A data correlation analysis method for AR glasses overall performance test

Abstract

The present application belongs to the technical field of optical equipment detection, and particularly relates to a data correlation analysis method for AR glasses whole machine performance test, comprising: S1, extracting geometric position features and luminosity response features from left and right eye images; S2, calculating luminosity competition load index by using a logarithmic form of Weber perception model; S3, constructing anisotropic parallax index by using an anisotropic elliptical field; S4, calculating visual fusion comfort score by using an S-shaped scoring function based on the anisotropic parallax index, and grading the whole machine performance of the AR glasses according to the scoring result. The present application establishes a deep coupling mechanism of multi-dimensional features, establishes a dynamic modulation relationship of luminosity data to geometric threshold value through an index attenuation coupling mechanism, solves the problem that the user is still dizzy due to binocular brightness or clarity unevenness although the geometric positions have been aligned, and ensures that each AR glasses leaving the factory can provide a comfortable visual fusion experience.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of optical equipment detection, and particularly relates to a data correlation analysis method for AR glasses whole machine performance test. BACKGROUND

[0002] The augmented reality glasses (AR glasses) project virtual images to the left eye and the right eye of a user through two independent optical display modules respectively, and the virtual images are fused in the cerebral cortex to form a stereoscopic virtual image. The consistency of binocular images is a key indicator determining the comfort of the user. If the spatial positions of the left eye image and the right eye image are greatly deviated, or the differences in the perception dimensions such as brightness and clarity are significant, the brain of the user cannot complete image fusion, and thus causes serious dizziness, double vision and even nausea and other adverse physiological reactions.

[0003] The current binocular image detection technology has some limitations in actual application: the existing technology usually directly calculates the linear difference of the brightness of the left eye and the right eye and sets a fixed threshold as the qualified standard, ignores the nonlinear characteristics that the human eye visual system follows the Weber-Fechner law in the perception of brightness, and causes strong discomfort caused by a small brightness difference in a low brightness scene, but the difference is easily ignored in detection. At the same time, the existing technology usually uses an isotropic circular region as the judgment boundary of the geometric position, and does not consider the physiological structure characteristics that the human eye is extremely sensitive to vertical parallax and has strong adjustment ability to horizontal parallax, which easily causes the missed detection of vertical deviation.

[0004] In addition, the existing detection process usually processes the luminosity consistency and the geometric position consistency as two independent processes in parallel, and lacks a multi-dimensional coupling judgment mechanism. In fact, visual perception is an overall process, when there is a luminosity competition between the left eye and the right eye, the brain will significantly reduce the fusion ability of geometric parallax in order to suppress the interference of low-quality images, that is, the worse the luminosity, the more accurate the geometric alignment must be. The existing technology lacks a mechanism for dynamically adjusting the geometric tolerance according to the luminosity difference, and it is difficult to accurately evaluate the real visual experience of the user, which causes the user to be uncomfortable with the shipped products. SUMMARY

[0005] The application provides a data correlation analysis method for AR glasses whole machine performance test, to solve the technical problems of ignoring the nonlinear perception characteristics of the human eye, lacking anisotropic parallax judgment, and splitting the luminosity and geometry judgment in the prior art.

[0006] To solve the above problems, the application adopts the following technical solutions:

[0007] A data correlation analysis method for AR glasses whole machine performance test, comprising the following steps:

[0008] S1, controlling the binocular camera to capture left and right eye images displayed by the AR glasses, and extracting geometric position features and luminosity response features from the left and right eye images;

[0009] S2, based on the luminosity response features, calculating a luminosity competition load index using a logarithmic form of a Weber perception model, the luminosity competition load index being used to evaluate the physiological burden of the human eye when processing luminosity differences between the left and right eyes;

[0010] S3, according to the geometric position features and the luminosity competition load index, constructing an anisotropic disparity index using an anisotropic elliptical field, wherein a tolerance threshold of geometric determination exponentially decays with an increase in the luminosity competition load index;

[0011] S4, based on the anisotropic disparity index, calculating a visual fusion comfort score using an S-shaped scoring function, and grading the overall performance of the AR glasses according to the scoring result.

