Naked eye 3D display optimization method based on real-time eyeball tracking

Abstract

The invention discloses a naked-eye 3D display optimization method based on real-time eyeball tracking, and particularly relates to the technical field of naked-eye 3D display, and the method comprises the steps: capturing eyeball movement data in real time through a visual angle tracking sensor, and collecting illumination information in combination with an ambient light sensor; depth perception parameters are calculated based on pupil diameter variation and frequency, and a time sequence prediction type dynamic compensation coefficient is generated in combination with eyeball movement acceleration; acquiring an initial fixation point coordinate by using an improved spherical projection mapping model, and performing compensation coefficient correction to obtain a real-time coordinate; and finally, according to the real-time coordinates, dynamically adjusting the refractive index distribution of the nanostructure layer, the rotation angle of the polarizer, the focal length of the optical lens and other optical modulation parameters. The real-time distance is calculated through the binocular parallax algorithm, the depth mapping value is generated by combining the focal length of the camera and the baseline distance, the method can adapt to the illumination change and the user view angle, and the 3D display effect is optimized.

Description

Technical Field

[0001] The present invention relates to the technical field of naked-eye 3D display technology, and more specifically, to a naked-eye 3D display optimization method based on real-time eye tracking. Background Art

[0002] In existing naked-eye 3D display technology, fixed grating or lens array solutions cannot dynamically respond to user position deviations, resulting in ghosting and reduced stereoscopic perception. The lack of ambient light adaptation significantly reduces display quality in strong or low-light environments. Its workflow pre-sets immutable optical modulation parameters and outputs parallax images based on a single preset viewing distance. The user must be strictly within a static optimal viewing area so that each eye can receive the image signal from the corresponding perspective, passively forming a stereoscopic perception through the human eye's persistence of vision. The entire process relies entirely on fixed hardware to achieve the naked-eye 3D effect, eliminating the need for auxiliary equipment, but lacks dynamic adjustment capabilities.

[0003] However, it still has some shortcomings in actual use, such as the inability to track the user's eye movement and position changes in real time. When the user deviates from the preset viewing area, the parallax images received by the left and right eyes are easily misaligned, resulting in ghosting or stereoscopic attenuation; the optical modulation parameters are fixed and cannot adapt to the ambient light intensity and spectral changes. In strong or low light environments, the display contrast decreases and details are lost, reducing the robustness of the visual experience; the real-time impact of dynamic changes in pupil diameter and eye movement acceleration is ignored, and depth perception only relies on static parallax algorithms, which are insufficient in accuracy. Long-term viewing causes visual fatigue, and it cannot adapt to the user's individual visual characteristics. Summary of the Invention

[0004] In order to overcome the above-mentioned defects of the prior art, the present invention provides a naked-eye 3D display optimization method based on real-time eye tracking, which solves the problems raised in the above-mentioned background technology through the following scheme.

[0005] To achieve the above objectives, the present invention provides the following technical solution: a naked-eye 3D display optimization method based on real-time eye tracking, comprising: S1: Detect the user's face through a view tracking sensor and capture eye movement data in real time; collect the light intensity and spectral distribution of the display scene through an ambient light sensor; the eye movement data includes the pupil center position, pupil diameter, and corneal reflection point coordinates; S2: Calculate depth perception parameters based on pupil diameter change and change frequency in consecutive image frames, and generate a dynamic compensation coefficient matrix based on eye movement acceleration; S3: According to the pupil center position and the corneal reflection point coordinates, the initial gaze point coordinates are obtained through a pre-trained long short-term memory neural network model; The initial gaze point coordinates are combined with the dynamic compensation coefficient matrix to correct the initial coordinates to obtain the real-time coordinates on the display panel; S4: Input the real-time coordinates, depth perception parameters, and light intensity and spectral distribution collected by the ambient light sensor into the multi-parameter arbitration controller, and generate a hardware execution sequence according to the following priority: A1: If the eye movement acceleration is greater than the preset threshold, the optical lens focus adjustment is performed first; A2: If the ambient light intensity is greater than the preset threshold, the polarizer rotation angle adjustment is performed first; A3: If the pupil diameter change rate is greater than a preset threshold, the refractive index adjustment of the nanostructure layer is performed first; A4: When multiple conditions are met at the same time, serial control is executed in sequence; S5: Dynamically adjust the optical modulation parameters of the naked-eye 3D display device according to the arbitration result.

