AR glasses binocular visual effect comprehensive optimization system and method, and computer readable storage medium

Abstract

The invention discloses an AR glasses binocular visual effect comprehensive optimization system and method, and a computer readable storage medium, and the system comprises a human-computer interface module, a parameter configuration module, a camera calibration module, an optical-mechanical image combination judgment module, a complete machine closed-loop adjustment module, a binocular consistency optimization module, and a permission and authorization module. All the modules work cooperatively through a unified data stream, and for a Micro-OLED light machine and waveguide sheet optical architecture, image combination precision judgment including translation, rotation, zooming and nonlinear distortion and closed-loop optimization of brightness consistency calibration are achieved, and quantized visual effect indexes are output.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of AR glasses, and particularly relates to an AR glasses binocular visual effect comprehensive optimization system and method and a computer readable storage medium. BACKGROUND

[0002] With the rapid development of augmented reality (AR) technology, intelligent display glasses are evolving towards thinness and high brightness. The optical architecture of "Micro-OLED light engine + waveguide sheet" has become the mainstream choice in the industry due to its excellent performance. The comprehensive performance of binocular visual effect (including image accuracy and brightness consistency) directly determines the product yield and user satisfaction. Optimization of related detection and correction technologies has become a key requirement in the industry. Insufficient image accuracy can cause ghosting and distortion, and poor brightness consistency can cause visual fatigue. Both need to be optimized to achieve a high-quality user experience.

[0003] In the prior art, solutions for binocular image alignment mainly focus on simplifying detection equipment or optimizing hardware structure to reduce errors, but there is no comprehensive optimization scheme for "image accuracy + brightness consistency". For example, Chinese invention patent application CN202211586810 discloses a binocular image alignment method based on a single lens. It captures left and right waveguide sheet images through a single lens, obtains displacement and angle values, and adjusts the light engine after comparing with the calibration picture to achieve image alignment. This scheme reduces the assembly tolerance of the double lens and reduces the hardware cost, but it can only cover simple deviations such as displacement and angle, and cannot adapt to the specific scaling and nonlinear distortion of the "Micro-OLED light engine + waveguide sheet" architecture. It does not provide quantitative image accuracy evaluation indicators, and does not involve detection and calibration of binocular brightness consistency, making it difficult to meet the comprehensive visual effect optimization requirements.

[0004] Another related prior art is the binocular image alignment detection device and method disclosed in Chinese invention patent application CN202211213202. It uses a single test camera to sequentially capture the display images of two sets of monocular optical modules, calculates the rotational parallax and vertical parallax to achieve detection. This scheme can only handle rotational and translational deviations, lacks detection capability for scaling and nonlinear distortion, and uses a linear detection process without a closed-loop verification mechanism. The reliability of the detection results depends on manual judgment and cannot be directly integrated into an automated production line. At the same time, this scheme does not address the binocular brightness consistency issue and does not provide quantitative analysis and calibration means for brightness differences, making it impossible to solve the problem of decreased visual experience caused by uneven brightness.

[0005] In summary, the prior art is a single-dimensional image combination detection or correction scheme, and there is no comprehensive optimization scheme for the specific architecture of "Micro-OLED light engine + waveguide sheet" to achieve "image combination accuracy + brightness consistency". It is difficult to calculate the compound deviation under this architecture, and there is a lack of quantitative detection and closed-loop calibration mechanism for brightness consistency. Moreover, there is no comprehensive quantitative judgment index with pixel-level accuracy and automatic adaptation capability for production lines, resulting in poor product debugging consistency, low yield, and uneven user experience. Therefore, there is an urgent need for a binocular visual effect comprehensive optimization scheme that is specifically adapted to the "Micro-OLED light engine + waveguide sheet" architecture, can accurately detect compound deviation and brightness difference, has comprehensive quantitative judgment capability, and is adapted to automatic production lines to solve the limitations of the prior art. SUMMARY

[0006] The technical problem to be solved by the present application is to provide an AR glasses binocular visual effect comprehensive optimization system and method, and a computer readable storage medium, which can be used for the "Micro-OLED light engine + waveguide sheet" optical architecture to realize image combination accuracy judgment including translation, rotation, scaling and nonlinear distortion, and closed-loop optimization of brightness consistency calibration, and output quantitative visual effect indicators.

[0007] To solve the above technical problems, the first aspect of the present application discloses an AR glasses binocular visual effect comprehensive optimization system, which comprises a human-computer interface module, a parameter configuration module, a camera calibration module, a light engine image combination judgment module, a complete machine closed-loop adjustment module, a binocular consistency optimization module and a permission and authorization module. Each module works cooperatively through unified data flow, and is used for the "Micro-OLED light engine + waveguide sheet" optical architecture to realize image combination accuracy judgment including translation, rotation, scaling and nonlinear distortion, and closed-loop optimization of brightness consistency calibration, and output quantitative visual effect indicators.

[0008] As an optional implementation, in the first aspect of the present application, the camera calibration module comprises an image acquisition unit and a calibration calculation unit: feature points are acquired through asymmetric circle point detection, monocular remapping matrix and stereo correction mapping are calculated, and the reprojection error is controlled within a preset range; the calibration calculation unit supports automatic verification of calibration results, and reacquires and detects if the results do not meet the requirements.

[0009] As another optional implementation, in the first aspect of the present application, the light engine image combination judgment module comprises a calibration parameter loading unit, an image correction unit, a circle point detection unit, a monocular pose detection unit and a dual light engine image combination judgment unit; the dual light engine image combination judgment unit calculates deviation statistics by comparing left and right eye circle point coordinates, and outputs pixel-level image combination accuracy quantitative results.

[0010] As a further optional implementation, in the first aspect of the application, the whole machine closed loop adjustment module comprises an adjustment vector calculation unit, a Bluetooth instruction sending unit, and an adjustment effect verification unit, the adjustment vector calculation unit uses the least square method to solve the optimal compensation parameters of translation, rotation, and scaling, and each unit cooperates to realize the closed loop process of "image combination determination-optical machine parameter adjustment-second image combination determination" until the image combination precision meets the standard.

[0011] As a further optional implementation, in the first aspect of the application, the binocular consistency optimization module is a matching unit of the binocular vision effect comprehensive optimization, comprising a full brightness detection unit, a brightness uniformity analysis unit, and a brightness calibration unit; the full brightness detection unit performs ROI sampling on the corrected image in a 4x12 grid, the brightness uniformity analysis unit calculates the min and max indexes of the binocular brightness difference, and the brightness calibration unit performs calibration operation according to the difference.

