Image adaptive calibration method for calibrating naked-eye 3D parameters and related device

Abstract

The application relates to the technical field of naked-eye 3D display, in particular to an image self-adaptive calibration method for calibrating naked-eye 3D parameters and related equipment, which comprises the following steps: a terminal screen projects a positioning pattern with known geometric features, a camera collects and calculates real three-dimensional space parameters of the camera relative to the screen; current 3D calibration parameters are loaded, a calibration image is generated and projected in combination with the real parameters; an actual calibration image is collected, key indicators such as crosstalk rate and viewpoint position error are quantitatively analyzed to obtain a performance error vector; parameter adjustment amount is calculated with the minimum error vector as the target and the real space parameters as the constraint; the parameters are updated and the above process is repeated until the final parameters are fixed after the error meets the threshold value. The application helps to automatically compensate for 3D calibration parameter deviation caused by individual differences of equipment, non-standard camera position and environmental changes through intelligent algorithms, and realizes personalized and high-precision naked-eye 3D visual experience.

Description

Technical Field

[0001] This application relates to the field of naked-eye 3D display technology, and in particular to an image adaptive calibration method and related equipment for calibrating naked-eye 3D parameters. Background Technology

[0002] Glasses-free 3D display technology uses optical elements such as lenticular lenses and gratings to precisely allocate multi-viewpoint images to specific locations in space, allowing viewers to experience stereoscopic visual effects without wearing 3D glasses. It is widely used in consumer electronics, automotive displays, and other fields. The core of its display effect relies on the accuracy of 3D calibration parameters (including viewpoint separation distance, image offset, crosstalk suppression coefficient, and grating phase compensation value). These parameters directly determine the spatial allocation accuracy of multi-viewpoint images, thus affecting the clarity, stereoscopic effect, and viewing comfort of the stereoscopic image.

[0003] However, existing naked-eye 3D parameter calibration technologies face two major technical challenges that severely restrict the widespread adoption of naked-eye 3D technology: (a) Internal parameter drift problem The static calibration parameters of the equipment at the factory will slowly drift due to factors such as temperature changes, material aging, and physical stress. As the usage time increases, the cumulative drift will lead to problems such as increased viewpoint crosstalk, deviation of the optimal viewing position, and distortion of depth perception, causing the 3D display effect to gradually deteriorate.

[0004] (ii) External reference bias problem For mobile terminals with integrated cameras, the camera is the core reference "sensor" in the naked-eye 3D calibration process, and its relative position to the screen directly determines the accuracy of the calibration reference information. However, in actual production and assembly, there are unavoidable tolerances in the assembly position of cameras for different models, and even for different devices of the same model. Traditional calibration methods use the "ideal reference point assumption," which assumes that the camera is located in a standard position directly in front of the center of the screen. This "one-size-fits-all" design completely ignores the actual assembly tolerances, causing the calibration algorithm to obtain incorrect reference information and calculate inaccurate 3D calibration parameters. This results in inherent errors in the 3D effect from the initial stage of calibration.

[0005] Existing technologies cannot simultaneously solve the two coupled problems of internal parameter drift and external reference bias, resulting in extremely poor consistency and stability of naked-eye 3D experience. Therefore, there is an urgent need for a naked-eye 3D parameter adaptive calibration method that can accurately measure and compensate for external reference bias, synchronously correct internal parameter drift, and break the coupling effect between the two, so as to solve the core defects of existing technologies. Summary of the Invention

[0006] To help automatically compensate for deviations in 3D calibration parameters caused by individual device differences, non-standard camera positions, and environmental changes through intelligent algorithms, and to achieve a personalized, high-precision naked-eye 3D visual experience, this application provides an image adaptive calibration method and related equipment for calibrating naked-eye 3D parameters.

[0007] Firstly, this application provides an image adaptive calibration method for calibrating naked-eye 3D parameters, which adopts the following technical solution: An image adaptive calibration method for calibrating naked-eye 3D parameters includes: S101. Control the terminal screen to project a positioning pattern with known geometric features, acquire an image of the positioning pattern through the terminal camera, and calculate the real three-dimensional spatial parameters of the camera relative to the screen based on a computer vision algorithm; S102. Load the current 3D calibration parameters, generate a position-aware calibration image according to the real three-dimensional spatial parameters, and project the calibration image according to the current 3D calibration parameters. S103. Acquire the actual projected calibration image through the camera, quantitatively analyze the key quality indicators of the calibration image, calculate the deviation between the key quality indicators and the preset ideal value, and obtain the performance error vector. S104. With minimizing the performance error vector as the objective and the real three-dimensional space parameters as constraints, calculate the adjustment amount of the 3D calibration parameters through an optimization algorithm; S105. Apply the parameter adjustment amount to the 3D display engine to update the 3D calibration parameters, repeat steps S102 to S104 until the performance error vector meets the preset threshold, and solidify the final 3D calibration parameters. The true 3D spatial parameters include the vertical distance between the camera and the screen, the horizontal offset, the pitch angle, and the roll angle; the calculation of the true 3D spatial parameters based on computer vision algorithms includes: Extract several non-coplanar feature points from the positioning pattern and record the known three-dimensional coordinates of each feature point; The acquired positioning pattern image is subjected to noise filtering to remove false feature points caused by environmental interference; Establish the correspondence between the camera intrinsic parameter matrix and the 3D coordinates of feature points and the image pixel coordinates, and solve for the extrinsic parameter matrix; The rotation matrix and translation vector are decomposed from the extrinsic parameter matrix, and the vertical distance, horizontal offset, pitch angle and roll angle are calculated. Several frames of positioning pattern images are continuously acquired and repeatedly processed to calculate the mean and variance of several sets of processing results; If the variance is less than the set stability threshold, the mean will be used as the final true three-dimensional space parameter. If the variance is greater than or equal to the preset stability threshold, re-acquire several frames of positioning pattern images and repeat the calculation process.