[0006] Preferably, the eye movement data includes: the view angle tracking sensor uses the grayscale difference between the pupil area and the iris area to separate the pupil area using a threshold segmentation algorithm, and performs ellipse fitting on the area to determine the pupil center position and pupil diameter; The view tracking sensor forms a reflection point on the corneal surface through a built-in infrared light source, uses binocular vision or multiple cameras to synchronously capture the reflection point image, combines the triangulation principle to calculate the three-dimensional coordinates of the reflection point, and eliminates noise fluctuations through filtering to obtain stable corneal reflection point coordinates.

[0007] Preferably, the depth perception parameters and the dynamic compensation coefficient matrix are specifically analyzed as follows: Performing time-attenuated weighting on pupil diameter changes over multiple consecutive frames and summing the results to obtain a weighted pupil diameter change; calculating a frequency of pupil diameter changes between the multiple consecutive frames; and combining the weighted pupil diameter change with the change frequency according to a preset weight ratio to obtain the depth perception parameter; The dynamic compensation coefficient matrix is ​​generated by superimposing a correction term on a unit matrix, and the correction term is calculated by coupling the depth perception parameter with each component of the eye movement acceleration according to a preset axial sensitivity coefficient.

[0008] Preferably, the real-time coordinates are specifically analyzed including: The input of the long short-term memory neural network model is a normalized feature vector formed by splicing the pupil center position coordinates and corneal reflection point coordinates captured in real time; The real-time coordinates are obtained by applying the dynamic compensation coefficient matrix to the initial gaze point coordinates output by the long short-term memory neural network model.

[0009] Preferably, the priorities include: When the eye movement acceleration is greater than a preset threshold, the optical lens focal length is adjusted first. The preset threshold is dynamically set based on the matching relationship between the human eye's persistence of vision and the display system's response time. When the ambient light intensity is greater than a preset threshold, the polarizer rotation angle is adjusted first. The preset threshold is set according to the minimum readable brightness requirement of the naked-eye 3D display device in an outdoor environment. When the pupil diameter change rate is greater than a preset threshold, the refractive index of the nanostructure layer is adjusted preferentially. The preset threshold is set based on the physiological adjustment characteristics of the human eye and the system noise tolerance.

[0010] Preferably, the dynamically adjusting optical modulation parameters includes: When performing optical lens focus adjustment, the focus adjustment amount is positively correlated with the amplitude of eye movement acceleration, and a nonlinear function is used to limit its maximum adjustment amplitude; When performing polarizer rotation angle adjustment, the rotation angle adjustment amount is linearly proportional to the ambient light intensity and includes a base bias angle; When the refractive index of the nanostructure layer is adjusted, the refractive index adjustment amount is positively correlated with the absolute value of the pupil diameter change rate, and a nonlinear saturation function is used to limit the adjustment range to between plus and minus 0.2; When multiple adjustments need to be performed, each adjustment is performed in sequence according to the order specified in S4, and the final optical modulation parameter value is the cumulative result of each adjustment amount.

[0011] Preferably, the light intensity and spectral distribution are specifically analyzed including: The ambient light sensor converts the light intensity into an electrical signal through a photodiode, and obtains the light intensity value through analog-to-digital conversion; The ambient light sensor decomposes the collected light into multiple spectral channels according to wavelength through a built-in micro-spectrometer, calculates the light intensity ratio of each channel, forms a spectral distribution curve, and obtains spectral distribution data.

[0012] Preferably, the pre-trained long short-term memory neural network model includes: Collect continuous time series of pupil center position coordinates, corneal reflection point coordinates and corresponding actual gaze point coordinate data, and construct training samples after filtering, denoising and splicing; A multi-layer long short-term memory network structure was used, with mean square error as the loss function and Adam optimizer for training until the model's gaze point coordinate prediction error on the validation set was less than 0.5 mm.

