Naked-eye 3D display system and control method based on metasurface cylindrical lens

Abstract

The application discloses a kind of based on super surface columnar lens naked eye 3D display system and control method, system includes super surface columnar lens layer, display panel layer and the super surface aberration compensation layer between the two;Compensation layer includes the nanometer column array formed by titanium dioxide, nanometer column width, center spacing, depth and depth ratio are in preset range, and same towards angle.The system also includes AI algorithm layer, for timing eye movement prediction, dynamic light field modeling and real-time parallax rendering;And the driving chip layer of integrated image signal processing, neural network processing, 3D acceleration and backlight control.The application effectively inhibits residual aberration and crosstalk under wide viewing angle by super surface aberration compensation layer, while combining AI perception and integrated drive, realizes the naked eye 3D display of structure thin, wide viewing angle and rapid response.

Description

Technical Field

[0001] This invention relates to naked-eye 3D display, and more particularly to a naked-eye 3D display system and control method based on metasurface cylindrical lenses. Background Technology

[0002] Current naked-eye 3D display technologies mainly include solutions such as parallax barriers, lenticular lenses, and directional light fields.

[0003] The parallax barrier solution controls the opening and closing of light through a liquid crystal layer, but the brightness loss can reach more than 65%, which seriously affects the viewing experience.

[0004] The lenticular lens solution guides light from different subpixels to the eyes through a microlens array, offering advantages such as minimal brightness loss and a wide viewing angle. However, traditional lenticular lens solutions have inherent drawbacks: 2D / 3D switching relies on the physical displacement of the optical film, resulting in a complex structure and low reliability; the field of view is limited, and even small movements by the viewer can degrade the 3D effect; and the product is relatively thick, contradicting the trend towards thinner and lighter consumer electronics.

[0005] Directional light field technology performs well in terms of resolution protection and low crosstalk, but its core components (directional backlight module, light control device and driving system) are highly customized, resulting in high cost and a narrow viewing angle.

[0006] Furthermore, Samsung, in collaboration with Pohang University of Science and Technology, published a report in *Nature* in 2026 on a switchable 2D / 3D display system based on metasurface cylindrical lenses. This system utilizes the polarization properties of light to dynamically switch between concave lenses (2D mode) and convex lenses (3D mode), reducing the optical panel thickness to approximately 1.2 mm and achieving an effective area of ​​25 square centimeters. However, this solution still faces challenges in industrialization: metasurface lenses suffer from decreased spatial resolution and display crosstalk issues at wide viewing angles; and the system lacks real-time eye tracking and dynamic light field optimization capabilities.

[0007] Therefore, there is an urgent need for a complete naked-eye 3D display solution that combines excellent optical performance, high cost-effectiveness, wide viewing angle, and ease of mass production. Summary of the Invention

[0008] Purpose of the invention: The purpose of this invention is to provide a naked-eye 3D display system and control method based on metasurface cylindrical lenses that effectively compensates for aberrations over a wide viewing angle, has a thin and light structure, and can adaptively adjust the light field.

[0009] Technical solution: The naked-eye 3D display system based on metasurface cylindrical lenses of the present invention includes a metasurface cylindrical lens layer and a display panel layer, and a metasurface aberration compensation layer is provided between the metasurface cylindrical lens layer and the display panel layer; the metasurface aberration compensation layer includes a nanopillar array formed of titanium dioxide, wherein the width, center spacing, depth and aspect ratio of each nanopillar in the nanopillar array are set within a preset range, and all nanopillars have the same orientation angle.

[0010] By adding a metasurface aberration compensation layer composed of a titanium dioxide nanopillar array between the metasurface cylindrical lens layer and the display panel layer, and limiting the key geometric parameters (width, center spacing, depth, and aspect ratio) of the nanopillars within a preset optimized range, while ensuring that the orientation angle of all nanopillars remains consistent, effective wavefront correction and residual aberration compensation can be achieved for the divergent light field generated by the metasurface cylindrical lens layer under wide viewing angle conditions. Because the propagation speed of light is slowed down by the refractive index of a transparent medium, resulting in phase delay upon exiting, the size and orientation angle of the nanopillars determine the amount of phase delay. By precisely designing the size distribution of each nanopillar in the array, the light wavefront can be repaired at the atomic scale, restoring it from a distorted shape to an ideal shape, thus achieving wavefront correction and aberration compensation. This structure effectively solves the problems of spatial resolution degradation and display crosstalk under wide viewing angles, significantly improving image quality uniformity at large angles while maintaining the overall thinness of the optical structure.