[0012] Further, the extraction of the geometric position features from the left and right eye images includes:

[0013] The left and right eye images are preprocessed, the pixel coordinates of the central target are extracted using a sub-pixel level barycenter algorithm, and are converted into angle coordinates as the geometric position features.

[0014] Further, the extraction of the luminosity response features from the left and right eye images includes:

[0015] The average luminance value is calculated in the effective display area of the left and right eye images, the modulation transfer function is calculated using the tilted edge method, and the MTF value thereof at a preset spatial frequency is taken as the sharpness index, and the average luminance value and the extracted MTF value are taken as the luminosity response features.

[0016] Further, the spatial frequency is 20 cycles / degree, and the MTF50 value corresponding to the modulation transfer function at 20 cycles / degree is taken as the sharpness index.

[0017] Further, the luminosity competition load index satisfies the following relationship:

[0018] ;

[0019] In the formula, is the luminosity competition load index, is a natural logarithmic function, and are the average luminance values of the left and right eyes, respectively, is the arithmetic mean of the left and right eye average luminance values, is a dark vision compensation constant, a clarity weight coefficient, and respectively clarity indicators of left and right eyes, is an arithmetic mean of the left and right eye clarity indicator values, is a blur compensation constant.

[0020] Further, the anisotropic disparity index satisfies the following relationship:

[0021] ;

[0022] wherein, is an anisotropic disparity index, and respectively coordinates of the center target of the left eye image and the right eye image in the horizontal direction, and respectively coordinates of the center target of the left eye image and the right eye image in the vertical direction, is a horizontal reference tolerance threshold, is a vertical reference tolerance threshold, is a natural exponential function, and respectively coupling sensitivity coefficients in the horizontal direction and the vertical direction, is a luminosity competition load index.

[0023] Further, the vertical reference tolerance threshold is less than the horizontal reference tolerance threshold.

[0024] Further, the coupling sensitivity coefficient in the vertical direction is greater than the coupling sensitivity coefficient in the horizontal direction.

[0025] Further, the visual fusion comfort score satisfies the following relationship:

[0026] ;

[0027] wherein, is a visual fusion comfort score, is an anisotropic disparity index, is a decision steepness factor.

[0028] Further, the binocular camera is composed of left and right human eye simulation camera modules of a test device.

[0029] The beneficial effects are:

[0030] 1. The application establishes a deep coupling mechanism of multi-dimensional features, and establishes a dynamic modulation relationship of photometric data on geometric threshold value through an exponential decay coupling mechanism. This mechanism solves the problem that the user is still dizzy due to uneven binocular brightness or clarity although the geometric position has been aligned, and ensures that each AR glasses shipped can provide a comfortable visual fusion experience.

[0031] 2. Through the sub-pixel level algorithm and the inclined edge method, the geometric and photometric data can be accurately extracted, and the image noise interference is eliminated, thereby providing high-quality input source for subsequent calculation.

[0032] 3. By introducing the logarithmic Weber model, the application simulates the nonlinear perception characteristics of the human eye to brightness in industrial automation detection, avoids the over-sensitive misjudgment of traditional linear algorithms to small brightness differences in high brightness environment, and captures the real sensitivity differences of the human eye in low brightness environment.

[0033] 4. By constructing a dynamic elliptical judgment field, the physiological law that the human eye has high horizontal tolerance and low vertical tolerance is strictly followed. Compared with the rough circular judgment region of the prior art, the risk of diplopia missed detection caused by small deviation in the vertical direction is avoided, and the good product misjudgment caused by excessive strictness in the horizontal direction is also avoided.

[0034] 5. By converting the binary judgment result into a continuous numerical score, compared with the traditional hard threshold, the products in the critical risk area can be more sensitively identified. BRIEF DESCRIPTION OF DRAWINGS

[0035] Figure 1 The flowchart of the data correlation analysis method for AR glasses whole machine performance test;

[0036] Figure 2 The anisotropic dynamic judgment field diagram of binocular images is shown.

[0037] Figure 3 The prior art and the detection efficiency comparison chart of the application. DETAILED DESCRIPTION

[0038] The embodiment of the data correlation analysis method for AR glasses whole machine performance test provided by the application:

[0039] As shown in Figure 1 A data correlation analysis method for AR glasses whole machine performance test comprises the following steps:

[0040] S1, control the binocular camera to collect the left and right eye images displayed by the AR glasses, and extract the geometric position features and photometric response features from the left and right eye images.

