), the user will instinctively and repeatedly try to look at the obscured area. This visual compensation behavior of "trying to penetrate the obscuration" will cause the actual visual density in that area to spike abnormally, far exceeding the natural prediction value, thus accurately identifying it as a risk point.

[0133] Therefore, the current line-of-sight focus density of each sampling point is obtained, and sampling points where the difference between the current line-of-sight focus density and the expected line-of-sight focus density is greater than the risk threshold are identified as occlusion risk points.

[0134] Identify the projection overlap between application software windows and occlusion risk points in the interaction model, associate application software windows with projection overlap with occlusion risk points, and thus pinpoint the specific interference source causing visual abnormalities.

[0135] Calculate the shortest projection distance between interconnected application software windows and occlusion risk points, and combine this with software parameters to calculate the visual interference efficiency index of the application software windows.

[0136] The Visual Interference Efficiency Index is used to evaluate the visual interference efficiency of application software windows on points of occlusion risk (i.e., areas that are obstructed and attract the user's unusual attention). It comprehensively evaluates window size, distance, facial posture, and the visual prominence of the window.

[0137] Visual interference efficiency index The calculation formula is:

[0138] ;

[0139] In the formula, This is the facial sensitivity weighting coefficient; it represents the importance of the facial area where the occlusion risk point is located (e.g., the eye area has a greater weight than the cheek), and is used to reflect the difference in the impact of the location of occlusion on the visual experience.

[0140] The larger the area of ​​the application software window, the wider the obstruction range. The square relationship is used to enhance the visual obstruction effect of large windows.

[0141] The distance value is the shortest projection distance, which is the shortest straight-line distance from the edge of the window to the point of occlusion risk. The closer the distance, the stronger the interference, and the inversely proportional the decrease. Even if the window does not completely cover the point, the adjacent bright block will still distract visual attention.

[0142] The distance smoothing constant is a very small positive number; this prevents the window from being too close to the risk point. When the denominator is zero, the calculation will crash, thus ensuring the stability of the calculation.

[0143] The facial deflection angle represents the angle at which the user's head deviates from its position facing the mirror. Cosine attenuation is introduced; the larger the deflection angle, the smaller the effective obstruction of the mirror's projection, thus reducing interference. When the user turns their face to the side, the actual projected area of ​​the window in the user's field of vision will shrink.

[0144] The contrast factor is the edge brightness of the application software window, which is the difference in brightness between the window edge and the background environment. The higher the contrast, the more dazzling the window becomes, and the stronger the interference with the viewer's gaze, exhibiting a positive enhancement effect.

[0145] Calculate the visual interference efficiency index of each application window that is associated with the occlusion risk point, and substitute it into the formula to calculate the occlusion index between each application window and the occlusion risk point:

[0146] ;

[0147] In the formula, The occlusion index serves as a comprehensive risk rating for the current window causing visual obstruction; it triggers the obstacle avoidance threshold and determines the intensity of subsequent dimensionality reduction processing.

[0148] The hyperbolic tangent activation function acts as a nonlinear mapping function, compressing the linear combination of input values ​​into the (0,1) interval to prevent excessively large values ​​from overflowing, thus making the exponential output smooth and standardized.

[0149] It represents the interference sensitivity gain, serving as an amplification factor for physical interference; it is used to adjust the weighting of physical interference in the final decision.

[0150] The difference in visual focus density represents the difference between the current actual visual density and the predicted natural visual density. It reflects the degree of abnormal visual focus caused by occlusion, indicating that the occlusion is more severe.

[0151] This is the density difference normalization coefficient, used as a scaling factor for the density difference value; it is used to adjust the density difference value to a scale that is consistent with the normalization coefficient. Having the same order of magnitude ensures the mathematical validity of adding the two together.

[0152] As a negative exponential decay function, it achieves the effect of forgetting decay over time.

[0153] The forgetting decay rate is a constant used to control the decay rate; it determines how quickly the user's "tolerance" for the window being obscured increases over time.

[0154] This refers to the idle time of an application window without interaction, which is the time during which the user does not interact with the application window.

[0155] In the formula, the longer the window is idle, the more it indicates that the user does not currently need the window, and the more intolerable its occlusion becomes. This is mitigated by a negative exponential term. Values ​​that approach the maximum value (highest risk) are prioritized for cleanup or reduction.