[0008] Optionally, the key quality indicators include crosstalk rate and viewpoint position error; the key quality indicators for quantifying the calibration image include: Extract the first pixel region corresponding to the first identifier of the left eye viewpoint region and the second pixel region corresponding to the second identifier of the right eye viewpoint region from the calibration image; Obtain the first grayscale average value corresponding to the second identifier in the first pixel region and the second grayscale average value corresponding to the second identifier in the second pixel region; Based on the first grayscale mean and the second grayscale mean, the grayscale ratio is calculated, and the grayscale ratio is used as the crosstalk rate; The first pixel coordinates corresponding to the first identifier and the second pixel coordinates corresponding to the second identifier in the calibration image are obtained by using a preset image algorithm. Retrieve the first ideal coordinates corresponding to the first identifier and the second ideal coordinates corresponding to the second identifier from the preset ideal values; The viewpoint position error is obtained based on the first pixel coordinates, the second pixel coordinates, the first ideal coordinates, and the second ideal coordinates.

[0009] Optionally, obtaining the viewpoint position error based on the first pixel coordinates, the second pixel coordinates, the first ideal coordinates, and the second ideal coordinates includes: Based on the first pixel coordinates and the first ideal coordinates, obtain the first position error value; Based on the second pixel coordinates and the second ideal coordinates, obtain the second position error value; Determine whether the specified configuration command has been detected; If detected, the viewpoint position error is obtained based on the specified configuration instruction, the first position error value, or the second position error value. If not detected, determine whether a specified runtime environment requirement exists; If it exists, determine whether the specified operating environment requirement is a rigid requirement; If it is true, then the maximum value between the first position error value and the second position error value is selected as the viewpoint position error.

[0010] Optionally, after determining whether the specified operating environment requirement is a rigid requirement if it exists, the method further includes: If not, determine whether the terminal camera supports independent adjustment of dual-eye parameters; If not supported, the average of the first position error value and the second position error value is obtained as the viewpoint position error; If supported, obtain the daily usage requirements of the terminal camera; If the daily use requirement is an extended use requirement, then the average of the first position error value and the second position error value is obtained as the viewpoint position error; If the daily usage requirement is not the extended usage requirement, then the maximum value between the first position error value and the second position error value is selected as the viewpoint position error.

[0011] Optionally, repeating steps S102 to S104 until the performance error vector meets a preset threshold, and solidifying the final 3D calibration parameters, includes: Each time the 3D calibration parameters are updated, the updated parameters are reloaded to generate a calibration image and project it. The newly projected calibration images are captured by a camera, key quality indicators are quantitatively analyzed, and a new performance error vector is calculated. The new performance error vector is compared with a preset threshold to determine whether the crosstalk rate and viewpoint position error are both less than the corresponding preset threshold. If both indicators are less than the corresponding preset threshold, the iteration process is stopped and the current 3D calibration parameters are fixed. If any indicator is greater than or equal to the corresponding preset threshold, then determine whether the current iteration number exceeds the maximum iteration number; If the maximum number of iterations has not been exceeded, return to step S102 to continue the calibration process; If the maximum number of iterations has been exceeded, then the difference between the current performance error vector and the target value of the preset threshold is detected. If the target difference is less than or equal to the preset difference threshold, the current parameters will be used as a transitional solution, and the user will be prompted to recalibrate periodically. If the difference exceeds the preset difference threshold, switch to the basic 3D calibration parameters, disable the adaptive calibration function, and prompt for troubleshooting.

[0012] Optional, also includes: Record the performance error vector and corresponding 3D calibration parameters for each iteration to form an iteration history record; Calculate the difference in error vectors between two adjacent iterations to determine the error convergence trend; If the error difference shows a decreasing trend and the current error vector meets the preset threshold, then stop the iteration and fix the parameters; If the error difference tends to stabilize but does not meet the threshold, then reduce the parameter adjustment step size and continue iterating; If the error difference shows an increasing trend, then backtrack to the calibration parameters of the previous valid iteration and adjust the optimization direction; The number of iterations is counted. If the threshold is not met even after the number of iterations reaches the preset upper limit, the working status of the camera and screen is detected. If the equipment malfunctions, a fault alarm will be generated, and the calibration process will be terminated.

[0013] Secondly, this application also discloses an image adaptive calibration system for calibrating naked-eye 3D parameters, which adopts the following technical solution: An image adaptive calibration system for calibrating naked-eye 3D parameters includes: The spatial position detection module is used to control the terminal screen to project a positioning pattern with known geometric features, acquire images of the positioning pattern through the terminal camera, and calculate the real three-dimensional spatial parameters of the camera relative to the screen based on computer vision algorithms. The calibration image processing module is used to load the current 3D calibration parameters, generate a position-aware calibration image based on the real three-dimensional spatial parameters, and project the calibration image according to the current 3D calibration parameters. The display quality analysis module is used to acquire the actual projected calibration image through the camera, quantitatively analyze the key quality indicators of the calibration image, calculate the deviation of the key quality indicators from the preset ideal value, and obtain the performance error vector. The parameter joint optimization module is used to calculate the adjustment amount of 3D calibration parameters through an optimization algorithm, with the goal of minimizing the performance error vector and the real three-dimensional space parameters as constraints. The iterative update and solidification module is used to apply the parameter adjustment amount to the 3D display engine to update the 3D calibration parameters until the performance error vector meets the preset threshold, and then solidify the final 3D calibration parameters.

[0014] Thirdly, the computer device provided in this application adopts the following technical solution: A smart terminal includes a memory and a processor, wherein the memory stores a computer program that can run on the processor, and when the processor loads the computer program, it executes the method of the first aspect.

[0015] By adopting the above technical solution, a computer program is generated based on the method of the first aspect and stored in a memory for loading and execution by a processor. Thus, a smart terminal is made based on the memory and the processor, making it convenient for users to use.

[0016] Fourthly, the computer-readable storage medium provided in this application adopts the following technical solution: A computer-readable storage medium storing a computer program that, when loaded by a processor, executes the method of the first aspect.

[0017] By adopting the above technical solution, a computer program is generated based on the method of the first aspect and stored in a computer-readable storage medium for loading and execution by a processor. The computer-readable storage medium facilitates the reading and storage of the computer program.