[0012] The second aspect of the application discloses an AR glasses binocular vision effect comprehensive optimization method, which is realized based on the system of the first aspect of the application, and comprises the following steps: Step 1: the parameter configuration module and the camera calibration module cooperatively complete system initialization, and the camera calibration module loads camera parameters, stereo calibration parameters, and remapping matrices; Step 2: the camera calibration module collects left and right eye original images, and the optical machine image combination determination module corrects the images through the remapping matrices; Step 3: the optical machine image combination determination module performs image combination precision determination: coordinates are obtained through asymmetric dot detection, monocular pose and binocular image combination deviation statistics are calculated, and image combination precision results are outputted; Step 4: if the image combination precision does not meet the standard, the whole machine closed loop adjustment module calculates an adjustment vector and sends it to the optical machine through Bluetooth, and then steps 2 to 3 are repeated to verify the effect until the standard is met; Step 5: the binocular consistency optimization module performs brightness consistency optimization: full brightness detection and grid sampling are performed on the corrected image, binocular brightness difference is calculated, if the difference is out of standard, brightness calibration is performed, and then sampling detection is repeated until the standard is met.

[0013] As an optional implementation, in the second aspect of the application, the detection precision of the image combination precision in step 3 is pixel level.

[0014] As a further optional implementation, in the second aspect of the present application, the adjustment performed by the whole machine closed loop adjustment module in step 4 includes digital pre-distortion compensation of the image rendering engine, the adjustment vector calculation unit of the whole machine closed loop adjustment module solves the optimal compensation parameters of translation, rotation and scaling by least square method, the image rendering engine performs digital pre-distortion compensation based on the parameters and independently corrects the non-linear distortion by a 3rd order polynomial algorithm, and the four types of compound deviations are corrected together.

[0015] As a further optional implementation, in the second aspect of the present application, the camera parameters in step 1 are obtained by "asymmetric circle point detection + single target positioning + double target positioning", and the calibration process supports deleting invalid images and reacquiring.

[0016] The third aspect of the present application discloses a computer readable storage medium, which stores a computer program, and the computer program is executed by a processor to realize the steps in the method disclosed in the second aspect of the present application.

[0017] Compared with the prior art, the embodiment of the present application has the following beneficial effects: Compared with the prior art, the embodiment of the present application integrates the image precision judgment and the brightness consistency calibration into a closed loop optimization system through the cooperative work of the seven modules and the unified data flow, breaks through the limitation of single dimension optimization, simultaneously covers two major core influencing factors of binocular vision effect, solves the problem that the existing technology cannot consider multiple dimensions, is specially designed for the "Micro-OLED light machine + waveguide sheet" architecture, can accurately detect and process four types of compound deviations of translation, rotation, scaling and non-linear distortion which are specific to the architecture, has stronger architecture adaptability and more outstanding optimization pertinence compared with general systems, focuses on the quantitative visual effect index output, uses objective data as the judgment basis for the image precision and the brightness consistency, gets rid of the dependence on traditional artificial subjective judgment, and provides standardized basis for visual effect evaluation, and the built-in closed loop logic of "detection-adjustment-verification" realizes real-time verification and iteration of the optimization effect through module cooperation, ensures that the image precision and the brightness consistency finally meet the standards, and has better optimization stability compared with linear processes. BRIEF DESCRIPTION OF DRAWINGS

[0018] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed in the embodiment description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0019] Figure 1 is a structural schematic diagram of an AR glasses binocular vision effect comprehensive optimization system disclosed by the embodiment of the present application; Figure 2is a flowchart of an AR glasses binocular visual effect comprehensive optimization method disclosed by the embodiment of the present application. Figure 3 is a light-mechanical image combination flowchart disclosed by the embodiment of the present application. Figure 4 is a whole machine image combination judgment flowchart disclosed by the embodiment of the present application. Figure 5 is a binocular consistency detection flowchart disclosed by the embodiment of the present application. DETAILED DESCRIPTION

[0020] In order for those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0021] Embodiment one Referring to Figure 1 , the embodiment of the present application discloses an AR glasses binocular visual effect comprehensive optimization system, which comprises a man-machine interface module 101, a parameter configuration module 102, a camera calibration module 103, a light-mechanical image combination judgment module 104, a whole machine closed-loop adjustment module 105, a binocular consistency optimization module 106 and a permission and authorization module 107. Each module works cooperatively through unified data flow. For the "Micro-OLED light machine + waveguide sheet" optical architecture, the system realizes the combination precision judgment including translation, rotation, scaling and nonlinear distortion, and the closed-loop optimization of brightness consistency calibration, and outputs quantitative visual effect indicators.

[0022] In the embodiment of the present application, through the cooperative work of the seven modules and the unified data flow, the combination precision judgment and the brightness consistency calibration are integrated into a closed-loop optimization system, breaking through the limitation of single-dimensional optimization, covering two core influencing factors of binocular visual effect, solving the problem that the existing system cannot consider multiple dimensions; specially designed for the "Micro-OLED light machine + waveguide sheet" architecture, it can accurately detect and process four types of composite deviations, i.e. translation, rotation, scaling and nonlinear distortion, which are specific to this architecture. Compared with general systems, the architecture has stronger adaptability and more prominent optimization pertinence; the core function focuses on the output of quantitative visual effect indicators, and both the combination precision and the brightness consistency are determined based on objective data, which breaks the dependence on traditional artificial subjective judgment and provides standardized basis for visual effect evaluation; the built-in closed-loop logic of "detection-adjustment-verification" realizes real-time verification and iteration of optimization effect through module cooperation, ensuring that the combination precision and brightness consistency finally meet the standards, and the optimization stability is better than that of linear process.

[0023] Optionally, the unified data stream is transmitted using the TCP / IP protocol; the module triggering sequence is "the permission and authorization module 107 passes the verification → the parameter configuration module 102 is initialized → the camera calibration module 103 completes the calibration → the light machine combined image determination module 104 and the binocular consistency optimization module 106 work in parallel → the whole machine closed loop adjustment module 105 performs optimization", which ensures the order of the process.

[0024] Optionally, the quantitative visual effect index includes the combined image precision index (RMS, P95, Max deviation statistical quantity) and the brightness consistency index (binocular brightness min, max value, brightness difference value), and the index data format is unified as "value + unit", supporting real-time calling.

[0025] Optionally, the adaptive Micro-OLED light machine resolution range is 1920x1080-4K, and the waveguide sheet field of view angle range is 40°-120°; the parameter configuration module 102 supports adjusting the detection and calibration parameters according to the specific models of the light machine and the waveguide sheet.

[0026] Optionally, the data interaction between modules adopts the JSON format, including four core fields of data identification, acquisition timestamp, value and unit, to ensure data consistency and readability.

[0027] Optionally, the system automatically detects the hardware connection and function availability of each module when starting; the module that does not pass the self-check is prompted through the man-machine interface module 101, so as to avoid the influence of the fault on the optimization result.

[0028] In one optional embodiment, the camera calibration module 103 includes an image acquisition unit and a calibration calculation unit: the feature points are acquired through asymmetric circle point detection, the monocular remapping matrix and the stereo correction mapping are calculated, and the re-projection error is controlled within a preset range; the calibration calculation unit supports automatic verification of the calibration result, and reacquires the detection if the calibration result does not meet the standard.