[0156] For example, if a weather window that has been running for a long time blocks the eyelash area that the user is viewing, it will be judged as extremely high risk and dealt with first.

[0157] The line-of-sight focus density difference refers to the difference between the current line-of-sight focus density and the expected line-of-sight focus density.

[0158] Calculate the final occlusion index of the application software window. Physical visual interference of the merged window ( ), the degree of abnormal focus of the user's gaze ( ), and introduce a time-based forgetting decay mechanism.

[0159] Application windows with an occlusion index greater than a preset occlusion threshold are selected as high-risk occlusion windows.

[0160] Obtain the two-dimensional planar coordinates of the occlusion risk points associated with the high-risk occlusion window.

[0161] The vector from the center point coordinates of the high-risk obstruction window to the two-dimensional plane coordinates of the obstruction risk point is used as the risk direction vector, and the opposite direction of the risk direction vector is used as the obstacle avoidance direction, that is, the control window moves away from the area that the user is highly concerned about.

[0162] The occlusion index is multiplied by a preset step size coefficient to calculate the displacement scalar; the more severe the occlusion, the farther the movement distance. The obstacle avoidance direction is multiplied by the displacement scalar to obtain the spatial obstacle avoidance displacement vector.

[0163] The current spatial coordinates of high-risk occlusion windows are updated based on the spatial obstacle avoidance displacement vector. It is then determined whether the updated current spatial coordinates exceed the display boundary of the smart mirror display to prevent the window from being moved off the screen and becoming inoperable.

[0164] If the window does not exceed the display boundary, re-render the high-risk occlusion window according to the updated current spatial coordinates.

[0165] If the window exceeds the display boundary, the distance beyond the display boundary is multiplied by a preset scaling factor to analyze the size scaling ratio. The current display size is updated according to the size scaling ratio, and the high-risk occlusion window is re-rendered. That is, when it is impossible to move the window to avoid obstacles, the window volume is reduced to make room for visual space.

[0166] The occlusion index of the high-risk occlusion window after rendering is recalculated. If the occlusion index is still greater than the preset safety threshold, the opacity of the high-risk occlusion window is dynamically reduced according to the occlusion index until the occlusion index is no greater than the preset safety threshold. As a last resort for spatial obstacle avoidance failure, the window is made semi-transparent, allowing the user's line of sight to penetrate the window and see the facial details behind it, achieving seamless optimization.

[0167] The present invention also provides a data processing-based intelligent mirror control system, comprising:

[0168] The data acquisition module is used to collect environmental parameters, user characteristics, software parameters, and historical interaction logs when the smart mirror is in use, and to generate an interaction model that includes the digital face and the smart mirror.

[0169] The intelligent analysis module is used to set sampling points on the digital facial surface, analyze historical interaction logs to filter interaction records, and fit the gaze focus relationship for the sampling points based on the interaction records, thereby setting occlusion risk points.

[0170] The window management module is used to identify and associate application software windows with overlapping projections and occlusion risk points in the interaction model, analyze the shortest projection distance between them, calculate the visual interference efficiency index and occlusion index of each application software window, and filter high-risk occlusion windows.

[0171] The optimization processing module is used to calculate the spatial obstacle avoidance displacement vector and size scaling ratio of high-risk occlusion windows based on the occlusion index, update the software parameters based on the spatial obstacle avoidance displacement vector and size scaling ratio, and control the smart mirror display to re-render the high-risk occlusion windows until the occlusion index is lower than the preset safety threshold.

[0172] Example 1: In this example, the display panel of the smart mirror has a size of 800mm × 600mm and a resolution of 1920 × 1440 pixels. On the surface of the digital facial model, a total of 486 sampling points are generated based on the Poisson disk distribution algorithm at a density of approximately 2 sampling points per square centimeter. Among them, the sampling density in the eye area is increased to 4 sampling points per square centimeter, totaling 72 sampling points, to ensure precise monitoring of the core area of ​​visual function.

[0173] Sampling points cover all major facial areas, including the forehead, brow bone, around the eyes, bridge of the nose, nostrils, cheeks, corners of the mouth, and chin. The distance between any two sampling points is no less than 8mm to avoid evaluation bias caused by local clustering.