[0011] Preferably, the nanopillars in the nanopillar array have a width of 80~120nm, a depth of 1000~1500nm, and an aspect ratio greater than or equal to 10; the center-to-center distance between two adjacent nanopillars is 200~500nm.

[0012] Preferably, the aspect ratio of the nanopillar is greater than or equal to 20.

[0013] An aspect ratio of ≥20 means that the height of the nanopillar can be more than 20 times its width. The nanopillar needs to be tall enough to accumulate sufficient phase delay and achieve full control of light.

[0014] Preferably, all nanopillars have an orientation angle of 0°.

[0015] Setting the orientation angle of all nanopillars in the compensation layer to 0° is to make it easier to control the phase of light waves by simply changing the size of the nanopillars. In addition, the 0° orientation angle design is not sensitive to the angle deviations that may occur in processes such as photolithography and etching. As long as the size (width and depth) of the nanopillars meets the design, even if there is a slight shift in the orientation angle, the impact on the overall performance is far less than that of a design scheme that relies on precise angles.

[0016] Preferably, the angle between the sidewall of the metasurface aberration compensation layer and the bottom plane is 88°~90°.

[0017] By limiting the angle between the sidewalls of the metasurface aberration compensation layer and the bottom plane to within the range of 88° to 90°, the tolerance window for the manufacturing process was clearly defined. This limitation ensures that the nanopillars have near-vertical sidewalls, meeting the aspect ratio requirements to provide efficient phase control capabilities, while allowing for a process deviation of approximately 2° in actual production. This significantly reduces the stringent requirements for absolute perpendicularity in the etching process, and substantially improves the mass production yield and manufacturing cost-effectiveness of the product.

[0018] Preferably, the metasurface columnar lens includes a nanopillar array, an active linear polarizer, and a polarization control film stacked sequentially along the light incident direction.

[0019] Preferably, the nanopillar array of the metasurface cylindrical lens is based on the Pancharatnam-Berry phase principle, embedding both convex and concave lens phase distributions within the same array, and using an active linear polarizer to switch the polarization direction of the incident light to the polarization state corresponding to different lenses to trigger the lens phase distribution corresponding to the nanopillar array.

[0020] Preferably, an AI algorithm layer is also included, which includes:

[0021] The temporal eye-tracking module is used to predict the spatial position of both eyes in future time periods based on the acquired images of the viewer's eyes;

[0022] The AI ​​light field generation module is used to dynamically construct a joint geometric optical path model between the viewer, the metasurface lenticular lens layer and the display panel layer based on the predicted binocular spatial position, and calculate the effective projection angle and optimal viewing center coordinates of the metasurface lenticular lens.

[0023] The parallax rendering module is used to generate left and right eye parallax images that match the current viewing position in real time based on the output of the light field model, and output them to the display panel layer.

[0024] By further integrating an AI algorithm layer that includes temporal eye tracking, AI light field generation, and parallax rendering, the system gains the ability to proactively perceive and dynamically adapt to changes in the viewer's position. Temporal eye tracking predicts the viewer's future gaze position, providing the system with an advanced decision-making benchmark; the AI ​​light field generation module reconstructs the light path model in real time based on this, ensuring that the optimal viewing area for the 3D display always dynamically follows the viewer; and the parallax rendering module accurately outputs a parallax image adapted to the current position. These three components work together to solve the problems of 3D effect degradation and dizziness caused by viewer movement, achieving adaptive, low-latency intelligent glasses-free 3D display.

[0025] Preferably, it also includes a driver chip layer, which includes an image signal processor, a neural network processing unit, a 3D acceleration unit, and a backlight control unit integrated in the same chip;

[0026] Among them, the neural network processing unit is used to run the sequential eye-tracking model; the 3D acceleration unit is used to provide hardware acceleration for the parallax rendering module; and the backlight control unit is used to control the luminous intensity of the backlight module in zones and maintain frame-level synchronization with the polarization switching of the active linear polarizer and the image refresh of the display panel.

[0027] By heterogeneously integrating image signal processing, neural network computation, 3D rendering acceleration, and backlight control unit into a single driver chip, a tightly coupled hardware platform covering the entire chain from perception and decision-making to physical display has been constructed. This architecture significantly reduces the data interaction latency between various functional modules. In particular, the frame-level synchronization design of the backlight control unit with polarization switching and image refresh allows the end-to-end response latency of the entire system to be compressed to a low level, providing a solid hardware foundation for a smooth, dizziness-free naked-eye 3D experience.