[0029] In this embodiment, the feature points are acquired through asymmetric circle point detection, which can more clearly distinguish the feature matching relationship of the left and right eyes, reduce the calibration ambiguity, and improve the calculation accuracy of the monocular remapping matrix and the stereo correction mapping; the re-projection error is controlled within a preset range, which reduces the chain influence of the calibration error on the subsequent combined image determination and brightness calibration from the source, guarantees the accuracy benchmark of the system optimization, supports automatic verification of the calibration result, and reacquires the detection if the calibration result does not meet the standard, without the need for manual judgment of the calibration effectiveness, improves the automation degree, and avoids the invalid process caused by manual misjudgment.

[0030] Optionally, the preset range of the re-projection error is ≤0.5 pixels, which meets the basic requirement of pixel-level stereoscopic detection; the asymmetric circle point detection adopts a 10*14 array calibration board, the circle point diameter is 2-3 pixels, the center distance between adjacent circle points is 8-10 pixels, which ensures the stability of feature point extraction; the maximum automatic retry is 3 times when reacquiring, and if it still does not meet the standard, the sound and light alarm is sent through the man-machine interface module 101, which takes into account automation and abnormal processing; the calibration calculation unit adopts Zhang Zhengyou calibration algorithm, which improves the stability and accuracy of parameter calculation.

[0031] In yet another optional embodiment, the light machine stereoscopic image determination module 104 includes: a calibration parameter loading unit, an image correction unit, a circle point detection unit, a monocular pose detection unit, a double light machine stereoscopic image determination unit; the double light machine stereoscopic image determination unit compares the left and right eye circle point coordinates, calculates the RMS / P95 / Max deviation statistics, and outputs the pixel-level stereoscopic image precision quantization result.

[0032] In this embodiment, the units in the module have clear division of labor, forming a standardized process of “parameter loading-image correction-feature detection-pose calculation-stereoscopic image determination”, which ensures the standardization and repeatability of stereoscopic image precision determination; by comparing the left and right eye circle point coordinates, calculating the RMS, P95, Max multi-dimensional deviation statistics, compared with a single index, it can more comprehensively reflect the stereoscopic image quality and get rid of the dependence on subjective judgment; output the pixel-level stereoscopic image precision quantization result, provide accurate data support for the whole machine closed-loop adjustment, and ensure the strong pertinence of the adjustment operation.

[0033] Optionally, the threshold standard of stereoscopic image precision determination is: RMS≤1 pixel, P95≤1.5 pixel, Max≤2 pixel, which adapts to the actual use requirement of AR glasses; the center positioning error of the circle point detection unit is ≤0.1 pixel, which further improves the coordinate comparison accuracy; the monocular pose detection unit outputs the pitch angle and yaw angle in addition to the roll angle, and the detection accuracy is all ≤0.1°, which comprehensively covers the light machine pose deviation; the image correction unit adopts the bilinear interpolation algorithm, and the corrected image has no obvious distortion.

[0034] In yet another optional embodiment, the whole machine closed-loop adjustment module 105 includes: an adjustment vector calculation unit, a Bluetooth instruction sending unit, and an adjustment effect verification unit; the adjustment vector calculation unit uses the least square method to solve the optimal compensation parameters of translation, rotation, and scaling, and the units cooperatively realize the closed-loop process of “stereoscopic image determination→light machine parameter adjustment→secondary stereoscopic image determination”, until the stereoscopic image precision meets the standard.

[0035] In this embodiment, the three units of adjustment vector calculation, Bluetooth instruction sending and adjustment effect verification are integrated to realize the "determination-adjustment-verification" closed loop process, ensuring that the adjustment of the image combination accuracy is not missed or repeated, and improving the adjustment efficiency. The adjustment instruction is sent through Bluetooth, without physical contact with the optical machine, avoiding additional interference or damage to the structure of the AR glasses during the adjustment process. The secondary image combination determination link verifies the adjustment effect in real time, ensuring that the image combination accuracy finally meets the standard, and the reliability is significantly improved compared with one-time adjustment.

[0036] Optionally, the Bluetooth communication adopts the BLE protocol to ensure real-time transmission of the adjustment instruction. The least squares method is used to solve the optimal parameters for the adjustment vector calculation, considering the adjustment efficiency and accuracy. When the closed loop adjustment is not up to standard for three consecutive times, the system automatically records the deviation data and prompts manual intervention for inspection, avoiding infinite loop. The determination standard for the adjustment effect verification is consistent with the initial image combination determination, ensuring the uniformity of the optimization results.

[0037] The whole machine closed loop adjustment module adopts optimization algorithms such as the least squares method, taking "minimum image combination deviation" as the goal, and iteratively solving the optimal compensation parameters: Translation compensation amount : Take the mean or weighted mean of the translation deviation of all feature points to offset the overall horizontal and vertical deviation of the binocular image; Rotation compensation angle : Calculate the angle deviation of the left and right eye images through rigid transformation fitting of the feature point set to correct the image rotation caused by the installation tilt of the optical machine; Scaling compensation factor s: Calculate the distance ratio between the left and right eye feature points to match the size difference of the binocular images and ensure the consistency of the image size.

[0038] In another optional embodiment, the binocular consistency optimization module 106 is a supporting unit for the comprehensive optimization of binocular vision effect, including a full brightness detection unit, a brightness uniformity analysis unit, and a brightness calibration unit. The full brightness detection unit samples the ROI in a 4x12 grid for the corrected image. The brightness uniformity analysis unit calculates the min and max indicators of the binocular brightness difference. The brightness calibration unit performs calibration according to the difference.

[0039] In this embodiment, as a supporting unit for comprehensive optimization, it is specially designed for brightness consistency problems, filling the gap of existing technology in single optimization of image combination accuracy, and realizing dual optimization of "image combination + brightness"; The full brightness detection unit uses 4x12 grid ROI sampling to uniformly cover the effective display area of the image, which is more accurate in reflecting the brightness distribution difference of the binoculars compared to local sampling; The min and max indicators quantify the brightness difference, providing a clear basis for brightness calibration, which can reduce visual fatigue caused by uneven brightness after calibration.

[0040] Optionally, the allowed range of luminance difference is binocular luminance min / max ratio >= 0.8, ensuring that the human eye has no obvious perception; 4x12 grid sampling covers 100% of the effective display area, and the edge grid is at least 5 pixels away from the image edge to avoid edge distortion affecting sampling accuracy; the luminance calibration unit adjusts the luminance by adjusting the light machine current output, with a current adjustment accuracy of 1mA to ensure fine calibration; the luminance uniformity analysis unit uses a mean filter algorithm to process sampling data to reduce noise interference.

[0041] In yet another optional embodiment, the license and authorization module 107 includes an authorization key verification unit that implements authorization verification of the bound device by calling the keygen.sh script to provide permission support for system functions.

[0042] In this embodiment, the authorization key verification realizes the authorization of the bound device, effectively protects the system intellectual property rights, prevents unauthorized devices from being used illegally, and reduces the risk of technology leakage; the keygen.sh script is called to automatically verify the key without manual input, improving the convenience of use and avoiding authorization failure caused by input errors; permission support is provided for system functions to ensure that only authorized devices can fully call the functions of each module, ensuring the safe and stable operation of the system.