[0013] The technical effects and advantages of the present invention are as follows: 1. Through real-time eye tracking and ambient light perception, the optical modulation parameters are dynamically adjusted to solve the ghosting and stereoscopic attenuation problems caused by user position deviation in traditional naked-eye 3D technology. 2. Combined with ambient light sensor data, the display parameters are dynamically optimized based on light intensity and spectral distribution. This enhances contrast in bright light environments and preserves details in low-light environments, overcoming the display quality degradation caused by traditional fixed optical parameters. 3. Generate depth perception parameters and dynamic compensation coefficients based on pupil diameter change, change frequency, and eye movement acceleration, predict gaze point coordinates through LSTM neural network, and optimize 3D depth of field performance in combination with real-time compensation mechanism. 4. A multi-parameter arbitration controller implements priority scheduling based on preset thresholds such as eye movement acceleration and pupil diameter change rate to avoid parameter conflicts. A serial execution mechanism ensures orderly and efficient hardware adjustments, improving system responsiveness. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 It is a schematic diagram of the overall structure of the present invention.

[0015] Figure 2 Schematic diagram of the eye movement data capture principle of the present invention.

[0016] Figure 3 This is a logic diagram of the multi-parameter arbitration controller of the present invention. DETAILED DESCRIPTION

[0017] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0018] refer to Figure 1-Figure 3 A naked-eye 3D display optimization method based on real-time eye tracking is shown, comprising: S1: Detect the user's face through a view tracking sensor and capture eye movement data in real time; collect the light intensity and spectral distribution of the display scene through an ambient light sensor; the eye movement data includes the pupil center position, pupil diameter, and corneal reflection point coordinates; S2: Calculate depth perception parameters based on pupil diameter change and change frequency in consecutive image frames, and generate a dynamic compensation coefficient matrix based on eye movement acceleration; S3: According to the pupil center position and the corneal reflection point coordinates, the initial gaze point coordinates are obtained through a pre-trained long short-term memory neural network model; The initial gaze point coordinates are combined with the dynamic compensation coefficient matrix to correct the initial coordinates to obtain the real-time coordinates on the display panel; S4: Input the real-time coordinates, depth perception parameters, and light intensity and spectral distribution collected by the ambient light sensor into the multi-parameter arbitration controller, and generate a hardware execution sequence according to the following priority: A1: If the eye movement acceleration is greater than the preset threshold, the optical lens focus adjustment is performed first; A2: If the ambient light intensity is greater than the preset threshold, the polarizer rotation angle adjustment is performed first; A3: If the pupil diameter change rate is greater than a preset threshold, the refractive index adjustment of the nanostructure layer is performed first; A4: When multiple conditions are met at the same time, serial control is executed in sequence; S5: Dynamically adjust the optical modulation parameters of the naked-eye 3D display device according to the arbitration result.

[0019] S1: detecting the user's face through a viewing angle tracking sensor and capturing eye movement data in real time; collecting the light intensity and spectral distribution of the display scene through an ambient light sensor; the eye movement data includes the pupil center position, pupil diameter, and corneal reflection point coordinates; Among them, after the view tracking sensor is activated, it first uses the built-in algorithm to locate the user's face and lock the tracking range, so that it can accurately collect eye movement data.

[0020] It should be further explained that the pupil and iris areas are separated by using the grayscale difference between the pupil and the surrounding iris under infrared light, and the pupil and iris areas are separated by a threshold segmentation algorithm (such as the Otsu adaptive threshold method). The segmented pupil area is fitted with an ellipse using the least squares method, and the center coordinates of the ellipse obtained by the fitting result are the pupil center position. .

[0021] Calculate the pupil diameter based on the ellipse fitting result of the pupil area .

[0022] Among them, the human pupil may appear elliptical in the non-emmetropic state (for example, when strabismic), and the short axis direction is closer to the actual physiological diameter of the pupil. The long axis may be elongated due to the tilt of the viewing angle, resulting in a larger measurement value, while the short axis is less affected by perspective distortion and has higher accuracy; therefore, the pupil diameter is defined as the actual length of the short axis of the ellipse.