[0174] Eye tracking employs the pupil-corneal reflection method, using an infrared supplemental light to create a Pullman spot reflection point on the user's cornea, and acquiring eye images at a sampling frequency of 30Hz. Preset time window. Set to 3 seconds, corresponding to the total number of sampling frames. It is 90 frames per second.

[0175] Weighting coefficient Take 0.6, A value of 0.4 indicates that gaze duration accounts for a higher proportion in the calculation of visual focus density, which is suitable for user scenarios where users primarily observe closely when looking in the mirror.

[0176] For environmental parameter acquisition, this embodiment selects two brightness factors: illuminance and color temperature. Illuminance is weighted at 0.35, and color temperature at 0.65, with color temperature given a higher weight to align with the human eye's greater sensitivity to color changes. The difference coefficient is preset to a threshold of 150 to filter historical interaction records with similar lighting environments.

[0177] The study used a typical scenario of observing facial skincare routines performed by users in a bathroom environment. Under warm light conditions with an ambient illuminance of 320 lux and a color temperature of 3000 K, 28 historical records with a difference coefficient less than 150 were selected from historical interaction logs. Linear regression fitting was performed with the gaze focus density at each sampling point as the dependent variable and two brightness factors as independent variables to obtain the gaze focus relationship for each sampling point.

[0178] Substituting the current brightness factor into the various relationships, the expected focal density distribution under natural conditions is calculated. The expected density for the eye area is approximately 0.28 to 0.35, for the lip area it is approximately 0.22 to 0.26, and for the cheek area it is approximately 0.08 to 0.12, presenting an overall natural attention distribution centered on the facial features.

[0179] At this moment, the weather information application window running on the smart mirror is located on the right side of the mirror, with a display area of ​​[missing information]. It measures 12,000 square millimeters, approximately 2.5% of the total mirror area, with a brightness contrast factor at the window edge. The value is 0.45. The left edge of the window exactly covers the corresponding area of ​​the user's right cheek, with the shortest projection distance being 0.45. A value of 0 indicates that the window edge coincides with the point of potential obstruction. Facial deflection angle. The angle is 5 degrees, and the user is basically facing the mirror. Facial sensitivity weighting coefficient. The value is 0.7 in the cheek region, representing the distance smoothing constant. Take 0.01.

[0180] Substituting into the formula for the visual interference efficiency index, we can obtain:

[0181] ;

[0182] Calculated Approximately Interference sensitivity gain Pick Density difference normalization coefficient Take 0.15, forgetting decay rate Take 0.12. Since the window has been idle for 42 seconds without any activity, that is... =42 seconds, and the difference in visual focus density in the obscured cheek area was actually detected. The value is approximately 0.31, which is far higher than the risk threshold of 0.15, and is therefore identified as an occlusion risk point.

[0183] Substitute into the formula to calculate: The occlusion index is approximately 0.018. Due to the long idle time, the negative exponent keeps the occlusion index at a low level. However, combined with the occlusion threshold set at 0.01, the occlusion index of this window is still higher than the occlusion threshold, and therefore it is judged as a high-risk occlusion window.

[0184] A three-level obstacle avoidance mechanism is activated. The first level performs spatial translation obstacle avoidance, using the direction opposite to the occlusion risk point from the window center as the obstacle avoidance direction. The step size is set to 80 pixels, and the occlusion index of 0.018 multiplied by the step size yields a displacement scalar of approximately 1.44 pixels. Considering the actual displacement is too small, the system adopts a minimum displacement step size of 20 pixels, translating the window 20 pixels to the right of the mirror image. After translation, the occlusion index is recalculated. Due to the increased shortest projection distance, the occlusion index drops to approximately 0.005, which is below the safety threshold of 0.01, and obstacle avoidance is complete.

[0185] If the window touches the right edge of the display after panning, a second level of scaling is triggered. The 12-pixel distance exceeding the boundary is multiplied by a scaling factor of 0.8, resulting in a scaling ratio of approximately 96%, and the window shrinks accordingly. If the occlusion index is still higher than the safety threshold after shrinking, a third level of transparency adjustment is further triggered, linearly reducing the window opacity according to the proportion of the occlusion index exceeding the threshold until it drops to around 60%, allowing the user's line of sight to penetrate the window and observe the facial image behind them.