[0041] Specifically, the binocular camera here is composed of left and right human-eye-simulating camera modules of the test device. In specific operation, first, the left and right human-eye-simulating camera modules of the test device are controlled to move along the linear motor guide rail, the camera interpupillary distance is adjusted to the standard pupillary distance, for example, the standard pupillary distance is 64 mm, and it is ensured that the camera optical axis is aligned with the design visual axis of the AR glasses. Subsequently, the AR glasses are driven to display a dedicated comprehensive test card, the comprehensive test card contains a central cross target, a chessboard distributed around, and color blocks of different gray scales, wherein the cross target is used for geometric positioning, and the color blocks of different gray scales are used for photometric measurement. Then, the binocular camera is triggered to synchronously expose, and left eye images and right eye images are respectively acquired.

[0042] When the geometric position feature is extracted, the left eye image and the right eye image are first respectively subjected to grayscale and Gaussian filter pretreatment, so as to remove the noise points in the images. Then, the pixel coordinates of the central cross target of the left eye image and the pixel coordinates of the central cross target of the right eye image are extracted by using a sub-pixel level barycenter algorithm. Finally, the obtained pixel coordinates are converted into angle coordinates according to the focal length and pixel size parameters of the camera, so as to facilitate the standardized calculation conforming to the human eye visual field angle.

[0043] When the photometric response feature is extracted, the average luminance value in the effective display area of the left eye image and the right eye image is respectively calculated, wherein the unit of the average luminance value is nit; for example, the effective display area is the central field of view range. At the same time, the modulation transfer function (MTF) of the left eye and the right eye is respectively calculated by using the black and white edge area in the image by using the inclined edge method, the MTF value at a preset spatial frequency is taken as the definition index, for example, the MTF50 value at a spatial frequency of 20 cycle / degree is taken as the definition index, and the average luminance value and the definition index constitute the photometric response feature.

[0044] The step S1 can extract the basic feature data reflecting the optical performance of the AR glasses, eliminate the interference of environmental noise and system error, provide accurate and reliable input basis for subsequent multi-dimensional data correlation analysis, and ensure the authenticity of the final evaluation result.

[0045] S2, based on the photometric response feature, a photometric competition load index is calculated by using a logarithmic form Weber perception model, the photometric competition load index is used to evaluate the physiological burden of the human eye when processing the luminosity difference between the left eye and the right eye.

[0046] Specifically, the photometric competition load index satisfies the following relationship:

[0047]

[0048] In the formula, The photometric competition load index is a dimensionless value. , , , are the average luminance values of the left eye and the right eye respectively, , are the sharpness indicators of the left eye and the right eye respectively. is a dark vision compensation constant, which may be 0.1 nit for example, used to simulate the dark noise level of the retina in a completely dark environment and prevent the denominator from being zero. is a blur compensation constant, which may be 0.01 for example. is a sharpness weight coefficient, which gives a higher weight to the sharpness difference term, which may be 1.5 for example.

[0049] In one example, assuming that the average luminance value of the left eye is nit and the average luminance value of the right eye is nit, then nit. The sharpness indicator of the left eye is , and the sharpness indicator of the right eye is , then .

[0050] First, calculate the luminance difference term: , and then calculate the natural logarithm of this term: .

[0051] Next, calculate the sharpness difference term: . Square it to get , and then multiply it by the weight coefficient sharpness weight coefficient to calculate .

[0052] Finally, calculate the total load as .

[0053] By using the logarithmic form of the Weber perception model, the physiological stress of the human eye when facing binocular images with inconsistent luminosity can be evaluated, solving the problem of poor adaptability of traditional linear judgment under different brightness backgrounds, making the evaluation result more in line with the real visual experience of the human eye.

[0054] S3, according to the geometric position feature and the luminosity competition load index, an anisotropic disparity index is constructed by using an anisotropic elliptical field, wherein the tolerance threshold of geometric judgment exponentially decays with the increase of the luminosity competition load index.