[0023] The view tracking sensor has a built-in infrared light source. The infrared light emitted by the light source illuminates the corneal surface, forming 1-4 corneal reflection points. The sensor collects reflection point images through binocular vision or multiple cameras synchronously, and calculates the three-dimensional coordinates of the reflection points based on the principle of triangulation. The collected reflection point coordinates are filtered (such as Kalman filtering) to eliminate coordinate fluctuations caused by slight eye movement or sensor noise, and retain stable corneal reflection point coordinates. .

[0024] The ambient light sensor converts the light intensity into an electrical signal through a photodiode, and obtains the light intensity value through analog-to-digital conversion. The sensor has a built-in micro-spectrometer that decomposes the collected light into multiple spectral channels according to wavelength, calculates the light intensity ratio of each channel, forms a spectral distribution curve, and obtains the spectral distribution value. .

[0025] S2: calculating depth perception parameters based on pupil diameter changes and change frequencies in consecutive image frames, and generating a dynamic compensation coefficient matrix in combination with eye movement acceleration; Set continuous The pupil diameter of the frame is , then the change in pupil diameter is: Frequency of change: Calculate depth perception parameters based on pupil diameter change and change frequency: ,in , is the weight matrix.

[0026] Dynamic compensation coefficient matrix , is the identity matrix, is the depth perception gain coefficient, Acceleration coupling coefficient matrix, represents a diagonal matrix, Eye movement acceleration scalar.

[0027] in, The sensitivity coefficients of each axis are determined through calibration experiments: the user is asked to look at a target with a changing depth gradient, the pupil diameter change rate and eye movement acceleration are recorded simultaneously, and a multivariate regression model is established to fit the coefficients of each axis.

[0028] Among them, the pupil diameter change weight matrix is: ,in is the attenuation coefficient, and ; It needs to be further explained that , The optimal value is 0.8. From the perspective of the temporal correlation of the physiological visual system, the human pupil has a short-term memory characteristic in response to depth changes. The latest pupil diameter change is most closely related to the current focus state. The attenuation coefficient of 0.8 can make the weight of each frame decay reasonably over time. In terms of balancing noise suppression and data stability, the value of 0.8 not only retains the effective information of the last 3 frames (the sum of the weights of the first 3 frames accounts for 76% of the total weight of the first 5 frames), but also effectively suppresses noise outside the 5 frames, avoiding Too small will lead to loss of effective historical information, and too large will introduce too much historical noise; pupil diameter change Reflects recent focus adjustments. Newer frames are more closely associated with current depth perception. Using exponentially decaying weights enhances the contribution of recent changes while suppressing early noise interference.

[0029] Among them, the change frequency weight matrix: ,in, is the overall scaling factor for the frequency term, and , is the frequency attenuation coefficient, and ; It should be further explained that the frequency of change Reflects the pupil adjustment speed, but its impact on depth perception is weaker than the change itself (for example, the frequency weight of fast and small jitters should be reduced). Therefore: use Overall, the contribution weight of the frequency term is reduced to avoid high-frequency noise dominating the depth parameters; at the same time Exponential decay is used to ensure that recent frequency changes are prioritized; Change in pupil diameter Directly reflects the amplitude of lens adjustment (such as the significant dilation of the pupil when looking from near to far), which is more directly related to the distance of the target depth; the frequency of change Reflects the speed of adjustment (such as jitter when quickly switching the focus target), and is more of a dynamic trend indicator of depth change, so The overall contribution of the frequency term can be made lower than that of the variation term; and The same magnitude; whether the change in pupil diameter Or change frequency , the newer the frame data in the time series, the stronger the correlation with the current depth perception (the visual system has a higher memory weight for recent dynamics). and If the magnitude difference is too large, the weight of the long-term data of the frequency term will be excessively attenuated, destroying the consistency in the time dimension. and Same level, , to ensure the consistency of time decay characteristics.