[0186] Example 2: The experimental environment was set as a simulated bathroom scene. A smart mirror was installed on the wall, and an adjustable lighting system was set in front of it, which could independently control the illuminance and color temperature. Twenty subjects with normal vision were recruited, aged between 22 and 45 years old, with 10 men and 10 women.

[0187] Each subject was tested under three lighting conditions in turn: low-illuminance warm light environment (illuminance 80 lux, color temperature 2700 K), medium-illuminance neutral light environment (illuminance 450 lux, color temperature 4000 K), and high-illuminance cool light environment (illuminance 1800 lux, color temperature 6500 K).

[0188] Under each lighting condition, the system randomly popped up application windows at different locations on the mirror surface. The window area ranged from 5% to 25% of the total mirror surface area, with a total of 60 occluded scenarios and 40 unoccluded scenarios set up. Subjects were asked to perform normal facial observation actions, including checking makeup, observing skin details, and adjusting their hairstyles.

[0189] Occlusion identification was performed using both the method of this invention and the fixed threshold method of existing technologies. The existing method uses a fixed line-of-sight density threshold of 0.2 as the judgment criterion, without considering changes in ambient lighting or historical habitual benchmarks.

[0190] In low-light, warm-light environments, due to the overall dimness, users' natural gaze is relatively concentrated. Existing methods achieve an occlusion recognition accuracy of approximately 71.3% and a false positive rate of approximately 18.5%, primarily due to misinterpreting normal, focused observation as occlusion. In contrast, the method of this invention establishes a dynamic benchmark through historical lighting matching, achieving an occlusion recognition accuracy of 94.7% and reducing the false positive rate to 3.2%, representing an improvement of over 23 percentage points compared to existing technologies.

[0191] Under moderate illumination and neutral light conditions, existing methods achieve an accuracy of approximately 78.6% and a false positive rate of approximately 12.8%. The method of this invention achieves an accuracy of 96.2% and a false positive rate of 2.5%.

[0192] In high-illuminance, cool-light environments, with ample ambient light, users' line-of-sight distribution is more dispersed. Existing methods struggle to adapt to baseline drift, resulting in an accuracy rate dropping to 65.9% and a false positive rate rising to 24.3%. The method of this invention adaptively adjusts the predicted density through a dynamically fitted line-of-sight focus relationship, maintaining an accuracy rate of 93.5% and a false positive rate of only 4.1%, demonstrating a significantly greater advantage.

[0193] Statistical analysis under three lighting conditions showed that the average occlusion recognition accuracy of the method of this invention was 94.8%, and the average false positive rate was 3.3%; while the average accuracy of existing methods was 71.9%, and the average false positive rate was 18.5%. This invention improves the recognition accuracy by approximately 22.9 percentage points and reduces the false positive rate by approximately 15.2 percentage points, fully demonstrating that the technical solution provided by this invention can effectively adapt to complex and changing environmental lighting conditions, significantly improving the accuracy and reliability of occlusion recognition.

[0194] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0195] Finally, it should be noted that the above descriptions are merely preferred embodiments 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 foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A smart mirror control method based on data processing, characterized in that: The method includes: When the S100 and smart mirror are in use, environmental parameters, user characteristics, software parameters, and historical interaction logs are collected; an interaction model containing a digital face and smart mirror is generated based on the user characteristics. S200: Set sampling points on the digital facial surface, analyze environmental parameters and historical interaction logs to filter interaction records; fit the gaze focus relationship formula for the sampling points based on the interaction records and calculate the expected gaze focus density; and take the sampling points where the difference between the current gaze focus density and the expected gaze focus density is greater than the risk threshold as occlusion risk points. S300: Identify and associate application software windows with overlapping projections and occlusion risk points in the interaction model, analyze the shortest projection distance between them and calculate the visual interference efficiency index of each application software window, calculate the occlusion index by combining the difference between the current visual focus density and the expected visual focus density, and screen out high-risk occlusion windows. S400 calculates the spatial obstacle avoidance displacement vector and size scaling ratio of high-risk occlusion windows based on the occlusion index, thereby updating the software parameters and controlling the smart mirror display to re-render the high-risk occlusion windows.