[0055] Specifically, the anisotropic disparity index satisfies the following relationship:

[0056]

[0057] In the formula, is the anisotropic disparity index, and are the coordinates of the center target in the horizontal direction of the left-eye image and the right-eye image, respectively, and are the coordinates of the center target in the vertical direction of the left-eye image and the right-eye image, respectively, is a horizontal reference tolerance threshold, and an example of the horizontal reference tolerance threshold is 1.0°; is a vertical reference tolerance threshold, and an example of the vertical reference tolerance threshold is 0.3°; is a natural exponential function; and are coupling sensitivity coefficients in the horizontal direction and the vertical direction, respectively, and examples of the coupling sensitivity coefficients are = 0.5, = 0.8.

[0058] Taking the example of S1, the luminosity competition load index is known to be , it is assumed that the measured horizontal parallax is , and the vertical parallax is .

[0059] First, the horizontal dynamic threshold is calculated as follows:

[0060] wherein is the horizontal dynamic threshold.

[0061] Next, the vertical dynamic threshold is calculated as follows: wherein is the vertical dynamic threshold.

[0062] The anisotropic parallax index is calculated by substituting the relationship of the anisotropic parallax index as follows:

[0063] The square of the horizontal component is .

[0064] The square of the vertical component is .

[0065] .

[0066] Since , it is still determined to be qualified after considering the luminosity load tightening threshold. It can be seen that the existence of the luminosity load compresses the original tolerance threshold, for example, the horizontal reference tolerance threshold 1.0 is compressed to 0.806, and the vertical reference tolerance threshold 0.3 is compressed to 0.212, achieving the dynamic tightening of the standard; if the luminosity difference is larger, the threshold will be further reduced.

[0067] By introducing anisotropic disparity index and exponential decay coupling mechanism, not only the physiological characteristics of human eyes that are more sensitive to vertical disparity are reflected, preventing missed detection in the vertical direction, but also adaptive judgment logic that the poorer the photometric quality, the higher the geometric alignment requirement is realized, thus intercepting the hidden hazard products that seem to meet the geometric position but cause dizziness due to photometric differences.

[0068] S4, based on the anisotropic disparity index, a visual fusion comfort score is calculated using an S-shaped scoring function, and the overall performance of the AR glasses is graded according to the scoring results.

[0069] Specifically, in this embodiment, in order to provide more fine quality management means for the production line, the binary decision result is converted into a continuous numerical score. Among them, the visual fusion comfort score satisfies the following relationship: ;

[0070] In the formula, is the decision steepness factor, which is exemplarily taken as 5.0, used to simulate the critical mutation characteristics of human eyes from comfort to dizziness.

[0071] In the example of S3, it is known that .

[0072] First, calculate the exponential term inside: .

[0073] Then calculate the denominator: .

[0074] Finally, calculate the score: .

[0075] The score of 87.7 is close to 88 points, indicating good visual comfort, which can be classified as good products, such as being divided into A level.

[0076] In other embodiments, if the anisotropic disparity index is slightly greater than 1, such as 1.1, , the exponential term is , the denominator is 3.85, and the score drops to about 26 points, reflecting severe punishment for products that exceed the standard.

[0077] In an optional embodiment, referring to Figure 2 , Figure 2 shows a schematic diagram of the anisotropic dynamic judgment field of binocular images. The horizontal axis represents horizontal geometric disparity, and the vertical axis represents vertical geometric disparity. The gray dashed circle represents the fixed threshold judgment boundary of the prior art, which is circular and does not distinguish between horizontal and vertical differences. The colored solid ellipse represents the dynamic judgment boundary of the present application, which is a flat ellipse, intuitively showing the anisotropic feature that the tolerance in the vertical direction is much smaller than that in the horizontal direction. Figure 2The image shows three ellipses of different colors (green, orange, and red), corresponding to low, medium, and high photometric competition loads, respectively. As the photometric load increases, the area of ​​the ellipses shrinks significantly, intuitively revealing the logic that the worse the photometric consistency, the tighter the geometric alignment standard automatically becomes.