[0030] S3: obtaining the initial gaze point coordinates through a pre-trained long short-term memory neural network model according to the pupil center position and the corneal reflection point coordinates; The initial gaze point coordinates are combined with the dynamic compensation coefficient matrix to correct the initial coordinates to obtain the real-time coordinates on the display panel; Among them, the pupil center position coordinate vector captured in real time and the corneal reflection point coordinate vector Splice by dimension, splice vector , the input vector is normalized to eliminate the sensor dimension difference.

[0031] The preprocessed vector is input into the pre-trained long short-term memory neural network (LSTM), and the model is based on the hidden state of the previous moment. and the current input splicing vector Calculate the current hidden state: ; The fully connected layer outputs the initial gaze point coordinates: ,for is the LSTM hidden state, is the output weight matrix, is the bias vector; The dynamic compensation coefficient matrix Acting on the initial gaze point coordinates , get the real-time coordinates .

[0032] The purpose of normalization is to eliminate this dimensional difference and map coordinate data of different dimensions into a unified numerical range. This process makes each feature equally important in the model training and prediction process, and prevents the value of a certain dimension from dominating the model learning process due to its excessive magnitude. A suitable normalization method (such as Min-Max normalization) converts each value in the feature vector to a specific interval such as 0 to 1 or -1 to 1 according to a certain formula. The normalization calculation formula is: ,in The minimum value of the coordinate data of this dimension within the training sample or actual acquisition range; It is the maximum value of the coordinate data of this dimension within the training sample or actual collection range.

[0033] It needs to be further explained that the pre-trained long short-term memory neural network model is used to capture the temporal correlation of eye movements. During pre-training, the coordinates of the pupil center and corneal reflection point in continuous frames are first collected as input, and the actual gaze point coordinates are simultaneously recorded as labels. After filtering and denoising, they are spliced ​​into feature vectors to construct continuous time series samples; the model adopts a multi-layer LSTM structure, the input layer dimension matches the feature vector dimension, the hidden layer is used to capture the temporal dependency of eye movements, and the output layer outputs the two-dimensional initial gaze point coordinates. The mean square error is used as the loss and the Adam optimizer is used for training. The training and validation sets are divided into 8:2, and 50-200 rounds are iterated until the loss is stable, so that the validation set coordinate prediction error is less than 0.5mm, so as to accurately predict the initial gaze point.

[0034] S4: Input the real-time coordinates, depth perception parameters, and light intensity and spectral distribution collected by the ambient light sensor into the multi-parameter arbitration controller, and generate a hardware execution sequence according to the following priority: A1: If the eye movement acceleration is greater than the preset threshold, the optical lens focus adjustment is performed first; A2: If the ambient light intensity is greater than the preset threshold, the polarizer rotation angle adjustment is performed first; A3: If the pupil diameter change rate is greater than a preset threshold, the refractive index adjustment of the nanostructure layer is performed first; A4: When multiple conditions are met at the same time, serial control is executed in sequence; Among them, the real-time coordinates, depth perception parameters, and light intensity and spectral distribution collected by the ambient light sensor are input into the multi-parameter arbitration controller, and the hardware adjustment instructions of the naked-eye 3D display device are generated through the priority judgment mechanism.

[0035] The eye movement acceleration is greater than the preset threshold, that is: ,but , prioritize the optical lens focus adjustment; otherwise , no adjustment is made; The ambient light intensity is greater than the preset threshold, that is: but , polarizer rotation angle adjustment is performed first; otherwise , no adjustment is made; The pupil diameter change rate is greater than the preset threshold, that is: but , the refractive index adjustment of the nanostructure layer is performed first; otherwise , no adjustment is made; Execution sequence vector , 1 means to execute the corresponding operation, 0 means not to execute When multiple conditions are met, the serial execution order corresponds to the vector index order ( ).

[0036] Among them, based on the preset priority rules: the optical lens focal length adjustment when the eye movement acceleration is greater than the preset threshold has the highest priority, the polarizer rotation angle adjustment when the ambient light intensity is greater than the preset threshold is second, and the nanostructure layer refractive index adjustment when the pupil diameter change rate is greater than the preset threshold has a relatively low priority; by serial control in this index order, it can be ensured that when multiple conditions coexist, the hardware adjustment is carried out in an orderly manner, parameter conflicts are avoided, and efficient optimization of the optical modulation parameters of the naked-eye 3D display device is guaranteed.