[0028] The naked-eye 3D display control method based on metasurface cylindrical lenses of the present invention includes:

[0029] Collect images of the viewer's eyes and predict the spatial coordinates of both eyes in future time periods based on a time-series prediction model;

[0030] Based on the predicted binocular spatial coordinates, a joint geometric optical path model is constructed between the viewer, the metasurface lenticular lens layer, and the display panel layer, and the effective projection angle and optimal viewing center coordinates of the metasurface lenticular lens are calculated.

[0031] Based on the calculation results of the optical path model, the following operations are performed synchronously: control the metasurface lenticular lens layer to switch to a 2D mode or 3D mode that matches the display content; generate left and right eye parallax images corresponding to the predicted viewing position and output them to the display panel layer; perform zonal adjustment of the backlight and keep it synchronized with polarization switching and image refresh.

[0032] By introducing temporal eye-tracking prediction to drive the display system, and using the predicted coordinates as a unified benchmark for subsequent light field calculations and display control, the fundamental problem of system response lagging behind human eye movement is solved. Based on this, a complete prediction-decision-execution control process is formed by constructing a joint geometric optical path model and simultaneously executing three core operations: polarization mode switching, parallax image refresh, and zoned backlight adjustment. This method ensures that optical layer modulation, image layer rendering, and backlight layer illumination accurately match the viewer's dynamic position both temporally and spatially, thus providing a consistently clear, tear-free naked-eye 3D visual experience during continuous motion.

[0033] Beneficial effects: By introducing a titanium dioxide nanopillar array with specific preset geometric parameters as a metasurface aberration compensation layer between the metasurface columnar lens layer and the display panel layer, and ensuring that all nanopillars have the same orientation angle, residual aberrations under wide viewing angles are effectively compensated, and spatial resolution degradation and display crosstalk are significantly suppressed. Based on this, combined with a time-predictive AI algorithm layer and a highly integrated driver chip layer, the system can adaptively adjust the light field and achieve low-latency response according to the dynamic changes in the viewer's position. Attached Figure Description

[0034] Figure 1 This is a block diagram of the overall structure of the naked-eye 3D display system of the present invention;

[0035] Figure 2 This is a diagram of the layered architecture of the optical layer;

[0036] Figure 3 A sectional view of the optical layers;

[0037] Figure 4 This diagram illustrates the working principle of polarization switching between 2D and 3D modes using a metasurface cylindrical lens.

[0038] Figure 5 This is a data flow diagram for temporal eye tracking and light field prediction.

[0039] Figure 6 This defines the logical relationships and data interaction processes between the various modules of the AI ​​algorithm layer.

[0040] Figure 7 A schematic diagram of the internal framework of a dedicated 3D display acceleration chip for driving the chip layer. Detailed Implementation

[0041] As shown in the figure, the naked-eye 3D display system based on metasurface lenticular lenses of the present invention includes a metasurface lenticular lens layer, a metasurface aberration compensation layer, and a display panel layer. The metasurface aberration compensation layer is disposed between the metasurface lenticular lens layer and the display panel layer. The three layers are arranged in a sandwich-like vertical stacking structure from top to bottom (with the light incident direction being downward), together forming an optical layer.

[0042] The metasurface columnar lens layer is located at the top of the optical layer. This layer includes a polarization control film, an active linear polarizer, and a nanopillar array stacked sequentially along the light incident direction (to distinguish it from the metasurface aberration compensation layer, the nanopillar array of the metasurface columnar lens layer will be referred to as the first nanopillar array). The first nanopillar array is based on the Pancharatnam–Berry (PB) phase principle, embedding two phase distributions within the same array: a convex lens (corresponding to the 3D mode) and a concave lens (corresponding to the 2D mode). During operation, the active linear polarizer receives control signals from the driving chip. When it receives the first control signal (such as the positive driving voltage +V), the active linear polarizer switches the polarization direction of the incident light to the first polarization state, aligning the output linear polarization direction with 0°, triggering the concave lens phase distribution of the first nanopillar array. This cancels the refractive force of the original convex lens (the display panel itself (or its underlying optical components in front of it) is designed to have specific basic converging characteristics, which can be equivalent to the original convex lens), making the overall refraction approach zero, allowing light to pass through in approximately a straight line, thus achieving 2D display. When a second control signal (such as a negative driving voltage -V) is received, the polarization direction is switched to the second polarization state, the output linear polarization direction is aligned with 90°, triggering the phase distribution of the convex lens, and light refraction forms multiple viewpoints to achieve 3D display. This process involves no physical displacement, and the switching time is approximately 10 milliseconds. The first nanopillar array is composed of nanoscale superatoms (TiO2 nanopillars) arranged in a specific periodic pattern, where each nanopillar has a width D1 of 80-200 nm, a center-to-center spacing P1 of 250-450 nm between adjacent nanopillars, a depth H1 of 600-1500 nm, an aspect ratio ranging from 3:1 to 10:1, and a preset orientation angle of 0° to 180° to achieve PB phase encoding.