[0043] Optionally, the authorization key is generated using the AES-256 encryption algorithm, with a key length of 256 bits, providing high anti-cracking strength; the bound device identifier is a hardware unique serial number, supporting 1-5 devices to be bound, and adapting to different use scenarios; the key validity period is configurable, supporting 1 year, 3 years, and permanent modes.

[0044] In yet another optional embodiment, the human-machine interface module 101 provides a visual monitoring interface that synchronously displays the coincidence precision index, luminance difference data, and closed-loop optimization state.

[0045] In this embodiment, a visual monitoring interface is provided to synchronously present the coincidence precision index, luminance difference data, and closed-loop state, making it easy for users to intuitively understand the system operation and reducing the operation threshold; real-time display of optimization process data facilitates quick identification of the root cause of problems during system anomalies, improving troubleshooting efficiency; the data export function is supported, and optimization data can be stored in CSV format for subsequent quality traceability and data analysis.

[0046] Optionally, the interface data update frequency is 1 time per second to ensure data real-time; out-of-limit data is highlighted in red, and the alarm volume can be adjusted between 30-80 decibels; the interface supports Chinese-English switching to adapt to different use scenarios; historical optimization data trend charts can be displayed, and near 7-day data query and comparison is supported.

[0047] Embodiment Two Reference Figure 2This invention discloses a method for comprehensively optimizing the binocular visual effect of AR glasses, based on the system described in Embodiment 1, and includes the following core steps: Step 1: System initialization: Load camera parameters, stereo calibration parameters and remapping matrix.

[0048] Optionally, camera parameters include camera focal length, principal point coordinates, and distortion coefficients, used to correct imaging distortion of a single camera; stereo calibration parameters include rotation matrix and translation vector between binocular cameras, used to ensure the accuracy of coordinated matching of left and right eye images; the remapping matrix is ​​a 3×3 matrix pre-generated based on the camera calibration results, used for quick invocation of subsequent image correction; during loading, the system automatically verifies the integrity of parameters, and if any parameter is missing, initialization is paused and prompts for supplementation are sent through the human-machine interface module 101.

[0049] Step 2: Acquire the original images of the left and right eyes, and correct the images using a remapping matrix.

[0050] Optionally, the acquired raw image format is RAW, and the resolution is consistent with the output resolution of the "Micro-OLED optical engine" to ensure that the original image information is not lost. The core of image correction is to correct camera distortion and binocular parallax through the remapping matrix. The distortion error of the corrected image is ≤0.3 pixels, providing a unified benchmark for subsequent detection and optimization.

[0051] Step 3: Perform image merging accuracy determination: Obtain coordinates through asymmetric dot detection, calculate monocular pose and binocular image merging deviation statistics, and output the image merging accuracy result.

[0052] Optionally, asymmetric dot detection uses a 10×14 array calibration board with a feature point extraction rate ≥98%, ensuring the stability of coordinate acquisition; monocular pose calculation uses the PnP algorithm, outputting roll angle, pitch angle, and yaw angle, with detection accuracy ≤0.1°; the binocular merging deviation statistics are calculated in the following order: RMS (root mean square deviation, the square root of the average of the sum of squares of the binocular coordinate deviations of all matched feature points (such as asymmetric dots) → P95 (95th percentile deviation, the deviation values ​​of all matched feature points are sorted from smallest to largest, and the deviation value located at the 95th percentile is taken, i.e., 95% of the feature point deviations are less than or equal to this value, and only 5% of the feature point deviations are greater than this value) → Max (i.e., the maximum value among the deviation values ​​of all matched feature points)", with the result retained to two decimal places, and the judgment thresholds are RMS≤1 pixel, P95≤1.5 pixels, and Max≤2 pixels.

[0053] Where n is the number of matching feature points, ( , ) represents the coordinates of the feature point of the left eye. , ) is the corresponding feature point coordinate of the right eye.

[0054] Step 4: Perform closed-loop adjustment: if the image accuracy does not meet the standard, calculate the adjustment vector and send it to the optical machine through Bluetooth, and then repeat steps 2 to 3 to verify the effect until the standard is met.

[0055] Optionally, the adjustment vector is solved by least squares method, including X / Y direction translation, rotation angle, and scaling coefficient, with parameter accuracy of 0.1 pixel, 0.05°, and 0.1% respectively; Bluetooth communication uses BLE protocol to avoid device interference caused by physical contact; the upper limit of the number of repeated verifications is 5, and if the standard is still not met, record the data and prompt manual intervention to avoid infinite loop.

[0056] Step 5: Perform brightness consistency optimization: perform full brightness detection and grid sampling on the corrected image, calculate the binocular brightness difference, and if the difference exceeds the standard, perform brightness calibration, and then repeat sampling detection until the standard is met.

[0057] Optionally, the image brightness range of full brightness detection is controlled within 100-1000 nits, 4x12 grid is used for ROI sampling on the effective display area of the image, each sub-block size is ≥10x10 pixels, and there is no sampling blind area; the brightness difference is calculated by min, max values and min / max ratio of binocular brightness, and the determination threshold is min / max ratio ≥0.8; brightness calibration is achieved by adjusting the current output of the optical machine, with current adjustment accuracy of 1mA, and repeated sampling detection after calibration until the difference meets the standard.

[0058] The perception of brightness by the human eye is nonlinear, and the sensitivity to relative brightness difference is much higher than the absolute brightness value. When min / max ratio ≥0.8, the relative difference of binocular brightness is ≤20% (when min / max=0.8, relative difference=(max-min) / max=1-0.8=0.2, using the maximum brightness as the reference, measuring the fluctuation range of the full field of view brightness, which is more in line with the visual perception habit of the human eye taking the "brightest area" as the reference). Within this range, the visual system of the human eye will automatically "fuse" the brightness information of the left and right eyes, and cannot distinguish the obvious brightness difference, and will not produce the discomfort of "one eye bright and one eye dark", and can also avoid visual fatigue caused by uneven brightness - this is the core basis of the determination threshold.

[0059] In this embodiment of the invention, a core process of "initialization - image correction - closed-loop adjustment of image merging accuracy - closed-loop optimization of brightness consistency" is designed. This process transforms the system's comprehensive optimization concept of "image merging + brightness" into executable steps. Through two closed-loop iterations, the system ensures that both types of visual effects meet the standards, resulting in a more comprehensive and reliable optimization. Each step is logically progressive and without redundancy. Image correction provides a unified benchmark for subsequent detection and optimization. Optimization of image merging accuracy and brightness consistency is carried out in an orderly manner. Combined with the system's high-speed data processing capabilities, the optimization process can be completed quickly. Each step clearly defines the core operation content and follows a standardized process to ensure consistent optimization results for different devices and batches of products, avoiding result fluctuations caused by non-standard processes. Based on the system architecture design of Embodiment 1, the steps can be personalized according to the optical parameters of the AR glasses without complex manual intervention. Only parameter configuration needs to be initialized, and subsequent processes are executed automatically, resulting in a low threshold for implementation.