[0018] In summary, this application includes the following beneficial technical effects: To address the issue of external reference deviation, a positioning pattern is projected and its image is acquired. Based on computer vision algorithms, the true three-dimensional spatial parameters of the camera relative to the screen are accurately calculated. This breaks the limitations of the traditional method's "ideal reference point assumption," eliminating reference information errors caused by camera assembly tolerances at the source and providing an accurate benchmark for subsequent parameter calibration. To address the issue of internal parameter drift, a closed-loop iterative process is used, constrained by true three-dimensional spatial parameters, to continuously quantify and analyze display quality and dynamically optimize 3D calibration parameters until performance errors meet thresholds and parameters are solidified. This achieves real-time compensation for parameter drift caused by factors such as temperature changes and material aging, ensuring that 3D calibration parameters are always in optimal condition. Furthermore, intelligent algorithms automatically compensate for 3D calibration parameter deviations caused by individual device differences, non-standard camera positions, and environmental changes, achieving personalized "one device, one calibration" customization. This ensures stable and consistent 3D display effects throughout the device's lifecycle and significantly improves display accuracy, ultimately providing users with a personalized, high-precision naked-eye 3D visual experience. Attached Figure Description

[0019] Figure 1 This is a main flowchart of an image adaptive calibration method for calibrating naked-eye 3D parameters according to an embodiment of this application; Figure 2 This is a flowchart of the steps for solving real 3D spatial parameters based on computer vision algorithms; Figure 3 This is a flowchart of the steps for quantitatively analyzing the key quality indicators of the calibrated image; Figure 4 This is a flowchart of the steps to obtain the viewpoint position error; Figure 5 This is a flowchart of the data iteration and parameter solidification steps; Figure 6 This is a block diagram of an image adaptive calibration system for calibrating naked-eye 3D parameters according to an embodiment of this application.

[0020] Explanation of reference numerals in the attached figures: 1. Spatial position detection module; 2. Calibration image processing module; 3. Display quality analysis module; 4. Parameter joint optimization module; 5. Iterative update and solidification module. Detailed Implementation

[0021] In a first aspect, this application discloses an image adaptive calibration method for calibrating naked-eye 3D parameters.

[0022] Reference Figure 1 An image adaptive calibration method for calibrating naked-eye 3D parameters includes steps S101 to S105: Step S101: Control the terminal screen to project a positioning pattern with known geometric features, acquire images of the positioning pattern through the terminal camera, and calculate the real three-dimensional spatial parameters of the camera relative to the screen based on computer vision algorithms.

[0023] Specifically, the real three-dimensional spatial parameters include the vertical distance between the camera and the screen, the horizontal offset, the pitch angle, and the roll angle. In this embodiment, the positioning pattern refers to an image with a clear and known geometric structure (such as a checkerboard, concentric circles, or an array of coded points), whose feature points have predefined three-dimensional coordinates to provide a benchmark for spatial position calculation. The computer vision algorithm uses an algorithm based on feature point matching and pose estimation (such as the PnP algorithm) to infer the spatial position of the camera by establishing a correspondence between the image pixel coordinates and the real three-dimensional coordinates.

[0024] When the device starts or triggers the calibration process, the screen automatically displays a high-contrast checkerboard positioning pattern, which the front-facing camera captures in real time. The processor calls OpenCV's findChessboardCorners corner detection function and solvePnP pose estimation function to accurately calculate the three-dimensional spatial offset (ΔX, ΔY, ΔZ) and rotation angle (θx, θy, θz) of the camera relative to the screen plane by precisely analyzing the actual position and geometric deformation characteristics of the checkerboard corners in the captured image. This set of data will serve as the core "true reference benchmark" for the entire calibration process, providing a precise basis for subsequent parameter optimization.

[0025] Step S102: Load the current 3D calibration parameters, generate a position-aware calibration image based on the real three-dimensional spatial parameters, and project the calibration image according to the current 3D calibration parameters.

[0026] Specifically, in this embodiment, the current 3D calibration parameters refer to the core parameters currently stored in the terminal for controlling the naked-eye 3D display, including virtual raster pitch, tilt angle, and viewpoint mapping matrix. The position-aware calibration image is a dedicated calibration map adapted to the actual position of the camera, for example, marking the left-eye viewpoint area with an "L" label and the right-eye viewpoint area with an "R" label to ensure that the camera can acquire a crosstalk-free label image under ideal parameters. The system first loads the 3D calibration parameters currently stored in the terminal, covering key configurations such as virtual raster pitch, tilt angle, and viewpoint mapping matrix. Combining the previously acquired real spatial position data of the camera, the system dynamically generates a dedicated calibration image—marking the left-eye viewpoint area with an "L" label and the right-eye viewpoint area with an "R" label. Since the plane offset (ΔX, ΔY) of the camera is clear, the system can predefine the clear and non-overlapping visual state that the "L" and "R" labels should present under ideal parameters, establishing a reference standard for subsequent effect evaluation.

[0027] Step S103: Acquire the actual projected calibration image through the camera, quantitatively analyze the key quality indicators of the calibration image, calculate the deviation between the key quality indicators and the preset ideal value, and obtain the performance error vector.

[0028] Specifically, in this embodiment, key quality indicators include crosstalk rate and viewpoint position error; quantitative analysis refers to converting visual effects into calculable numerical indicators through image algorithms; preset ideal values ​​are the optimal target values ​​for naked-eye 3D display (such as an ideal value of 0 for crosstalk rate and an ideal value of 0 for viewpoint position error); the performance error vector is a set of deviations between key quality indicators and ideal values, comprehensively reflecting the degree of inaccuracy of the current 3D display.

[0029] The camera then captures the actual calibrated image projected onto the screen. The processor uses image analysis algorithms to quantify and evaluate two core quality indicators: crosstalk rate, determined by detecting the grayscale leakage of the "R" marker within the "L" marker area; and viewpoint position error, calculated by comparing the deviation between the centroid pixel coordinates of the "L" and "R" markers and their theoretical coordinates. The actual measured values ​​of the two indicators are compared with the ideal values ​​(crosstalk rate = 0, position error = 0) to construct a multi-dimensional performance error vector, comprehensively reflecting the inaccuracy of the current 3D display.

[0030] Step S104: With the goal of minimizing the performance error vector and the real three-dimensional space parameters as constraints, calculate the adjustment amount of the 3D calibration parameters through an optimization algorithm.