[0043] The metasurface aberration compensation layer comprises an array of nanopillars formed from titanium dioxide (TiO2), a high-refractive-index material, which may be referred to as the second nanopillar array. The width D2, center-to-center spacing P2, depth H2, and aspect ratio of each nanopillar in the array are all set within a preset optimization range, and all nanopillars have the same orientation angle.

[0044] In one specific embodiment, the width D2 of the nanopillars is 80~120nm, the depth H2 is 1000~1500nm, and the aspect ratio is not less than 10:1, preferably not less than 20:1, to provide stronger aberration compensation capability; the center-to-center distance P2 between two adjacent nanopillars is 200~500nm. The orientation angle of all nanopillars can be a full geometric phase coverage from 0 to 2π, and the specific orientation angle is determined according to the target phase distribution required by the metasurface aberration compensation layer; in this embodiment, the orientation angle of the nanopillars is uniformly set to 0°. Under the above setting requirements, the angle S between the sidewall and the bottom plane of the metasurface aberration compensation layer is allowed to be between 88° and 90°, that is, a process deviation of about 2° is allowed, which is more convenient for actual industrial production. This compensation layer, with the help of the high refractive index and low loss characteristics of titanium dioxide material, can perform wavefront correction and compensation for residual aberrations and chromatic aberrations of the light field emitted by the metasurface columnar lens layer at a wide viewing angle (such as 100 degrees), effectively suppressing spatial resolution degradation and display crosstalk. The upper and lower surfaces of the metasurface aberration compensation layer are tightly bonded to the metasurface columnar lens layer and the display panel layer, respectively, using high-transmittance OCA optical adhesive. The geometric centers of each layer are kept on the same optical axis. The OCA optical adhesive has a refractive index of approximately 1.52 and a thickness of approximately 25 μm to achieve refractive index matching of the optical path. Its fabrication can employ a mixed-gas stepwise inductively coupled plasma (ICP) etching method, combining the high selectivity of CHF3 with the high anisotropy of BCl3 to achieve the required high aspect ratio and sidewall angle.

[0045] The display panel layer is located at the bottom of the three-layer optical structure and serves as the source of the system's image content. The display panel layer can be a TFT-LCD panel or an OLED panel. It receives left and right eye parallax image data from the driving chip, emits light pixel by pixel, and the light sequentially passes through the metasurface aberration compensation layer and the metasurface lenticular lens layer before entering the human eye.

[0046] The metasurface columnar lens layer and the TiO2 metasurface aberration compensation layer involved in this invention can both be mass-produced using large-area roll-to-plate nanoimprint technology. The specific steps are as follows:

[0047] Step 1: Master fabrication – Using electron beam lithography, a reverse pattern of metasurface columnar lenses is constructed on a quartz or silicon substrate;

[0048] Step 2: Material Coating – Spin-coating a high-refractive-index UV-curable resin onto a transparent glass substrate or OLED encapsulation layer;

[0049] Step 3: Imprinting and Transfer – The master pattern is imprinted onto the resin layer using a Roll-to-Plate device. The imprinting cycle is approximately 2 minutes.

[0050] Step 4: UV curing – UV curing is performed simultaneously during the imprinting process to maintain the fidelity of the pattern;

[0051] Step 5: Demolding and Protection – After precise demolding, a protective coating is applied to the surface of the lens array;

[0052] Step 6: TiO2 layer deposition – On the substrate where the metasurface columnar lens has been fabricated, a TiO2 nanopillar compensation layer is formed by atomic layer deposition or nanoimprinting.