[0060] Taking the comprehensive optimization of binocular visual effects of a certain AR smart glasses as an example, this AR glasses adopts a "Micro-OLED optical engine + waveguide plate" architecture. The optical engine resolution is 2560×1440, and the waveguide plate field of view is 60°. During production line testing, it was found that there was slight ghosting in the image and inconsistent brightness between the left and right eyes. Optimization was required using the method of this invention. The optimization environment was a room temperature of 25℃, a relative humidity of 45%, and no strong light interference. The specific optimization process is as follows: Step 1: System initialization, loading camera parameters, stereo calibration parameters and remapping matrix. Specific parameters to load: a. Camera parameters: focal length 6.0mm, principal point coordinates (1280, 720), distortion coefficients k1=-0.012, k2=0.006, k3=-0.003; b. Stereo calibration parameters: Binocular rotation matrix The translation vector is (52.3, 0.4, 2.1) mm; c. Remapping matrix (3×3): ; Verification result: Parameters are complete, values ​​are within a reasonable range, initialization was successful.

[0061] Step 2: Acquire raw images for both eyes and correct the images using a remapping matrix. Image acquisition: A 5-megapixel global shutter camera was used to simultaneously acquire raw images of the left and right eyes (RAW format, resolution 2560×1440), with an acquisition time of 0.28 seconds; Image correction: a. Use the remapping matrix to correct distortion and disparity, and use bilinear interpolation to supplement pixels; b. Verification after correction: The distortion error of the left eye image is 0.2 pixels, the distortion error of the right eye image is 0.18 pixels, the size of the left and right eye images is consistent, and there is no obvious blur.

[0062] Step 3: Perform image alignment accuracy determination and output the image alignment accuracy result. Feature point detection: Using a 10×14 array asymmetric calibration board, 42 feature points were extracted (extraction rate 97.7%), with a center positioning error of 0.08 pixels; Monocular pose calculation: Left eye roll angle 0.06°, pitch angle -0.04°, yaw angle 0.03°; Right eye roll angle 0.12°, pitch angle -0.05°, yaw angle 0.02°; Calculation of deviation statistics: a. RMS (Root Mean Square Deviation): 1.2 pixels; b.P95 (95th percentile deviation): 1.7 pixels; c.Max (maximum deviation): 2.3 pixels; Judgment result: Not met (threshold RMS≤1 pixel, P95≤1.5 pixel, Max≤2 pixel), closed-loop adjustment is required.

[0063] Step 4: Perform closed-loop adjustments and verify the results. Adjustment vector solution: Calculated using the least squares method, the adjustment vector is: X translation 0.4 pixels, Y translation -0.3 pixels, rotation angle -0.07°, scaling factor 1.004; Command transmission: The adjustment vector is encapsulated into a Bluetooth command using the BLE protocol, with a transmission delay of 42ms and a response time of 0.8 seconds after the optical engine receives the command. Closed-loop verification: a. Repeat steps 2-3 to reacquire images and calculate the deviation statistics: RMS = 0.9 pixels, P95 = 1.3 pixels, Max = 1.8 pixels; b. Judgment result: Meets the standard, double image is eliminated.

[0064] Step 5: Perform brightness consistency optimization until the target is met. Full brightness detection and sampling: Control the optical engine to output a full brightness image (brightness 400 nits), sample 48 points with a 4×12 grid, and each sub-block is 10×10 pixels; Brightness difference calculation: a. Left eye sampling values: min=380 nits, max=420 nits; b. Right eye sampling values: min=350 nits, max=410 nits; c. Binocular brightness min / max ratio = 350 / 420≈0.83, which meets the standard (threshold ≥ 0.8). Optimization result: Brightness consistency meets requirements, no additional calibration is needed.

[0065] Final optimization results: 1. Image combining accuracy: RMS=0.9 pixels, P95=1.3 pixels, Max=1.8 pixels, meeting the threshold standard, and ghosting is completely eliminated; 2. Brightness consistency: The binocular brightness min / max ratio is 0.83, with no significant difference in visual brightness; 3. Total time: 28.7 seconds (including initialization, data collection, adjustment, and verification). 4. Actual wearing experience: Users experience no double vision or visual fatigue after wearing the product, and their binocular vision is clear and coordinated.

[0066] In an optional embodiment, the detection accuracy of the image merging accuracy in step 3 is at the pixel level.

[0067] In this embodiment, the image alignment accuracy detection reaches the pixel level, which can accurately capture tiny image alignment deviations. Compared with low-precision detection, it significantly improves the upper limit of AR glasses visual effect optimization. The single detection time is less than 3 seconds, which meets the high-speed detection requirements of automated production lines and reduces production line operating costs. It balances high precision and high efficiency, ensuring product quality without affecting the production cycle, thus achieving a balance between quality and efficiency.

[0068] Optionally, the pixel-level detection accuracy is specifically defined as an image alignment deviation detection error ≤ 0.5 pixels to ensure reliable detection results; the testing environment for detection time is room temperature 25℃±5℃, relative humidity 40%-60%, and no strong light interference to ensure data repeatability; the time consumption includes the entire process of image acquisition, correction, feature extraction, and deviation calculation, excluding data transmission delay; the time consumption of multiple batches of detection fluctuates ≤ 0.3 seconds to ensure detection stability.

[0069] In another optional embodiment, the adjustment performed by the whole-machine closed-loop adjustment module includes digital pre-distortion compensation of the image rendering engine. The adjustment vector calculation unit of the whole-machine closed-loop adjustment module solves the optimal compensation parameters for translation, rotation and scaling by the least squares method. The image rendering engine performs digital pre-distortion compensation based on the parameters and independently corrects nonlinear distortion by a third-order polynomial algorithm, thereby achieving the correction of four types of composite deviations.

[0070] The least squares method is the core algorithm of the closed-loop adjustment module, used to solve for three types of rigid body transformation parameters: translation compensation, rotation angle, and scaling factor. These parameters are solved sequentially in the order of "translation compensation → rotation angle → scaling factor." Nonlinear distortion correction is not within the scope of the least squares method and is independently implemented by the image rendering engine through third-order polynomial digital pre-distortion compensation. The two methods work together to complete the composite deviation correction of the "Micro-OLED optical engine + waveguide sheet" architecture. The least squares method is the core algorithm of the adjustment vector calculation unit in the whole-machine closed-loop adjustment module, specifically used to solve for the optimal compensation parameters under the objective of "minimizing image alignment deviation". Module collaborative logic: After the adjustment vector calculation unit obtains the parameters using the least squares method, the Bluetooth command sending unit transmits them to the optical engine, and then the adjustment effect verification unit completes the closed-loop verification to ensure that the image alignment accuracy meets the standards (RMS≤1 pixel, P95≤1.5 pixels, Max≤2 pixels). The solution follows a progressive logic of "first solving positional offset → then correcting angle tilt → finally matching size ratio", ensuring that each step of correction lays the foundation for subsequent optimization. Step 1: Solve for the translation compensation (Δx, Δy) Objective: To counteract the overall horizontal / vertical offset of the left and right eye images, and to initially align the feature points in spatial position.