[0037] in, , obtained through statistics of 50 sets of eye movement experiments; , is the minimum readable brightness requirement for naked-eye 3D displays in outdoor daylight; , is the optimal solution to balance physiological characteristics, system noise tolerance and hardware reliability, ensuring that key optical parameter optimization is triggered only when the user actively adjusts depth perception.

[0038] It needs to be further explained that , represents the total angular acceleration of eye movement; Indicates the status of the corresponding adjustment instruction. Through a clear priority arbitration mechanism, it converts complex multi-parameter inputs into orderly hardware adjustment instructions, avoiding disordered operations when multiple factors conflict, and ensuring that the naked-eye 3D display device can quickly and accurately optimize optical parameters according to user status and environmental changes.

[0039] S5: dynamically adjusting the optical modulation parameters of the naked-eye 3D display device according to the arbitration result.

[0040] in, is the focal length of the lens, is the polarizer angle, is the refractive index of the nanostructure layer; The real-time coordinates (i.e., the precise location of the user's current gaze point on the display panel) are used to determine the spatial action benchmark for adjusting the optical modulation parameters: the focal length adjustment of the optical lens will perform local focusing optimization with the real-time coordinates as the center area; the refractive index adjustment of the nanostructure layer will perform dynamic, gradient refractive index distribution reconstruction with the pixel unit corresponding to the real-time coordinates as the benchmark.

[0041] The depth perception parameters (reflecting the user's current visual focus state and depth perception strength) are used to dynamically correct the hardware adjustment amount to make it more consistent with the user's real-time physiological state.

[0042] when When , the focal length adjustment amount is positively correlated with the amplitude of eye movement acceleration, and the focal length adjustment amount is: ,in is the focal length adjustment coefficient matrix; when When , the rotation angle is linearly related to the ambient light intensity, and the normalized value of the light intensity is ,in The maximum light intensity that the device can withstand; Polarizer rotation angle adjustment: ,in is the angle adjustment coefficient matrix; when When , the refractive index adjustment is related to the absolute value of the pupil diameter change rate. The refractive index adjustment of the nanostructure layer is: ,in is the refractive index adjustment coefficient matrix; is the depth perception bias coefficient; the refractive index change value of the nanostructure layer is limited to plus or minus 0.2.

[0043] When multiple conditions are met simultaneously ( When multiple elements in are 1), the serial execution order corresponds to the vector index order ( ), the final parameters are obtained by accumulating the adjustment amount: Original optical parameter vector According to the focal length adjustment amount, the polarizer rotation angle adjustment amount and the nanostructure layer refractive index adjustment amount, the preset adjustment amount of each cloth is obtained. Adjusted optical parameters ,in The matrix is ​​adjusted serially.

[0044] Among them, the depth perception bias coefficient is a key compensation parameter used to optimize the refractive index adjustment of the nanostructure layer in naked-eye 3D display. Its core role is to solve the problems of individual visual differences and system-inherent deviations.

[0045] It needs to be further explained that ,in This is the attenuation factor. When the acceleration is too large, it can prevent the focus from adjusting too much and causing severe shaking of the image. Is the basic proportional coefficient, which is obtained by measuring the acceleration of the lens The optimal adjustment amount is obtained by reverse deduction; ,in It is used to maintain a certain basic angle in low light and enhance the adjustment range in high light to ensure good anti-reflection effect; is the efficiency coefficient of the polarizer, and the polarization degree measured by the polarizer Sure; in As the pupil diameter change rate increases, the refractive index adjustment gradually approaches saturation, avoiding exceeding the physical limit of the material. The basic adjustment coefficient is determined by the refractive index adjustment range of the nanomaterial.

[0046] After the adjustment is completed, the new eye movement data and ambient light data are collected in real time through the S1 sensor, and the error vector of the adjusted parameters is calculated: ,in is the ideal parameter vector.