[0053] Step 7: Optical performance testing – Sampling tests are conducted on key optical indicators such as field of view, resolution retention, and crosstalk rate for each batch of products, with a yield target set at ≥92%.

[0054] The system of this invention also includes an AI algorithm layer. The AI ​​algorithm layer mainly includes a temporal eye-tracking module, an AI light field generation module, and a parallax rendering module.

[0055] The temporal eye-tracking module acquires images of the viewer's face and eyes using at least one camera. It then employs a trained bidirectional long short-term memory (Bi-LSTM) neural network to predict the spatial coordinates of both eyes for the next 80 milliseconds based on the eye movement trajectories over the past 60-80 milliseconds. The output of this module serves as the reference input for the optical layer metasurface cylindrical lens polarization switching and AI light field generation modules, as well as the 2D real-time to 3D AI engine module.

[0056] The AI ​​light field generation module performs SBP dynamic optimization based on predicted binocular spatial coordinates, calculates the light field distribution in real time, and dynamically constructs a joint geometric light path model of "viewer (human eye) - metasurface lenticular lens layer - display panel layer". It dynamically calculates parameters such as the effective projection angle of the metasurface lenticular lens and the optimal center coordinates of the viewing area. Specifically, it can refer to the layered light field architecture concept of EyeReal of Fudan University, and adopts a dynamic spatial bandwidth product optimization strategy to focus the limited display information budget on the area where the viewer's line of sight falls in real time, rather than broadcasting it uniformly to the entire field of view. The calculated θ parameters are then passed to the parallax rendering module.

[0057] The parallax rendering module, based on the output of the AI ​​light field generation module, calls the sub-pixel rearrangement algorithm in real time to dynamically generate left and right eye parallax images that match the current viewing position, and outputs them to the display panel layer, while maintaining frame-level synchronization with the polarization state of the metasurface lenticular lens layer.

[0058] In addition, the AI ​​algorithm layer may also include a 2D real-time to 3D AI engine module, which uses a deep neural network to perform real-time depth estimation on the input standard 2D video stream and generate left and right eye parallax images. Simultaneously, it triggers the metasurface lenticular lens layer to switch from 2D mode to 3D mode, realizing the real-time 3D conversion of existing 2D content. The deep neural network here can be any of the following existing technologies, or other existing neural networks with the same function: 1. MiDaS (Multiple Depth Estimation Accuracy with Single Network): A classic monocular depth estimation model that can estimate relative depth maps from a single 2D image with high accuracy. Subsequent upgraded versions, such as DPT (Dense Prediction Transformer), can further improve accuracy; 2. Depth Pro: A zero-shot monocular depth estimation model developed by Apple, capable of generating high-precision 3D depth maps from a single 2D image at high speed; 3. Publication number CN121309793A (or CN121262350A).

[0059] To achieve efficient collaboration among the modules, this system also includes a driver chip layer. The driver chip layer contains at least one embedded 3D display acceleration chip, which integrates an image signal processor (ISP), a neural network processing unit (NPU), a 3D acceleration unit, and a backlight control unit into a single chip.

[0060] The ISP is used to preprocess the images captured by the eye-tracking camera. This module connects the external dual-camera input signal lines and the internal output data stream to the NPU module.

[0061] The NPU has a computing power of at least 1 TOPS, which is used to run Bi-LSTM temporal eye-tracking models and 2D-to-3D depth estimation models.

[0062] The 3D acceleration unit is used to hardware accelerate disparity map calculations and subpixel rearrangement operations.

[0063] The backlight control unit is used to control the luminous intensity of the backlight module in zones and maintain strict frame-level synchronization with the polarization switching of the active linear polarizer and the image refresh of the display panel.

[0064] The control method flow of the present invention is as follows:

[0065] First, images of the viewer's eyes are captured in real time by a camera. After preprocessing by an ISP, the images are input into a Bi-LSTM time-series prediction model to predict the spatial coordinates of the eyes within the next 80 milliseconds.

[0066] Next, based on the predicted eye positions, the AI ​​light field generation module is activated to dynamically construct a joint geometric optical path model between the viewer, the metasurface lenticular lens layer, and the display panel layer, and to calculate the effective projection angle and optimal viewing center coordinates of the metasurface lenticular lens.