[0031] Specifically, in this embodiment, the optimization algorithm can be gradient descent algorithm, genetic algorithm, etc., the core of which is to find the optimal solution in the parameter adjustment space; the 3D calibration parameter adjustment amount refers to the correction value of the current 3D calibration parameters to reduce performance error, such as fine-tuning the translation amount of the viewpoint mapping matrix, the pitch parameter of the virtual raster, etc., to ensure that the adjusted parameters can adapt to the real position of the camera.

[0032] This step uses the norm of the performance error vector (such as root mean square error) as the objective function to minimize the core optimization algorithm, with the camera's real-space position as the constraint. Through intelligent optimization techniques such as gradient descent and genetic algorithms, the 3D calibration parameters are iteratively adjusted (e.g., fine-tuning the viewpoint mapping matrix by 0.5 pixels in the X direction each time), simultaneously simulating the impact of parameter adjustments on performance error. The algorithm continuously searches for and locks the optimal parameter combination that minimizes the error, achieving precise matching between position, quality, and parameters.

[0033] Step S105: Apply the parameter adjustment amount to the 3D display engine to update the 3D calibration parameters, repeat steps S102 to S104 until the performance error vector meets the preset threshold, and solidify the final 3D calibration parameters.

[0034] Specifically, in this embodiment, the 3D display engine is the core module in the terminal responsible for parsing 3D calibration parameters and driving the screen to display 3D effects; the preset threshold is a pre-set performance qualification standard (such as crosstalk rate <2%, viewpoint position error <0.5 pixels) to ensure that the display effect reaches the accuracy acceptable to the user; the fixed parameters are the current optimal parameters written into the terminal's non-volatile memory to ensure that the parameters are not lost after the device restarts or is powered off, and are used as the default parameters for subsequent 3D display.

[0035] The optimal parameter combination obtained through optimization is written into the 3D display driver. The system automatically repeats the process of "parameter-sensitive image projection - display effect quality assessment - joint parameter optimization" to ensure that the parameter optimization results converge and stabilize. When the performance error vector meets the preset threshold (e.g., crosstalk rate <2%, viewpoint position error <0.5 pixels), the calibration process terminates, and the finally optimized parameters are fixed and stored as the default benchmark parameters for subsequent naked-eye 3D display on the terminal, ensuring long-term stable visual effects.

[0036] Reference Figure 2 In one embodiment of this example, the step S101 of solving the real three-dimensional spatial parameters based on computer vision algorithms includes steps S201 to S207: Step S201: Extract several non-coplanar feature points from the positioning pattern and record the known three-dimensional coordinates of each feature point.

[0037] Specifically, feature points refer to easily identifiable key locations in the positioning pattern (such as the corners of a checkerboard or the center of a coding point), and their three-dimensional coordinates are already defined when designing the positioning pattern; non-coplanar feature points refer to feature points that are not on the same plane, which can ensure the uniqueness and accuracy of spatial position calculation and avoid ambiguity caused by planar feature points; in this embodiment, six non-coplanar feature points can be extracted.

[0038] Step S202: Perform noise filtering on the acquired positioning pattern image to remove false feature points caused by environmental interference.

[0039] Specifically, in this embodiment, noise filtering uses a Gaussian filtering algorithm to eliminate image interference caused by changes in ambient light and camera sensor noise; false feature points are non-real feature points that are misidentified in the image (such as false corner points formed by shadows). By comparing the geometric distribution pattern of feature points, false feature points that do not meet the preset conditions are eliminated to ensure the accuracy of subsequent calculations.

[0040] Step S203: Establish the correspondence between the camera intrinsic parameter matrix and the 3D coordinates of feature points and the image pixel coordinates, and solve for the extrinsic parameter matrix.

[0041] Specifically, in this embodiment, the camera intrinsic parameter matrix includes inherent camera parameters such as focal length and principal point coordinates, which have been calibrated before leaving the factory; the correspondence between the three-dimensional coordinates of feature points and the pixel coordinates of the image is established by matching the real feature points of the positioning pattern with the feature points identified in the image; the extrinsic parameter matrix is ​​a matrix describing the position and orientation of the camera relative to the screen coordinate system, which is obtained by solving the PnP algorithm.

[0042] Step S204: Decompose the rotation matrix and translation vector from the extrinsic parameter matrix, and calculate the vertical distance, horizontal offset, pitch angle and roll angle.

[0043] Specifically, in this embodiment, the rotation matrix reflects the camera's attitude angles (pitch angle, roll angle, etc.), the translation vector reflects the spatial offset of the camera relative to the screen origin; the vertical distance is the vertical distance (ΔZ) from the camera's optical center to the screen plane, the horizontal offset is the horizontal offset of the camera on the screen plane (ΔX), the pitch angle is the rotation angle of the camera around the horizontal axis (θx), and the roll angle is the rotation angle of the camera around the vertical axis (θz), which are obtained through matrix decomposition and trigonometric function conversion.

[0044] Step S205: Continuously acquire several frames of positioning pattern images and repeatedly solve them, and calculate the mean and variance of several sets of solution results.

[0045] Specifically, in this embodiment, 10-20 frames of images are continuously acquired, and each frame of image is independently calculated to obtain multiple sets of three-dimensional spatial parameters; the mean is the arithmetic mean of multiple sets of calculation results, reflecting the average level of the parameters; the variance is an indicator that measures the degree of dispersion of multiple sets of calculation results, and the smaller the variance, the more stable the calculation results are.

[0046] Step S206: If the variance is less than the set stability threshold, the mean is used as the final true three-dimensional spatial parameter.

[0047] Specifically, in this embodiment, the stability threshold is preset to 0.05 (the unit is adapted according to the parameter type, such as distance in mm and angle in degrees). When the variance is less than this threshold, it indicates that the calculation result fluctuates little and has high reliability. At this time, the mean is taken as the final parameter to ensure the accuracy of the benchmark for subsequent calibration.

[0048] Step S207: If the variance is greater than or equal to the preset stability threshold, re-acquire several frames of positioning pattern images and repeat the calculation process.

[0049] Specifically, in this embodiment, a variance greater than or equal to the set stability threshold means that the solution result is greatly disturbed, which may be caused by sudden changes in ambient light or temporary occlusion of the camera; when re-acquiring, the number of acquisition frames can be increased (e.g., increased to 30 frames) to further improve the stability of the solution result until the variance meets the stability threshold requirement.