[0047] Determine whether the error is within the allowable error range. If the error is less than the allowable error, then the adjustment is valid and the adjustment process ends here; if the error is greater than or equal to the allowable error, it is necessary to re-adjust the operation, that is, recalculate the adjustment amount of each parameter and make an integrated adjustment, and then perform feedback verification again until the error is less than the allowable error, ensuring that the adjusted optical parameters can enable the naked-eye 3D display device to achieve a good display effect.

[0048] Among them, in the initial state, Preset gold standard parameters based on the hardware performance indicators of the naked-eye 3D display device and the human visual characteristics, including: Optical lens focal length: Set a benchmark value based on the comfortable viewing distance range of the human eye (e.g. 30-100cm), and calculate the initial ideal focal length based on the display panel size. ; Polarizer rotation angle: preset reference angle based on optimal contrast requirements under standard ambient light (such as 500 lux indoors) (e.g. 45°); Nanostructure layer refractive index: Based on the standard depth of field range of the displayed content (such as 0-3m virtual depth), set the initial refractive index distribution benchmark value (e.g. 1.5±0.1).

[0049] Get the ideal parameter vector: .

[0050] It should be further explained that the allowable error is: the focal length error of the optical lens is less than or equal to 0.5mm. Based on the human visual system's perception threshold of spatial resolution, the 0.5mm error threshold not only meets the indifference range of the human eye's subjective visual perception, but also matches the mechanical adjustment accuracy of mainstream optical lenses, thus balancing display quality and hardware cost. The polarizer rotation angle error is less than or equal to 2°. The polarization efficiency of the polarizer decays in a cosine law with the angle deviation. When the angle error is less than or equal to 2°, the polarization efficiency attenuation is less than or equal to 3%, and the human eye cannot perceive the change in image brightness and contrast. At the same time, this threshold matches the step accuracy of the polarizer drive motor to ensure that the hardware can stably achieve it. The refractive index error of the nanostructure layer is less than or equal to 0.02. The refractive index change of the nanostructure layer directly affects the refraction angle and phase difference of light. When the refractive index deviation is less than or equal to 0.02, the depth of field offset of the 3D image is less than or equal to 0.5%, which is far below the resolution threshold of the human eye for depth perception, and this range is within the physical adjustment capability of nanomaterials.

[0051] Secondly: The drawings of the embodiments disclosed in the present invention only involve structures related to the embodiments disclosed in the present invention. Other structures may refer to conventional designs. The same embodiment and different embodiments of the present invention may be combined with each other without conflict. Finally: The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A naked-eye 3D display optimization method based on real-time eye tracking, characterized in that: include: S1: Uses a view tracking sensor to detect the user's face and capture eye movement data in real time; The ambient light sensor is used to collect the light intensity and spectral distribution of the display scene; the eye movement data includes the pupil center position, pupil diameter and corneal reflection point coordinates; S2: Calculate depth perception parameters based on pupil diameter change and change frequency in consecutive image frames, and generate a dynamic compensation coefficient matrix based on eye movement acceleration; S3: According to the pupil center position and the corneal reflection point coordinates, the initial gaze point coordinates are obtained through a pre-trained long short-term memory neural network model; The initial gaze point coordinates are combined with the dynamic compensation coefficient matrix to correct the initial coordinates to obtain the real-time coordinates on the display panel; S4: Input the real-time coordinates, depth perception parameters, and light intensity and spectral distribution collected by the ambient light sensor into the multi-parameter arbitration controller, and generate a hardware execution sequence according to the following priority: A1: If the eye movement acceleration is greater than the preset threshold, the optical lens focus adjustment is performed first; A2: If the ambient light intensity is greater than the preset threshold, the polarizer rotation angle adjustment is performed first; A3: If the pupil diameter change rate is greater than a preset threshold, the refractive index adjustment of the nanostructure layer is performed first; A4: When multiple conditions are met at the same time, serial control is executed in sequence; S5: Dynamically adjust the optical modulation parameters of the naked-eye 3D display device according to the arbitration result.