2. The intelligent mirror control method based on data processing according to claim 1, characterized in that: User features include facial features and facial rotation angles captured by a depth camera; The historical interaction log includes records of interactions performed by the smart mirror each time it runs under different environmental parameters. Each interaction record includes environmental parameters and user characteristics; the user refers to the user of the smart mirror. Environmental parameter data includes different brightness factors, which are physical indicators that quantify the lighting conditions of the environment in which the smart mirror is located; The software parameters include the display area of ​​the application window, edge brightness and contrast factor, idle time without interaction, current spatial coordinates, and current display size.

3. The intelligent mirror control method based on data processing according to claim 2, characterized in that: The digital display panel of the smart mirror is used as a two-dimensional coordinate plane; By analyzing the eye images in facial features, the relative vector between the pupil center and the corneal reflection point is extracted based on the pupil-corneal reflection method, and the gaze direction vector is calculated. The gaze direction vector is intersected with the two-dimensional coordinate plane of the smart mirror to obtain the gaze landing point coordinate sequence as gaze tracking data; The gaze focus density distribution is obtained by analyzing user characteristics of each interaction record in the historical interaction log. The gaze focus density distribution refers to the duration of time that the user's gaze lingers on different positions on the surface of the digital facial model within a preset time window. and number of fixations The normalized value distribution; Eye Focus Density The calculation formula is: ； In the formula, The total duration of the preset time window; The total number of sampling frames within the preset time window; and These are preset weighting coefficients; The gaze tracking data is processed into gaze focus coordinates on a two-dimensional coordinate plane; the gaze focus coordinates are mapped onto the user's facial surface to generate a digital facial model with superimposed gaze focus density distribution, thereby constructing an interactive model that includes the digital facial model and the smart mirror model.

4. The intelligent mirror control method based on data processing according to claim 3, characterized in that: S200 includes: S201. Set the sampling density, acquire digital facial surface mesh data, and determine the total number of line-of-sight sampling points based on the total area and sampling density of the surface mesh data. S202. Based on the Poisson disk distribution algorithm, a total number of sampling points are uniformly and iteratively generated on the surface grid data; a mapping relationship between the sampling points and the coordinates of the viewpoint is established. S203. Obtain the values ​​of various brightness factors of the smart mirror at the current moment. And the values ​​of various brightness factors under the environmental parameters in each interaction record in the historical interaction log. ; Substitute into the formula to calculate the difference coefficient of each interaction record. : ； In the formula, The number of terms in the brightness factor; For the first The weighting coefficient of the brightness factor; Filter out interaction records with a difference coefficient less than a preset threshold from the historical interaction logs; obtain the gaze focus density of each sampling point based on the gaze focus density distribution of each interaction record; S204. Extract the data pairs consisting of the gaze focus density and brightness factor at the same moment in the interaction record as sample data segments, and fit the gaze focus relationship between the gaze focus density and brightness factor for each sampling point. S205. Obtain the brightness factor in the environmental parameters at the current moment, and substitute it into the line-of-sight focus relationship of each sampling point to obtain the expected line-of-sight focus density; S206. Obtain the current line-of-sight focus density of each sampling point, and designate sampling points whose difference between the current line-of-sight focus density and the expected line-of-sight focus density is greater than the risk threshold as occlusion risk points.

5. The intelligent mirror control method based on data processing according to claim 4, characterized in that: S204 includes: S2041, Number of sample data segments in the interaction record for statistical sampling points The instantaneous values ​​of various brightness factors at the same time in each sample data segment are extracted as independent variables, and the line-of-sight focus density is used as the dependent variable and packaged into training samples. S2042. Establish a linear regression model, setting the initial intercept and regression coefficients; [This...] Using all independent variables from the training samples as input data, and given the parameters of intercept and regression coefficients, the model outputs the predicted line-of-sight focus density. S2043. Calculate the difference between the dependent variable and the expected visual focus density in each training sample as the visual gap coefficient. Sum the visual gap coefficients of all training samples to obtain the visual gap index. S2044. Using the line-of-sight gap index as the objective function for hyperparameter optimization, the intercept and regression coefficients are iteratively adjusted. S2045. By taking the values ​​of the intercept and regression coefficient corresponding to the minimum fixed line-of-sight gap index, the fitted line-of-sight focus relationship is obtained.