[0071] Solution logic: Take the mean of the coordinate deviations of all matching feature points (asymmetric circles), which is essentially minimizing the "sum of squared deviations", as shown in the formula: ; Where n is the number of feature points, ( , ) is the coordinate of the left eye, ( ', (') represents the coordinates of the right eye).

[0072] Step 2: Solve for the rotation angle (θ) Objective: To correct image rotation deviations caused by the tilt of the optical engine installation, so that the left and right eye images are in the same pose.

[0073] Solution logic: Based on the rigid body rotation transformation formula (x'=xcosθ-ysinθ, y'=xsinθ+ycosθ), after substituting the coordinates of the feature points, the sum of squared deviations after rotation is minimized through nonlinear least squares iteration (Gauss-Newton method) to obtain the optimal θ (accuracy ≤0.05°).

[0074] Step 3: Solve for the scaling factor (s) Objective: To match the size ratio of the left and right eye images and avoid "inconsistent image size" caused by differences in optical engine output.

[0075] Solution logic: Minimize the sum of squared deviations between the scaled feature point distances and the target distance using the least squares method. The formula is: The accuracy is ≤0.1%.

[0076] The image rendering engine is an image processing unit integrated into AR glasses or optimization systems. It can receive deviation compensation parameters output by the closed-loop adjustment module of the whole machine in real time, accurately transform the pixel coordinates of the original image, and then transmit the processed image to the optical engine for projection onto the waveguide sheet.

[0077] The compensation operation is performed in the order of "basic coordinate transformation → nonlinear distortion correction", and the parameters are all calculated by the whole machine closed-loop adjustment module based on the image deviation statistics. Translational Deviation Correction: To compensate for the horizontal / vertical offsets in the left and right eye images, the rendering engine performs translational compensation on the original image pixel coordinates. The formula is as follows: , ( The translation compensation amount output by the closed-loop adjustment module of the whole machine (with an accuracy of ≤0.5 pixels) is used to align the left and right eye images in spatial position.

[0078] Rotation Deviation Correction: To address image rotation caused by optical engine mounting angle deviation, the rendering engine corrects pixel coordinates using a rotation transformation matrix. The formula is as follows: ( To compensate for rotation angle (accuracy ≤ 0.1°), correcting image tilt issues.

[0079] Scaling Deviation Correction: To address the issue of inconsistent magnification / reduction ratios in binocular images, the rendering engine adjusts pixel coordinates using a scaling factor, as shown in the formula. , ( To achieve a scaling compensation factor (accuracy ≤ 1%), the left and right eye image sizes are matched.

[0080] Nonlinear distortion correction: Specifically addresses the nonlinear distortions unique to the "Micro-OLED optomechanical + waveguide" architecture (such as pincushion / barrel distortion caused by waveguide transmission). These distortions do not fall under the category of rigid body transformations (translation / rotation / scaling) and cannot be solved using the least squares method. The implementation method is third-order polynomial digital pre-distortion compensation. A mapping relationship between "distorted pixel coordinates → ideal pixel coordinates" is established through reverse modeling, using the following formula: , ; Where, x, y: pixel coordinates of the distorted image (the original image coordinates that are distorted after being output by the optomechanical system and transmitted through the waveguide); x′, y′: ideal pixel coordinates of the corrected image (the target coordinates that meet the requirements of binocular image merging after distortion cancellation); k1, k2, k3: third-order distortion coefficients (calculated by the camera calibration module through asymmetric dot detection + Zhang Zhengyou calibration algorithm, with different coefficients corresponding to different optomechanical-waveguide combinations).

[0081] Execution process: First, the rigid body transformation corrections for translation, rotation, and scaling are completed using the least squares method; Substitute the corrected feature point coordinates (x, y) into the above third-order polynomial to calculate the ideal coordinates (x', y'). The ideal coordinates (x', y') after mapping may be floating-point numbers (non-integer pixel positions), while image display needs to be based on integer pixels, resulting in missing or misaligned pixel positions. To address this, a weighted average of the gray values ​​of the four adjacent integer pixels around the ideal coordinates (x', y') is calculated to fill the pixel value at the floating-point position, achieving pixel resampling (i.e., pixel resampling is completed through bilinear interpolation algorithm), and generating a corrected image without nonlinear distortion.

[0082] After each compensation process, the system will re-acquire the image and determine the image merging accuracy. If the accuracy is not met, the whole machine closed-loop adjustment module will iteratively optimize the compensation parameters, and the rendering engine will repeatedly perform the correction operation until the deviation meets the threshold requirement.

[0083] In this embodiment, digital pre-distortion compensation using an image rendering engine is employed. Compared to traditional mechanical adjustments, it can more flexibly and accurately correct four types of composite deviations, and is especially suitable for fine correction of minute deviations. It is specifically designed for four types of deviations: translation, rotation, scaling, and nonlinear distortion, covering the main deviation types of the "Micro-OLED optical engine + waveguide sheet" architecture, with a comprehensive correction range. Software-level compensation does not require modification of the hardware structure, reduces dependence on mechanical assembly precision, reduces hardware adjustment wear, and extends the lifespan of AR glasses.

[0084] In another optional embodiment, the camera parameters in step 1 are obtained through "asymmetric dot detection + single-target calibration + dual-target calibration", and the calibration process supports deleting invalid images and re-acquiring them.

[0085] In this embodiment, a combination of "asymmetric dot detection + single-eye calibration + dual-eye calibration" is used to obtain parameters. Compared with single calibration, it takes into account both monocular accuracy and binocular collaborative accuracy, thus improving parameter reliability. It supports the deletion of invalid images and the removal of low-quality calibration images such as blurry or occluded images to avoid affecting the accuracy of parameter calculation. The standardized parameter acquisition process ensures strong consistency of calibration results from different devices and batches, thereby improving the system's versatility.

[0086] Optionally, the calibration execution order is as follows: first perform single-eye calibration for both eyes, and then perform dual-eye calibration based on the single-eye calibration results to improve calibration efficiency; the criteria for invalid images are: reprojection error > 1 pixel or feature point extraction rate < 95%; the number of calibration images acquired is 10-20, the acquisition distance is 30-50cm, and the calibration board covers more than 80% of the camera's field of view; the calibration parameters are stored in XML format, which supports subsequent calling and modification.