[0050] Reference Figure 3 In one embodiment of this example, the key quality indicators of the quantitative analysis calibration image in step S103 include steps S301 to S306: Step S301: Extract the first pixel region corresponding to the first identifier of the left eye viewpoint region and the second pixel region corresponding to the second identifier of the right eye viewpoint region in the calibration image.

[0051] Specifically, in this embodiment, the left eye viewpoint region is the area on the screen specifically projecting the left eye image, and the first identifier is the "L" symbol in this area; the right eye viewpoint region is the area specifically projecting the right eye image, and the second identifier is the "R" symbol in this area; the first pixel region is the set of all pixels where the "L" symbol is located in the calibration image, and the second pixel region is the set of all pixels where the "R" symbol is located, extracted by an image segmentation algorithm.

[0052] Step S302: Obtain the first grayscale mean value corresponding to the second identifier in the first pixel region and the second grayscale mean value corresponding to the second identifier in the second pixel region.

[0053] Specifically, the grayscale mean is the arithmetic mean of the grayscale values ​​of all pixels within a pixel region, with a grayscale value range of 0-255 (0 being pure black and 255 being pure white); the first grayscale mean is the average grayscale value of the "R" symbol pixels mixed in within the first pixel region, reflecting the degree of interference from the right eye image to the left eye region; the second grayscale mean is the average grayscale value of the "R" symbol itself within the second pixel region, serving as a reference benchmark.

[0054] Step S303: Calculate the grayscale ratio based on the first grayscale mean and the second grayscale mean, and use the grayscale ratio as the crosstalk rate.

[0055] Specifically, in this embodiment, the crosstalk rate is calculated as follows: Crosstalk rate = first grayscale mean / second grayscale mean × 100%. This ratio directly reflects the degree of crosstalk between the left and right eye images. The smaller the ratio, the less crosstalk there is and the clearer the 3D display effect.

[0056] Step S304: Obtain the first pixel coordinates corresponding to the first identifier and the second pixel coordinates corresponding to the second identifier in the calibration image using a preset image algorithm.

[0057] Specifically, in this embodiment, the preset image algorithm uses the centroid extraction algorithm, which calculates the gray-scale centroid of the pixels in the marked area to obtain the center pixel coordinates of the mark; the first pixel coordinates are the center pixel coordinates (x1, y1) of the "L" symbol, and the second pixel coordinates are the center pixel coordinates (x2, y2) of the "R" symbol. The coordinates are with the upper left corner of the screen as the origin, the horizontal rightward x-axis as the x-axis, and the vertical downward y-axis as the y-axis.

[0058] Step S305: Retrieve the first ideal coordinates corresponding to the first identifier and the second ideal coordinates corresponding to the second identifier from the preset ideal values.

[0059] Specifically, in this embodiment, the preset ideal coordinates are theoretically optimal coordinates calculated based on the camera's real three-dimensional spatial parameters, that is, the pixel position that the identifier should be when the 3D parameters are fully adapted; the first ideal coordinates (x1', y1') are the ideal center coordinates of the "L" symbol, and the second ideal coordinates (x2', y2') are the ideal center coordinates of the "R" symbol, which are stored in the terminal's local parameter library.

[0060] Step S306: Obtain the viewpoint position error based on the first pixel coordinates, the second pixel coordinates, the first ideal coordinates, and the second ideal coordinates.

[0061] Specifically, in this embodiment, the viewpoint position error is the core indicator for measuring the deviation between the actual position and the ideal position of the marker. The smaller the deviation, the more accurate the spatial allocation of the 3D image and the better the stereoscopic visual experience for the viewer.

[0062] Reference Figure 4In one embodiment of this example, step S306, based on the first pixel coordinates, the second pixel coordinates, the first ideal coordinates, and the second ideal coordinates, obtains the viewpoint position error, including steps S401 to S407: Step S401: Obtain the first position error value based on the first pixel coordinates and the first ideal coordinates.

[0063] Specifically, the first position error value reflects the pixel distance between the actual position and the ideal position of the "L" symbol.

[0064] Step S402: Obtain the second position error value based on the second pixel coordinates and the second ideal coordinates.

[0065] Specifically, the second position error value reflects the pixel distance between the actual position and the ideal position of the "R" symbol. In this embodiment, both the first position error value and the second position error value can be calculated using Euclidean distance.

[0066] Step S403: Determine whether the specified configuration instruction has been detected.

[0067] Specifically, in this embodiment, the specified configuration instruction is a personalized parameter configuration instruction issued by the user or the system (such as prioritizing the optimization of the left eye display effect), and the existence of the instruction is detected through the system interface of the terminal.

[0068] Step S404: If detected, obtain the viewpoint position error based on the specified configuration command, the first position error value, or the second position error value.

[0069] Specifically, in this embodiment, if the instruction requires "prioritizing optimization of the left eye", then the first position error value is used as the viewpoint position error; if the instruction requires "prioritizing optimization of the right eye", then the second position error value is used as the viewpoint position error, ensuring that the parameter optimization fits the personalized needs.

[0070] Step S405: If not detected, determine whether there is a specified runtime environment requirement.

[0071] Specifically, in this embodiment, the specified operating environment requirement is the terminal's usage scenario requirement (such as in-vehicle display scenario, VR adaptation scenario), and the existence of the requirement is determined by reading the terminal's scenario mode parameters.

[0072] Step S406: If it exists, determine whether the specified operating environment requirement is a rigid requirement.

[0073] Specifically, in this embodiment, rigid requirements refer to scenarios with extremely high and uncompromising display accuracy requirements (such as professional 3D design display), while non-rigid requirements refer to scenarios with moderate accuracy requirements (such as daily movie watching). The priority level of the scenario requirements is used to determine the requirements.

[0074] Step S407: If it is true, then select the maximum value between the first position error value and the second position error value as the viewpoint position error.

[0075] Specifically, in this embodiment, under rigid requirements, it is necessary to ensure that the display effect of both the left and right eyes reaches a high precision. The maximum value is selected as the error index, which can force parameter optimization and reduce the maximum deviation between the two eyes, so as to avoid the single viewpoint effect from failing to meet the standard.

[0076] In one embodiment of this example, if step S406 exists, then after determining whether the specified operating environment requirement is a rigid requirement, steps S501 to S505 are further included: Step S501: If not, determine whether the terminal camera supports independent adjustment of eye parameters.