[0078] In one embodiment, refer to Figure 3 , Figure 3 A comparison chart of the detection performance of existing technologies and this invention is presented. The horizontal axis represents the photometric competition load index, and the vertical axis represents the comprehensive amplitude of geometric parallax. Green dots represent high-quality products, located below the judgment curve of this invention. Gray dots represent obviously defective products, located above the judgment line of the existing technology. Red crosses represent latent defects; these dots are located below the gray dashed line but above the blue solid line of this invention. This means that although these products may appear to meet the geometric error standards, the excessive photometric differences can actually cause dizziness in users.

[0079] The soft edge detection mechanism, built using the S-shaped scoring function, transforms complex detection data into an intuitive percentage score. This not only sensitively identifies products in critical risk zones but also provides quantifiable quality grading criteria for the production line, helping manufacturers optimize yield control strategies and improve overall shipment quality.

[0080] In addition, in the description of this specification, "multiple" means at least two, such as two, three or more, etc., unless otherwise expressly and specifically defined.

Claims

1. A data correlation analysis method for performance testing of AR glasses, characterized in that, Includes the following steps: S1, control the binocular camera to acquire left and right eye images displayed on the AR glasses, and extract geometric position features and photometric response features from the left and right eye images; S2, Based on the photometric response characteristics, the photometric competition load index is calculated using a logarithmic form of the Weber perception model. The photometric competition load index is used to assess the physiological burden on the human eye when processing the difference in photometric intensity between the left and right eyes. S3, Based on the geometric location features and the photometric competition load index, an anisotropic parallax index is constructed using an anisotropic elliptic field, wherein the tolerance threshold for geometric determination decreases exponentially with the increase of the photometric competition load index. S4. Based on the anisotropic parallax index, the visual fusion comfort score is calculated using the S-shaped scoring function, and the overall performance of the AR glasses is graded according to the score results. The photometric competition load index satisfies the following relationship: In the formula, The photometric competition load index. It is the natural logarithm function. and These are the average brightness values ​​for the left and right eyes, respectively. It is the arithmetic mean of the average brightness values ​​of the left and right eyes. This is the dark vision compensation constant. For clarity weighting coefficients, and These are the acuity indicators for the left and right eyes, respectively. The arithmetic mean of the acuity index values ​​for both eyes. For fuzzy compensation constants; The anisotropic parallax index satisfies the following relationship: In the formula, The anisotropic parallax index, and These are the horizontal coordinates of the center target in the left and right eye images, respectively. and These are the vertical coordinates of the center target in the left and right eye images, respectively. This is the horizontal baseline tolerance threshold. The vertical reference tolerance threshold, It is a natural exponential function. and These are the coupling sensitivity coefficients in the horizontal and vertical directions, respectively; Visual fusion comfort scores satisfy the following relationship: In the formula, Score the visual integration comfort level. This is the steepness factor of the judgment.

2. The data correlation analysis method for AR glasses overall performance testing according to claim 1, characterized in that, Extracting geometric location features from the left and right eye images includes: The left and right eye images are preprocessed, and the pixel coordinates of the central target are extracted using a subpixel-level centroid algorithm and converted into angular coordinates as geometric position features.

3. The data correlation analysis method for AR glasses overall performance testing according to claim 2, characterized in that, Extracting photometric response features from the left and right eye images includes: The average brightness value is calculated within the effective display area of ​​the left and right eye images. The modulation transfer function is calculated using the tilted edge method, and its MTF value at a preset spatial frequency is taken as the sharpness index. The average brightness value and the extracted MTF value are used as the photometric response feature.

4. The data correlation analysis method for AR glasses overall performance testing according to claim 3, characterized in that, The preset spatial frequency is 20 cycles / degree, and the MTF50 value corresponding to the modulation transfer function at 20 cycles / degree is used as the sharpness indicator.

5. The data correlation analysis method for AR glasses overall performance testing according to claim 3, characterized in that, The vertical reference tolerance threshold is less than the horizontal reference tolerance threshold.

6. The data correlation analysis method for AR glasses overall performance testing according to claim 5, characterized in that, The coupling sensitivity coefficient in the vertical direction is greater than that in the horizontal direction.

7. A data correlation analysis method for performance testing of AR glasses according to any one of claims 1-6, characterized in that, The binocular camera consists of left and right human-eye-like camera modules of the testing equipment.

Images