[0067] Finally, based on the calculation results of the AI ​​light field generation module, the following parallel operations are executed synchronously: a control voltage is applied to the metasurface lenticular lens layer, and the polarization state of the incident light is switched through an active linear polarizer, so that the metasurface lenticular lens switches to a 2D mode or 3D mode that matches the display content; the parallax rendering module is called to generate left and right eye parallax images corresponding to the predicted viewing position and output them to the display panel layer; the backlight control unit performs zoned adjustment of the backlight and keeps it synchronized with polarization switching and image refresh.

[0068] If the input source is 2D content, the method also includes a step of triggering a 2D real-time to 3D AI engine to convert the 2D image into a 3D parallax image in real time and link it with the 3D mode.

[0069] The entire process is performed in a frame-level synchronized manner, providing viewers with a continuous, dizziness-free, wide-angle, and high-brightness naked-eye 3D viewing experience.

[0070] The following examples illustrate this in detail.

[0071] Example 1

[0072] It uses a 27-inch 4K TFT-LCD panel as its display core.

[0073] The metasurface columnar lens layer employs a roll-to-plate nanoimprinting process, where UV-cured resin is spin-coated onto a transparent glass substrate, and the metasurface columnar lens structure is transferred using a nanoimprinting master. The lens array pitch is designed to match the sub-pixel layout of the display panel, the material's refractive index n is between 1.50 and 1.58, and the overall height is controlled at the micrometer level. Electron beam lithography is used to precisely control the geometry and spacing of the nanopillars to construct the metasurface functional areas, resulting in an overall optical panel thickness of approximately 1.2 millimeters.

[0074] A metasurface aberration compensation layer is set below the metasurface columnar lens layer. The aberration compensation layer composed of TiO2 nanopillars is constructed by atomic layer deposition or spin coating process to accurately correct residual aberration and chromatic aberration under a 100-degree wide viewing angle.

[0075] The eye-tracking system uses two 5-megapixel camera modules embedded on either side of the top bezel of the display (horizontally spaced approximately 15cm apart). The cameras capture images of the user's face at 120 frames per second (fps). The system first detects the facial region, extracts the region of interest (ROI) for the eyes, and uses a custom edge AI accelerator to extract pupil-iris feature points. The feature vector is then input into a bidirectional long short-term memory (Bi-LSTM) temporal prediction model.

[0076] Bi-LSTM model training: Before deployment, the model was pre-trained using a large-scale dataset containing over 10,000 real eye movement trajectories. During actual operation, the model predicts the spatial positions of both eyes in the next 80 milliseconds based on the eye movement sequence of the past 70 milliseconds, with the spatial prediction error controlled within ±3mm.

[0077] Real-time operation flow: When the user first looks at the screen, the system automatically completes face detection and initial calibration. During subsequent operation, the Bidirectional Long Short-Term Memory (Bi-LSTM) module continuously predicts the gaze position; the AI ​​light field generation module dynamically calculates the optimal optical path projection angle of the metasurface lenticular lens; the parallax rendering module synthesizes left and right eye parallax images in real time; and the driver chip, while outputting the rendered image to the display panel, simultaneously controls the polarization switching and backlight zoning of the metasurface lenticular lens via the electrode plate. When the user wears glasses or the ambient light changes, the system automatically adjusts the camera exposure parameters and image gain to maintain tracking robustness.

[0078] Example 2

[0079] Based on a 55-inch 8K TFT-LCD panel, this system is designed for multi-user scenarios. Four wide-angle cameras are positioned at the four corners to detect and locate the faces of multiple users. The AI ​​light field generation module employs a multi-user weighted optimization algorithm (which can utilize the weighted optimization algorithm proposed by Gao Chen (Mindu Innovation Lab)). It selects the average viewing position as the main reference viewpoint and simultaneously performs multi-view layered optimization of the projected light field of the metasurface cylindrical lens based on the secondary viewing position information.

[0080] The multi-view layered optimization includes the following steps:

[0081] (1) Based on the spatial positions of each viewer determined by the multi-person weighted optimization algorithm, the display area is divided into N view areas (N≥2).

[0082] (2) Calculate the light field parameters (including projection angle θ_i and center coordinates C_i) of the optimal viewing position for each viewing area independently.

[0083] (3) Allocate display resources according to the priority of each view area (e.g., weighted by the number of viewers or viewing distance);

[0084] (4) The light field parameters of each viewing area are fused to generate the driving signal matrix of the metasurface cylindrical lens.