[0077] Specifically, in this embodiment, the ability to independently adjust binocular parameters means that the terminal can adjust the 3D calibration parameters of the left and right eyes separately, and determine whether the terminal supports this function by querying the terminal's hardware capability parameters.

[0078] Step S502: If not supported, then obtain the average of the first position error value and the second position error value as the viewpoint position error.

[0079] Specifically, in this embodiment, when the parameters of both eyes cannot be adjusted independently, taking the average value can balance the deviation of both eyes, avoid excessive deviation of a single viewpoint, and ensure a balanced overall display effect.

[0080] Step S503: If supported, obtain the daily usage requirements of the terminal camera.

[0081] Specifically, in this embodiment, daily usage requirements are obtained using a dual-mode approach: "factory-customized configuration + subsequent usage record iteration and optimization." For factory calibration scenarios, the customized configuration parameters preset by the terminal are read first. If it is a specific customer order, the customer's requirements can be entered into the system during the production process (e.g., if the customer clearly states that the equipment is used for long-term 3D design work, it is directly calibrated as "extremely long usage requirements"; if the customer has no special requirements, the default configuration is "non-extremely long usage requirements"). After the equipment is delivered to the user, the system automatically accumulates the 3D function usage time and dynamically updates it during subsequent calibrations (periodic calibration, triggered calibration) according to the rule of "daily average usage exceeding 4 hours is judged as an extreme usage requirement." This satisfies the customized requirements of specific customers and adapts to the changes in daily usage scenarios of ordinary users.

[0082] Step S504: If the daily use requirement is an extended use requirement, then the average of the first position error value and the second position error value is obtained as the viewpoint position error.

[0083] Specifically, in this embodiment, in scenarios involving extended use, viewing comfort should be prioritized. Taking the average value can avoid visual fatigue caused by excessive differences in binocular deviation and improve the experience of long-term use.

[0084] Step S505: If the daily use requirement is not an extended use requirement, then select the maximum value between the first position error value and the second position error value as the viewpoint position error.

[0085] Specifically, in this embodiment, display accuracy can be prioritized in non-extended usage scenarios, and the maximum value can be selected as the error index, which can more quickly reduce the maximum deviation between the eyes and improve the sharpness of the 3D effect.

[0086] Reference Figure 5 In one embodiment of this example, steps S102 to S104 are repeated until the performance error vector meets a preset threshold, and the final 3D calibration parameters are solidified, including steps S601 to S609: Step S601: After each update of the 3D calibration parameters, reload the updated parameters to generate a calibration image and project it.

[0087] Specifically, in this embodiment, the current parameters in the 3D display engine are replaced immediately after the parameters are updated, and the calibration image is regenerated according to the new parameters to ensure real-time matching between the image and the parameters and avoid evaluation errors caused by parameter lag.

[0088] Step S602: Acquire the newly projected calibration image using a camera, quantify and analyze key quality indicators, and calculate a new performance error vector.

[0089] Specifically, in this embodiment, the data acquisition and analysis process is consistent with step S103, ensuring that the evaluation criteria are consistent for each iteration, and the new performance error vector directly reflects the effect changes after the parameter update.

[0090] Step S603: Compare the new performance error vector with the preset threshold to determine whether the crosstalk rate and viewpoint position error are both less than the corresponding preset threshold.

[0091] Specifically, in this embodiment, the preset thresholds include a crosstalk rate threshold (e.g., 2%) and a viewpoint position error threshold (e.g., 0.5 pixels). Both indicators must be met simultaneously to determine that the 3D display effect is qualified.

[0092] Step S604: If both indicators are less than the corresponding preset threshold, stop the iteration process and solidify the current 3D calibration parameters.

[0093] Step S605: If any indicator is greater than or equal to the corresponding preset threshold, determine whether the current iteration number exceeds the maximum iteration number.

[0094] Specifically, in this embodiment, the maximum number of iterations is preset to 20 times to avoid infinite loops in the iteration process and to balance calibration efficiency and effectiveness.

[0095] Step S606: If the maximum number of iterations has not been exceeded, return to step S102 to continue the calibration process.

[0096] Specifically, in this embodiment, after returning, the adjustment amount is further optimized based on the current performance error vector and parameter adjustment experience, continuously approaching the optimal parameters.

[0097] Step S607: If the maximum number of iterations has been exceeded, then detect the difference between the current performance error vector and the target value of the preset threshold.

[0098] Specifically, in this embodiment, the target gap refers to the difference between the current crosstalk rate and the threshold, and the difference between the current viewpoint position error and the threshold, which is used to evaluate the approximation degree of parameter optimization.

[0099] Step S608: If the target difference is less than or equal to the preset difference threshold, the current parameters are used as a transitional solution, and the user is prompted to recalibrate periodically.

[0100] Specifically, in this embodiment, the preset difference thresholds are crosstalk rate difference ≤1% and viewpoint position error difference ≤0.3 pixels. At this time, although the parameters are not fully up to standard, they are close to the qualified standard, and the transition scheme can meet the needs of daily use. The periodic calibration prompt is pushed through the terminal pop-up window or notification, and it is recommended that users calibrate once a month.

[0101] Step S609: If the difference is greater than the preset difference threshold, switch to the basic 3D calibration parameters, turn off the adaptive calibration function and prompt for troubleshooting.

[0102] Specifically, in this embodiment, the basic 3D calibration parameters are the general parameters used when the device leaves the factory, ensuring that users can still use the 3D function (the effect may be poor); the troubleshooting tips include cleaning the camera, ensuring the screen is not obstructed, and ensuring the ambient light is moderate. If the problem is still not resolved, it is recommended to check the hardware.

[0103] In one embodiment of this example, steps S701 to S707 are further included: Step S701: Record the performance error vector and the corresponding 3D calibration parameters for each iteration to form an iteration history record.

[0104] Specifically, in this embodiment, the iteration history records the timestamp of each iteration, the performance error vector (crosstalk rate, viewpoint position error), the adjustment amount of 3D calibration parameters, and the current parameter value, which facilitates tracing the optimization process and analyzing the convergence trend.

[0105] Step S702: Calculate the difference in error vectors between two adjacent iterations to determine the error convergence trend.