6. The intelligent mirror control method based on data processing according to claim 4, characterized in that: The S300 includes: S301. Identify the projection overlap between application software windows and occlusion risk points in the interaction model, and associate application software windows with projection overlap relationships with occlusion risk points. S302. Calculate the shortest projection distance between the interrelated application software windows and the occlusion risk points, and calculate the visual interference efficiency index of the application software windows in combination with the software parameters. S303. Based on the visual interference efficiency index and the difference in visual focus density, calculate the occlusion index of each application software window, and select application software windows with an occlusion index greater than the preset occlusion threshold as high-risk occlusion windows.

7. The intelligent mirror control method based on data processing according to claim 6, characterized in that: The formula for calculating the visual interference efficiency index is: ； In the formula, The visual interference efficiency index; This is the facial sensitivity weighting coefficient; The display area of ​​the application software window; This is the distance value for the shortest projection distance; This is the distance smoothing constant; For facial deflection angle; The edge brightness contrast factor of the application software window; Calculate the visual interference efficiency index of each application window that is associated with the occlusion risk point, and substitute it into the formula to calculate the occlusion index between each application window and the occlusion risk point: ； In the formula, For occlusion index; It is the hyperbolic tangent activation function; For interference sensitivity gain; This represents the difference in focal density of the line of sight. This is the density difference normalization coefficient; Forgetting decay rate; The idle time of an application window without interaction; The line-of-sight focus density difference refers to the difference between the current line-of-sight focus density and the expected line-of-sight focus density.

8. The intelligent mirror control method based on data processing according to claim 6, characterized in that: The S400 includes: Obtain the two-dimensional planar coordinates of the occlusion risk points associated with the high-risk occlusion windows; The vector from the center point coordinates of the high-risk occlusion window to the two-dimensional plane coordinates of the occlusion risk point is taken as the risk direction vector, and the opposite direction of the risk direction vector is taken as the obstacle avoidance direction. The occlusion index is multiplied by a preset step size coefficient to calculate the displacement scalar, and the obstacle avoidance direction is multiplied by the displacement scalar to obtain the spatial obstacle avoidance displacement vector. Update the current spatial coordinates of the high-risk occlusion window based on the spatial obstacle avoidance displacement vector, and determine whether the updated current spatial coordinates exceed the display boundary of the smart mirror display. If the high-risk occlusion window does not exceed the display boundary, re-render the high-risk occlusion window according to the updated current spatial coordinates; If the window exceeds the display boundary, the distance beyond the display boundary is multiplied by a preset scaling factor to analyze the size scaling ratio. The current display size is then updated based on the size scaling ratio, and the high-risk occlusion window is re-rendered. Recalculate the occlusion index of the high-risk occlusion window after rendering. If the occlusion index is still greater than the preset safety threshold, dynamically reduce the opacity of the high-risk occlusion window according to the occlusion index until the occlusion index is no greater than the preset safety threshold.

9. A data processing-based intelligent mirror control system, applied to the data processing-based intelligent mirror control method as described in claim 1, characterized in that: The system includes: The data acquisition module is used to collect environmental parameters, user characteristics, software parameters, and historical interaction logs when the smart mirror is in use, and to generate an interaction model that includes the digital face and the smart mirror. The intelligent analysis module is used to set sampling points on the digital facial surface, analyze historical interaction logs to filter interaction records, and fit the gaze focus relationship for the sampling points based on the interaction records, thereby setting occlusion risk points. The window management module is used to identify and associate application software windows with overlapping projections and occlusion risk points in the interaction model, analyze the shortest projection distance between them, calculate the visual interference efficiency index and occlusion index of each application software window, and screen high-risk occlusion windows. The optimization processing module is used to calculate the spatial obstacle avoidance displacement vector and size scaling ratio of high-risk occlusion windows based on the occlusion index, update the software parameters based on the spatial obstacle avoidance displacement vector and size scaling ratio, and control the smart mirror display to re-render the high-risk occlusion windows until the occlusion index is lower than the preset safety threshold.

Images