[0087] Example 3 This invention discloses a method for comprehensively optimizing the binocular visual effect of AR glasses, based on the system described in Embodiment 1, and includes the following steps: Step 1: After the licensing and authorization module completes the authorization verification of the bound device, the parameter configuration module and the camera calibration module work together to complete the system initialization. The camera calibration module loads the camera parameters, stereo calibration parameters and remapping matrix, and the human-machine interface module displays the initialization status synchronously. Step 2: The camera calibration module acquires the original images of the left and right eyes, and the optomechanical image combining determination module corrects the images through the remapping matrix; Step 3: The optomechanical image merging determination module performs image merging accuracy determination: it obtains coordinates through asymmetric dot detection, calculates the monocular posture and binocular image merging deviation statistics, outputs the image merging accuracy result, and the human-machine interface module simultaneously displays the image merging accuracy result and whether it meets the standard. Step 4: If the image merging accuracy does not meet the standard, the whole machine closed-loop adjustment module calculates the adjustment vector and sends it to the optical engine via Bluetooth. Then, steps 2 to 3 are repeated to verify the effect. The human-machine interface module displays the adjustment status in real time. If the adjustment fails to meet the standard for 5 consecutive times, the human-machine interface module issues an abnormal prompt and records the deviation data until the image merging accuracy meets the standard. Step 5: The binocular consistency optimization module performs brightness consistency optimization: it performs full brightness detection and grid sampling on the calibrated image, calculates the binocular brightness difference, and if the difference exceeds the standard, it performs brightness calibration and repeats the sampling detection until the standard is met; after optimization, the human-machine interface module synchronously displays the brightness consistency result and the overall optimization completion status.

[0088] In an optional embodiment, the authorization verification of the licensing and authorization module in step 1 is achieved by calling the keygen.sh script to verify the authorization key of the bound device. If the verification fails, the optimization process is terminated, and the human-machine interface module displays an authorization failure message.

[0089] Optionally, the authorization key is generated using the AES-256 encryption algorithm, with a key length of 256 bits, providing strong anti-cracking capabilities; the identifier of the bound device is the device's unique hardware serial number, and one authorization key can be bound to a maximum of 5 devices; the authorization verification supports an automatic retry mechanism, automatically retrying twice if the first verification fails, and terminating the optimization process if it still fails; the authorization failure prompt displayed by the human-machine interface module includes specific reasons, such as key expiration, device serial number mismatch, or invalid key.

[0090] In another optional embodiment, the human-machine interface module synchronously displays the following content: image merging accuracy indicators (RMS, P95, Max values), brightness difference data (min, max values ​​and ratio of binocular brightness), closed-loop optimization status (initialization, detection, adjustment, completion and abnormality), and supports exporting the above data in CSV format.

[0091] Optionally, the update frequency of the image merging accuracy index, brightness difference data, and closed-loop optimization status is once per second to ensure data real-time performance. When the image merging accuracy index or brightness difference data exceeds the judgment threshold, the corresponding data item is highlighted in red, accompanied by an adjustable audible alarm of 30-80 decibels. The data exported in CSV format includes five core fields: data acquisition timestamp, data type, specific value, judgment result, and device serial number. The exported data can be stored to a specified path on the local hard drive or an external storage device, and the storage path can be manually configured through the human-machine interface module. The human-machine interface module supports storing historical optimization data for the past 7 days and supports comparison and display of historical data with current data for easy trend analysis.

[0092] Example 4 See Figure 1 This invention discloses a comprehensive optimization system for binocular visual effects in AR glasses, comprising: The human-machine interface module 101 provides an operating interface for the user; The parameter configuration module 102 provides an input interface to the user, including: 1. Image merging determination threshold settings, including thresholds for monocular determination rotation angle, pitch angle, centering tolerance, etc., and binocular determination root mean square error, P95, and maximum error thresholds; 2. Left and right camera serial numbers, used to determine the camera's mounting position on the fixture; Camera calibration module 103 includes camera image acquisition and camera calibration calculation: 1. After the user clicks the "Start Camera" button, the system begins to capture and display the left and right images; 2. After the user places the asymmetrical dot pattern, click the "Shoot" button. The system saves the current frame and prompts that the shot was taken successfully, while also informing the user of the remaining number of images required to complete the calibration. The captured images are displayed in sequence on the interface. 3. Users can delete any captured image and re-capture it; 4. After the number of shots reaches the set value, the user clicks the "Start Calibration" button, and the system performs binocular camera calibration. The calibration results will be saved and displayed on the interface. The operation process is as follows: Step 1: Initialize the configuration module: Set the parameters required for calibration, create and configure the dot detector, verify system parameters, and ensure the rationality of input parameters; Step 2: World Coordinate Generation Module: Map physical coordinates to generate 3D coordinate points corresponding to the real world; unify the coordinate system to ensure a one-to-one correspondence between detection points and physical points; verify the correctness of the coordinate generation logic; Step 3: Dot detection module: Extract circular grid feature points from images; provide multiple detection strategies to improve success rate; automatically correct the order of detection points; generate visualization results of the detection process; Step 4: Data Acquisition Module: Batch read and process calibration images from folders, verify the data to ensure the validity of the collected data, and provide statistical analysis of the processing results; The optomechanical merging determination module 104 includes calibration parameter loading, camera image acquisition, image correction, dot detection, monocular pose detection, dual-optical-mechanical merging determination, and calculation and adjustment suggestions, i.e., the optomechanical adjustment shown in the flowchart of this document: 1. When the user clicks the "Start Camera" button, the system checks whether there are valid camera calibration parameters (if not, it prompts the user to calibrate first), then opens the camera and displays the left and right images and the image matching detection indicators (rotation angle, offset distance, X / Y direction offset) in real time, and performs real-time image matching judgment (red for failure, green for success). 2. After clicking the "Offset Recognition" button, the system calculates the direction and angle that the optical engine needs to be adjusted based on the current image alignment deviation, and the results will be saved and displayed; See Figure 3 The operation process is as follows: Step 1: Initialize the system: Load camera parameters, stereo calibration parameters, and correction parameters; preload the remapping matrix. Step 2: Input left and right eye images: The left and right eye images are the original, uncorrected images; Step 3: Image Correction The left and right eye images are corrected using a remapping matrix; Step 4: Dot detection: A connected component dot detector is used to detect a dot grid on the corrected image, and the coordinates of the dots detected in the left and right eye images are output. Step 5: Monocular pose detection: The pose (mainly roll angle) is calculated using the camera parameters of the left and right eyes and the coordinates of the detected dots. The pose results of the left and right eyes are output. Based on the coordinates of the dots detected by the left eye, the center of the dot set is calculated and compared with the expected center position. The centering result of the left eye is then output. Step 6: Dual-optical-machine image combination determination: Compare the coordinates of the dots detected by the left and right eyes, calculate the statistics of the relative deviation (RMS, P95, Max), and output the image combination result; Step 7: Calculate the operation adjustment vector: Based on the pose, centering, and merging results, calculate mechanical adjustment suggestions and output adjustment vectors and adjustment priorities; The closed-loop adjustment module 105 includes calibration parameter loading, camera image acquisition, image correction, dot detection, image coincidence detection (i.e., dual-optical-machine image coincidence determination), and calculation operations, and adjusts via Bluetooth (i.e., adjusts by sending Bluetooth commands): See Figure 4 The operation process is as follows: Step 1: Initialize the system: Load camera parameters, stereo calibration parameters, and correction parameters; preload the remapping matrix. Step 2: Input left and right eye images: The left and right eye images are the original, uncorrected images; Step 3: Image Correction The left and right eye images are corrected using a remapping matrix; Step 4: Dot detection: A connected component dot detector is used to detect a dot grid on the corrected image, and the coordinates of the dots detected in the left and right eye images are output. Step 5: Dual-optical imaging determination: Compare the coordinates of the dots detected by the left and right eyes, calculate the statistics of the relative deviation (RMS, P95, Max), and output the image combination result; Step 6: Calculate the operation adjustment vector: Based on the pose, centering, and merging results, calculate mechanical adjustment suggestions and output adjustment vectors and adjustment priorities; The binocular consistency optimization module 106 includes calibration parameter loading, camera image acquisition, image correction, and full brightness detection. See Figure 5 The operation process is as follows: Step 1: Initialize the system: Load camera parameters, stereo calibration parameters, and correction parameters; preload the remapping matrix. Step 2: Input left and right eye images: The left and right eye images are the original, uncorrected images; Step 3: Image Correction The left and right eye images are corrected using a remapping matrix; Step 4: Full Brightness Detection: For each detected rectangle, crop it to the image area to obtain a rectangular region of interest (ROI); divide the ROI into a 4×12 grid, with the last column / row possibly narrower to ensure it does not exceed the boundary; traverse each grid sub-block and take the pixel mean of that sub-rectangle as a "sample value"; summarize the mean values ​​of all sub-blocks into a list and count them as the number of samples; calculate the average / minimum / maximum value of the sample value list and give the uniformity min / max.