[0106] Specifically, in this embodiment, the error vector difference is the difference between the subsequent error vector and the previous error vector (e.g., crosstalk rate difference = current crosstalk rate - previous crosstalk rate). The convergence trend includes three cases: decreasing (error continues to decrease), stable (error fluctuation is less than 0.1%), and increasing (error continues to increase).

[0107] Step S703: If the error difference shows a decreasing trend and the current error vector meets the preset threshold, then stop the iteration and fix the parameters.

[0108] Specifically, in this embodiment, the decreasing trend indicates that the parameter adjustment direction is correct. If the indicator meets the target, the iteration can be stopped directly to avoid resource waste caused by over-optimization.

[0109] Step S704: If the error difference tends to stabilize but does not meet the threshold, then reduce the parameter adjustment step size and continue iterating.

[0110] Specifically, in this embodiment, the parameter adjustment step size is the magnitude of each parameter modification. Reducing the step size (such as starting from 1 / 2 of the original step size) can achieve fine optimization and help the error approach the threshold.

[0111] Step S705: If the error difference shows an increasing trend, backtrack to the calibration parameters of the previous valid iteration and adjust the optimization direction.

[0112] Specifically, in this embodiment, the increasing trend indicates that the current parameter adjustment direction is incorrect, the previous effective iteration refers to the historical iteration where the error is less than the current iteration, and the adjustment optimization direction refers to modifying the parameters in reverse (such as changing the original increase of the grating pitch to decrease) and trying optimization again.

[0113] Step S706: Count the number of iterations. If the number of iterations reaches the preset upper limit but the threshold is still not met, then detect the working status of the camera and screen.

[0114] Specifically, in this embodiment, the preset upper limit is consistent with the maximum number of iterations in step S605 (20 times); the working status detection includes whether the camera is normally acquiring images, whether the screen is normally projecting patterns, and whether the optical components (such as lenticular lens gratings) are intact.

[0115] Step S707: If the equipment is abnormal, generate a fault alarm message and terminate the calibration process.

[0116] Specifically, in this embodiment, the fault alarm information includes abnormal components (such as "camera acquisition abnormality") and troubleshooting suggestions, which are communicated to the user through terminal pop-ups, sound prompts, etc.; terminating the calibration process can avoid invalid iterations in abnormal states and protect the device hardware.

[0117] Secondly, this application also discloses an image adaptive calibration system for calibrating naked-eye 3D parameters.

[0118] Reference Figure 6 An image adaptive calibration system for calibrating naked-eye 3D parameters includes: The spatial position detection module 1 is used to control the terminal screen to project a positioning pattern with known geometric features, acquire images of the positioning pattern through the terminal camera, and calculate the real three-dimensional spatial parameters of the camera relative to the screen based on computer vision algorithms. The calibration image processing module 2 is used to load the current 3D calibration parameters, generate a position-aware calibration image based on the real three-dimensional spatial parameters, and project the calibration image according to the current 3D calibration parameters. Display quality analysis module 3 is used to acquire the actual projected calibration image through the camera, quantitatively analyze the key quality indicators of the calibration image, calculate the deviation of the key quality indicators from the preset ideal value, and obtain the performance error vector. The parameter joint optimization module 4 is used to calculate the adjustment amount of 3D calibration parameters through an optimization algorithm with the goal of minimizing the performance error vector and the real three-dimensional space parameters as constraints. The iterative update and solidification module 5 is used to apply the parameter adjustment amount to the 3D display engine to update the 3D calibration parameters until the performance error vector meets the preset threshold, and then solidify the final 3D calibration parameters.

[0119] Thirdly, this application discloses a smart terminal, including a memory and a processor. The memory stores a computer program that can run on the processor. When the processor loads the computer program, it executes an image adaptive calibration method for calibrating naked-eye 3D parameters as described in the above embodiment.

[0120] Fourthly, embodiments of this application disclose a computer-readable storage medium, wherein a computer program is stored in the computer-readable storage medium, and when the computer program is loaded by a processor, it executes an image adaptive calibration method for calibrating naked-eye 3D parameters as described above.

[0121] The above are all preferred embodiments of this application, and are not intended to limit the scope of protection of this application. Therefore, all equivalent changes made in accordance with the structure, shape and principle of this application should be covered within the scope of protection of this application.

Claims

1. An image adaptive calibration method for calibrating naked-eye 3D parameters, characterized in that, include: S101. Control the terminal screen to project a positioning pattern with known geometric features, acquire an image of the positioning pattern through the terminal camera, and calculate the real three-dimensional spatial parameters of the camera relative to the screen based on a computer vision algorithm; S102. Load the current 3D calibration parameters, generate a position-aware calibration image according to the real three-dimensional spatial parameters, and project the calibration image according to the current 3D calibration parameters. S103. Acquire the actual projected calibration image through the camera, quantitatively analyze the key quality indicators of the calibration image, calculate the deviation between the key quality indicators and the preset ideal value, and obtain the performance error vector. S104. With minimizing the performance error vector as the objective and the real three-dimensional space parameters as constraints, calculate the adjustment amount of the 3D calibration parameters through an optimization algorithm; S105. Apply the parameter adjustment amount to the 3D display engine to update the 3D calibration parameters, repeat steps S102 to S104 until the performance error vector meets the preset threshold, and solidify the final 3D calibration parameters. The true 3D spatial parameters include the vertical distance between the camera and the screen, the horizontal offset, the pitch angle, and the roll angle; the calculation of the true 3D spatial parameters based on computer vision algorithms includes: Extract several non-coplanar feature points from the positioning pattern and record the known three-dimensional coordinates of each feature point; The acquired positioning pattern image is subjected to noise filtering to remove false feature points caused by environmental interference; Establish the correspondence between the camera intrinsic parameter matrix and the 3D coordinates of feature points and the image pixel coordinates, and solve for the extrinsic parameter matrix; The rotation matrix and translation vector are decomposed from the extrinsic parameter matrix, and the vertical distance, horizontal offset, pitch angle and roll angle are calculated. Several frames of positioning pattern images are continuously acquired and repeatedly processed to calculate the mean and variance of several sets of processing results; If the variance is less than the set stability threshold, the mean will be used as the final true three-dimensional space parameter. If the variance is greater than or equal to the preset stability threshold, re-acquire several frames of positioning pattern images and repeat the calculation process.