[0085] For large-size panels, the metasurface aberration compensation layer uses a partition calibration strategy to compensate for the uniformity error of the metasurface cylindrical lens during the large-area manufacturing process, so as to ensure the consistency of color and the uniformity of 3D depth perception throughout the display area.

[0086] The partition calibration strategy includes the following steps:

[0087] (1) Divide the large-area metasurface aberration compensation layer into M×N calibration partitions in space (M≥2, N≥2);

[0088] (2) For each calibration partition, independently measure the deviation Δφ_ij(x,y) between its actual phase response and the design target value;

[0089] (3) Calculate compensation parameters independently for each partition based on the measurement deviation (such as adjusting the equivalent refractive index or local thickness of the nanopillars in that partition).

[0090] (4) Write the compensation parameters of each zone into the manufacturing equipment and implement zone-differentiated processing in subsequent batch production;

[0091] (5) Perform secondary measurement verification on the compensated partition and iterate until the phase uniformity of the whole region reaches the preset threshold.

Claims

1. A glasses-free 3D display system based on metasurface lenticular lenses, comprising a metasurface lenticular lens layer and a display panel layer, characterized in that: Between the metasurface columnar lens layer and the display panel layer, there is also a metasurface aberration compensation layer; the metasurface aberration compensation layer includes a nanopillar array formed of titanium dioxide, wherein the width, center spacing, depth and aspect ratio of each nanopillar in the nanopillar array are set within a preset range, and all nanopillars have the same orientation angle.

2. The system according to claim 1, characterized in that: The nanopillars in the array have a width of 80-120 nm, a depth of 1000-1500 nm, and an aspect ratio greater than or equal to 10; the center-to-center distance between two adjacent nanopillars is 200-500 nm.

3. The system according to claim 2, characterized in that: The aspect ratio of the nanopillars is greater than or equal to 20.

4. The system according to claim 1, characterized in that: All nanopillars have an orientation angle of 0°.

5. The system according to claim 2, characterized in that: The angle between the sidewall of the metasurface aberration compensation layer and the bottom plane is 88°~90°.

6. The system according to claim 1, characterized in that: The metasurface cylindrical lens comprises a polarization control film, an active linear polarizer, and a nanopillar array stacked sequentially along the light incident direction.

7. The system according to claim 6, characterized in that: The nanopillar array of the metasurface cylindrical lens is based on the Pancharatnam-Berry phase principle, embedding both convex and concave lens phase distributions within the same array. An active linear polarizer switches the polarization direction of the incident light to the polarization state corresponding to different lenses to trigger the lens phase distribution corresponding to the nanopillar array.

8. The system according to claim 1, characterized in that, It also includes an AI algorithm layer, which includes: The temporal eye-tracking module is used to predict the spatial position of both eyes in future time periods based on the acquired images of the viewer's eyes; The AI ​​light field generation module is used to dynamically construct a joint geometric optical path model between the viewer, the metasurface lenticular lens layer and the display panel layer based on the predicted binocular spatial position, and calculate the effective projection angle and optimal viewing center coordinates of the metasurface lenticular lens. The parallax rendering module is used to generate left and right eye parallax images that match the current viewing position in real time based on the output of the light field model, and output them to the display panel layer.

9. The system according to claim 8, characterized in that, It also includes a driver chip layer, which includes an image signal processor, a neural network processing unit, a 3D acceleration unit, and a backlight control unit integrated in the same chip; Among them, the neural network processing unit is used to run the sequential eye-tracking model; the 3D acceleration unit is used to provide hardware acceleration for the parallax rendering module; and the backlight control unit is used to control the luminous intensity of the backlight module in zones and maintain frame-level synchronization with the polarization switching of the active linear polarizer and the image refresh of the display panel.

10. A method for controlling naked-eye 3D displays based on metasurface cylindrical lenses, comprising: Collect images of the viewer's eyes and predict the spatial coordinates of both eyes in future time periods based on a time-series prediction model; Based on the predicted binocular spatial coordinates, a joint geometric optical path model is constructed between the viewer, the metasurface lenticular lens layer, and the display panel layer, and the effective projection angle and optimal viewing center coordinates of the metasurface lenticular lens are calculated. Based on the calculation results of the optical path model, the following operations are performed synchronously: control the metasurface lenticular lens layer to switch to a 2D mode or 3D mode that matches the display content; generate left and right eye parallax images corresponding to the predicted viewing position and output them to the display panel layer; perform zonal adjustment of the backlight and keep it synchronized with polarization switching and image refresh.

Images