[0093] Licensing and Authorization Module 107: 1. The system provides an authorization interface. Users need to import the authorization key file, which is generated from the machine code in the user authorization interface. 2. The keygen.sh script is automatically invoked to verify the authorization key when the system starts up; 3. After successful verification, the user will be taken to the main interface. If verification fails, the user will be prompted with an error message. 4. The authorization mechanism is designed to support device binding.

[0094] Example 5 This invention discloses a computer-readable storage medium storing a computer program. When the program is executed by a processor, it implements the steps of the AR glasses binocular visual effect comprehensive optimization method described in Embodiment 2 or Embodiment 3.

[0095] The content disclosed in the embodiments of this invention is only a preferred embodiment of the invention and is used only to illustrate the technical solutions of the invention, not to limit it. Although the invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this invention.

Claims

1. A comprehensive optimization system for binocular visual effects in AR glasses, characterized in that, It includes a human-machine interface module, a parameter configuration module, a camera calibration module, an optical-mechanical image matching determination module, a whole-machine closed-loop adjustment module, a binocular consistency optimization module, and a licensing and authorization module. All modules work together through a unified data flow to achieve image matching accuracy determination, including translation, rotation, scaling and nonlinear distortion, as well as closed-loop optimization of brightness consistency calibration for the "Micro-OLED optical engine + waveguide" optical architecture, and output quantitative visual effect indicators.

2. The AR glasses binocular visual effect comprehensive optimization system according to claim 1, characterized in that, The camera calibration module includes an image acquisition unit and a calibration calculation unit: it acquires feature points through asymmetric dot detection, calculates the monocular remapping matrix and stereo correction mapping, and controls the reprojection error within a preset range; the calibration calculation unit supports automatic verification of calibration results, and re-acquires and detects if the results do not meet the standards.

3. The AR glasses binocular visual effect comprehensive optimization system according to claim 1, characterized in that, The optical-mechanical image merging determination module includes: a calibration parameter loading unit, an image correction unit, a dot detection unit, a monocular posture detection unit, and a dual-optical-mechanical image merging determination unit; the dual-optical-mechanical image merging determination unit calculates the deviation statistics by comparing the dot coordinates of the left and right eyes, and outputs the pixel-level image merging accuracy quantification result.

4. The AR glasses binocular visual effect comprehensive optimization system according to claim 1, characterized in that, The whole machine closed-loop adjustment module includes: an adjustment vector calculation unit, a Bluetooth command sending unit, and an adjustment effect verification unit. The adjustment vector calculation unit uses the least squares method to solve for the optimal compensation parameters of translation, rotation, and scaling. Each unit works together to realize the closed-loop process of "image merging determination - optical-mechanical parameter adjustment - secondary image merging determination" until the image merging accuracy meets the standard.

5. The AR glasses binocular visual effect comprehensive optimization system according to claim 1, characterized in that, The binocular consistency optimization module is a supporting unit for the comprehensive optimization of binocular visual effects, including: a full brightness detection unit, a brightness uniformity analysis unit, and a brightness calibration unit; the full brightness detection unit performs ROI sampling of the corrected image in a 4×12 grid, the brightness uniformity analysis unit calculates the min and max indices of binocular brightness difference, and the brightness calibration unit performs calibration operations based on the difference.

6. A method for comprehensive optimization of binocular visual effects in AR glasses, characterized in that, The system implementation based on claim 1 includes the following steps: Step 1: The parameter configuration module and the camera calibration module work together to complete the system initialization. The camera calibration module loads the camera parameters, stereo calibration parameters and remapping matrix. Step 2: The camera calibration module acquires the original images of the left and right eyes, and the optomechanical image combining determination module corrects the images through the remapping matrix; Step 3: The optomechanical image merging determination module performs image merging accuracy determination: it obtains coordinates through asymmetric dot detection, calculates the monocular pose and binocular image merging deviation statistics, and outputs the image merging accuracy result; Step 4: If the image combining accuracy does not meet the standard, the whole machine closed-loop adjustment module calculates the adjustment vector and sends it to the optical engine via Bluetooth. Then, repeat steps 2 to 3 to verify the effect until the standard is met. Step 5: The binocular consistency optimization module performs brightness consistency optimization: it performs full brightness detection and grid sampling on the calibrated image, calculates the binocular brightness difference, and if the difference exceeds the standard, it performs brightness calibration and repeats the sampling detection until the standard is met.

7. The method for comprehensive optimization of binocular visual effects of AR glasses according to claim 6, characterized in that, The detection accuracy of the image merging accuracy mentioned in step 3 is at the pixel level.

8. The method for comprehensive optimization of binocular visual effects of AR glasses according to claim 6, characterized in that, The adjustments performed by the whole-machine closed-loop adjustment module in step 4 include digital pre-distortion compensation of the image rendering engine. The adjustment vector calculation unit of the whole-machine closed-loop adjustment module solves the optimal compensation parameters for translation, rotation and scaling by the least squares method. The image rendering engine performs digital pre-distortion compensation based on these parameters and independently corrects nonlinear distortion by a third-order polynomial algorithm, thereby achieving the correction of four types of composite deviations.

9. The method for comprehensive optimization of binocular visual effects of AR glasses according to claim 6, characterized in that, The camera parameters in step 1 are obtained through "asymmetric dot detection + single-target calibration + dual-target calibration". The calibration process supports deleting invalid images and re-acquiring them.

10. A computer-readable storage medium, characterized in that, The device contains a computer program that, when executed by a processor, implements the steps of the AR glasses binocular visual effect comprehensive optimization method as described in claims 6 to 9.

Images