2. The image adaptive calibration method for calibrating naked-eye 3D parameters according to claim 1, characterized in that, The key quality indicators include crosstalk rate and viewpoint position error; The key quality indicators for the quantitative analysis of the calibration image include: Extract the first pixel region corresponding to the first identifier of the left eye viewpoint region and the second pixel region corresponding to the second identifier of the right eye viewpoint region from the calibration image; Obtain the first grayscale average value corresponding to the second identifier in the first pixel region and the second grayscale average value corresponding to the second identifier in the second pixel region; Based on the first grayscale mean and the second grayscale mean, the grayscale ratio is calculated, and the grayscale ratio is used as the crosstalk rate; The first pixel coordinates corresponding to the first identifier and the second pixel coordinates corresponding to the second identifier in the calibration image are obtained by using a preset image algorithm. Retrieve the first ideal coordinates corresponding to the first identifier and the second ideal coordinates corresponding to the second identifier from the preset ideal values; The viewpoint position error is obtained based on the first pixel coordinates, the second pixel coordinates, the first ideal coordinates, and the second ideal coordinates.

3. The image adaptive calibration method for calibrating naked-eye 3D parameters according to claim 2, characterized in that, The step of obtaining the viewpoint position error based on the first pixel coordinates, the second pixel coordinates, the first ideal coordinates, and the second ideal coordinates includes: Based on the first pixel coordinates and the first ideal coordinates, obtain the first position error value; Based on the second pixel coordinates and the second ideal coordinates, obtain the second position error value; Determine whether the specified configuration command has been detected; If detected, the viewpoint position error is obtained based on the specified configuration instruction, the first position error value, or the second position error value. If not detected, determine whether a specified runtime environment requirement exists; If it exists, determine whether the specified operating environment requirement is a rigid requirement; If it is true, then the maximum value between the first position error value and the second position error value is selected as the viewpoint position error.

4. The image adaptive calibration method for calibrating naked-eye 3D parameters according to claim 3, characterized in that, If it exists, then after determining whether the specified operating environment requirement is a rigid requirement, the process further includes: If not, determine whether the terminal camera supports independent adjustment of dual-eye parameters; If not supported, the average of the first position error value and the second position error value is obtained as the viewpoint position error; If supported, obtain the daily usage requirements of the terminal camera; If the daily use requirement is an extended use requirement, then the average of the first position error value and the second position error value is obtained as the viewpoint position error; If the daily usage requirement is not the extended usage requirement, then the maximum value between the first position error value and the second position error value is selected as the viewpoint position error.

5. The image adaptive calibration method for calibrating naked-eye 3D parameters according to claim 1, characterized in that, The repeated steps S102 to S104 until the performance error vector meets the preset threshold, and the final 3D calibration parameters are solidified, include: Each time the 3D calibration parameters are updated, the updated parameters are reloaded to generate a calibration image and project it. The newly projected calibration images are captured by a camera, key quality indicators are quantitatively analyzed, and a new performance error vector is calculated. The new performance error vector is compared with a preset threshold to determine whether the crosstalk rate and viewpoint position error are both less than the corresponding preset threshold. If both indicators are less than the corresponding preset threshold, the iteration process is stopped and the current 3D calibration parameters are fixed. If any indicator is greater than or equal to the corresponding preset threshold, then determine whether the current iteration number exceeds the maximum iteration number; If the maximum number of iterations has not been exceeded, return to step S102 to continue the calibration process; If the maximum number of iterations has been exceeded, then the difference between the current performance error vector and the target value of the preset threshold is detected. If the target difference is less than or equal to the preset difference threshold, the current parameters will be used as a transitional solution, and the user will be prompted to recalibrate periodically. If the difference exceeds the preset difference threshold, switch to the basic 3D calibration parameters, disable the adaptive calibration function, and prompt for troubleshooting.

6. The image adaptive calibration method for calibrating naked-eye 3D parameters according to claim 1, characterized in that, Also includes: Record the performance error vector and corresponding 3D calibration parameters for each iteration to form an iteration history record; Calculate the difference in error vectors between two adjacent iterations to determine the error convergence trend; If the error difference shows a decreasing trend and the current error vector meets the preset threshold, then stop the iteration and fix the parameters; If the error difference tends to stabilize but does not meet the threshold, then reduce the parameter adjustment step size and continue iterating; If the error difference shows an increasing trend, then backtrack to the calibration parameters of the previous valid iteration and adjust the optimization direction; The number of iterations is counted. If the threshold is not met even after the number of iterations reaches the preset upper limit, the working status of the camera and screen is detected. If the equipment malfunctions, a fault alarm will be generated, and the calibration process will be terminated.

7. An image adaptive calibration system for calibrating naked-eye 3D parameters, used to perform the method according to any one of claims 1 to 6, characterized in that, include: The spatial position detection module (1) is used to control the terminal screen to project a positioning pattern with known geometric features, acquire the image of the positioning pattern through the terminal camera, and calculate the real three-dimensional spatial parameters of the camera relative to the screen based on computer vision algorithms. The calibration image processing module (2) is used to load the current 3D calibration parameters, generate a position-aware calibration image according to the real three-dimensional spatial parameters, and project the calibration image according to the current 3D calibration parameters. The display quality analysis module (3) is used to acquire the actual projected calibration image through the camera, quantitatively analyze the key quality indicators of the calibration image, calculate the deviation of the key quality indicators from the preset ideal value, and obtain the performance error vector. The parameter joint optimization module (4) is used to calculate the adjustment amount of 3D calibration parameters by means of optimization algorithm with the goal of minimizing the performance error vector and the real three-dimensional space parameters as constraints. The iterative update and solidification module (5) is used to apply the parameter adjustment amount to the 3D display engine to update the 3D calibration parameters until the performance error vector meets the preset threshold and the final 3D calibration parameters are solidified.

8. A smart terminal, comprising a memory and a processor, characterized in that, The memory is used to store computer programs that can run on the processor, and when the processor loads the computer program, it executes the method of any one of claims 1 to 6.

9. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is loaded by the processor, it executes the method of any one of claims 1 to 6.

Images