2. The method for optimizing naked-eye 3D display based on real-time eye tracking according to claim 1, characterized in that: The eye movement data includes: The view tracking sensor uses a threshold segmentation algorithm to separate the pupil area, and performs ellipse fitting on the area to determine the pupil center position and pupil diameter; The reflection point image is collected synchronously through binocular vision or multiple cameras, and the three-dimensional coordinates of the reflection point are calculated based on the principle of triangulation. After filtering to eliminate noise fluctuations, the coordinates of the corneal reflection point are obtained.

3. The method for optimizing naked-eye 3D display based on real-time eye tracking according to claim 1, characterized in that: The depth perception parameters and the dynamic compensation coefficient matrix are specifically analyzed as follows: Performing time-attenuated weighting on pupil diameter changes over multiple consecutive frames and summing the results to obtain a weighted pupil diameter change; calculating a frequency of pupil diameter changes between the multiple consecutive frames; and combining the weighted pupil diameter change with the change frequency according to a preset weight ratio to obtain the depth perception parameter; The dynamic compensation coefficient matrix is ​​generated by superimposing a unit matrix with a correction term, and the correction term is calculated by coupling the depth perception parameter with each component of the eye movement acceleration according to a preset axial sensitivity coefficient.

4. The method for optimizing naked-eye 3D display based on real-time eye tracking according to claim 1, wherein: The real-time coordinates are specifically analyzed as follows: The normalized feature vector obtained by splicing the pupil center position coordinates and corneal reflection point coordinates captured in real time is input into the long short-term memory neural network model; The real-time coordinates are obtained by applying the dynamic compensation coefficient matrix to the initial gaze point coordinates output by the long short-term memory neural network model.

5. The method for optimizing naked-eye 3D display based on real-time eye tracking according to claim 1, wherein: The priorities include: When the eye movement acceleration is greater than a preset threshold, the optical lens focal length is adjusted first. The preset threshold is dynamically set based on the matching relationship between the human eye's persistence of vision and the display system's response time. When the ambient light intensity is greater than a preset threshold, the polarizer rotation angle is adjusted first. The preset threshold is set according to the minimum readable brightness requirement of the naked-eye 3D display device in an outdoor environment. When the pupil diameter change rate is greater than a preset threshold, the refractive index of the nanostructure layer is adjusted preferentially. The preset threshold is set based on the physiological adjustment characteristics of the human eye and the system noise tolerance.

6. The method for optimizing naked-eye 3D display based on real-time eye tracking according to claim 1, wherein: The dynamically adjusting optical modulation parameters includes: When performing optical lens focus adjustment, the focus adjustment amount is positively correlated with the amplitude of eye movement acceleration, and a nonlinear function is used to limit its maximum adjustment amplitude; When performing polarizer rotation angle adjustment, the rotation angle adjustment amount is linearly proportional to the ambient light intensity and includes a base bias angle; When the refractive index of the nanostructure layer is adjusted, the refractive index adjustment amount is positively correlated with the absolute value of the pupil diameter change rate, and a nonlinear saturation function is used to limit the adjustment range to between plus and minus 0.2; When multiple adjustments need to be performed, each adjustment is performed in sequence according to the order specified in S4, and the final optical modulation parameter value is the cumulative result of each adjustment amount.

7. The method for optimizing naked-eye 3D display based on real-time eye tracking according to claim 1, wherein: Light intensity and spectral distribution, with specific analysis including: The ambient light sensor converts the light intensity into an electrical signal through a photodiode, and obtains the light intensity value through analog-to-digital conversion; The ambient light sensor decomposes the collected light into multiple spectral channels according to wavelength through a built-in micro-spectrometer, calculates the light intensity ratio of each channel, forms a spectral distribution curve, and obtains spectral distribution data.

8. The method for optimizing naked-eye 3D display based on real-time eye tracking according to claim 4, wherein: The pre-trained long short-term memory neural network model includes the following specific analysis: Collect continuous time series of pupil center position coordinates, corneal reflection point coordinates and corresponding actual gaze point coordinate data, and construct training samples after filtering, denoising and splicing; A multi-layer long short-term memory network structure was used, with mean square error as the loss function and Adam optimizer for training until the model's gaze point coordinate prediction error on the validation set was less than 0.5 mm.

Images