Three-dimensional image generation method based on optimization of parameters of a spherical reflection cavity

Abstract

The application belongs to the field of image generation, and particularly relates to a three-dimensional image generation method based on parameter optimization of a spherical reflection cavity, aiming at solving the problem of low imaging accuracy and poor adaptability caused by the lack of a cooperative optimization mechanism in the prior art. The application comprises: obtaining a reflection path offset contrast table of the optical properties of the inner wall of the cavity, measuring the deformation deviation of the cavity to construct a geometric compensation mapping relationship, and converting environmental disturbance into a noise intensity coefficient in real time; inputting a target three-dimensional image into a pre-compensation processor, synchronously performing optical path compensation, geometric distortion correction and environmental noise suppression, and outputting a pre-corrected image to a projection device; capturing the actual projection through an image sensor, extracting difference data from the pre-corrected image, and decomposing the difference data into a system error component and an environmental interference component; and feeding back the updated parameters to the processor to regenerate a final three-dimensional image. The application realizes cooperative correction and dynamic optimization of multiple error sources, and significantly improves the imaging accuracy and adaptability.

Description

Technical Field

[0001] The present invention belongs to the field of image generation, and in particular relates to a three-dimensional image generation method based on spherical reflection cavity parameter optimization. Background Art

[0002] The spherical reflective cavity, the core component of the virtual imaging system, is manufactured using an integrated molding process using a carbon fiber prepreg and PMI foam sandwich structure. The main cavity and the two earpiece enclosures are assembled with high precision to form a complete reflective surface, while aluminum alloy upper and lower side rails and earpiece enclosure side rails ensure optical surface accuracy.

[0003] In the field of three-dimensional imaging and display, utilizing spherical reflective cavity structures for image generation is a promising technology approach. These cavities, through the reflective properties of their inner walls, guide and reconstruct projected light, creating a surround-view stereoscopic visual effect. This approach offers unique advantages in immersive experiences, virtual reality, and advanced visualization. However, the practical application of this technology is significantly limited by multiple complex factors, which restrict its imaging accuracy and stability.

[0004] The primary challenge stems from the optical and geometric properties of the spherical reflective cavity itself. The optical properties of the inner wall of the cavity (such as reflectivity and scattering properties) are not ideally uniform, resulting in nonlinear path deviations for light rays at different incident angles during reflection. This deviation directly distorts the final image. At the same time, slight shape deviations are inevitable in the cavity during manufacturing, installation, or long-term use, that is, there is a difference between the actual geometric shape and the design benchmark. This geometric deformation introduces additional image distortion, and this distortion pattern changes dynamically with the position of the cavity. Existing methods usually use static geometric correction models, which are difficult to effectively adapt to this dynamic distortion caused by the combined action and mutual coupling of optical path deviation and cavity geometric deformation.

[0005] Secondly, environmental disturbances within the cavity are another key factor affecting image quality. Environmental factors such as mechanical vibrations exist in real time during system operation and are difficult to completely eliminate. These disturbances cause tiny, random fluctuations in the projection optical path, manifesting as dynamic noise in the final image, severely reducing image clarity and signal-to-noise ratio. Traditional noise suppression algorithms are often designed for general scenarios and lack the ability to specifically model and compensate for the dynamic variations in noise generation and intensity within the specific physical environment of a spherical reflective cavity.

[0006] Currently, solutions to these problems are often fragmented and lack systematic integration. Compensating for optical path offset, correcting for cavity geometric distortion, and suppressing ambient noise are typically treated as independent issues, each addressed using separate modules or algorithms. This fragmented approach not only increases system complexity but also makes it difficult to achieve coordinated optimization of these parameters.

[0007] Based on this, the present invention proposes a three-dimensional image generation method based on spherical reflection cavity parameter optimization. Summary of the Invention

[0008] In order to solve the above-mentioned problems in the prior art, namely, the prior art separately processes optical distortion, geometric deviation and environmental disturbance and lacks a collaborative optimization mechanism, resulting in low imaging accuracy and poor adaptability, the present invention provides a three-dimensional image generation method based on spherical reflection cavity parameter optimization, the method comprising:

[0009] Obtain the optical characteristic parameters of the inner wall of the spherical reflection cavity to calculate the deviation angle of the reflected light under different incident angles and generate a comparison table of the reflection path deviation; measure the deviation between the actual shape parameters of the cavity and the design reference shape parameters, and construct a geometric compensation mapping relationship;

[0010] Real-time monitoring of environmental disturbance parameters inside the cavity and conversion into noise intensity coefficients; inputting the reflection path offset comparison table, geometric compensation mapping relationship and noise intensity coefficient into an image pre-compensation processor;

[0011] The target three-dimensional image is input into the image pre-compensation processor, and the following operations are performed:

[0012] Performing optical path compensation on the original image according to the reflection path offset comparison table; performing geometric distortion correction based on the geometric compensation mapping relationship; applying the noise intensity coefficient to suppress environmental noise; and outputting the pre-corrected image to a projection device;

[0013] The image sensor is used to capture the actual projected distorted image, extract the difference data between the distorted image and the pre-corrected image, and decompose the difference data into a system error component and an environmental interference component;

[0014] The system error component is used to update the reflection path offset comparison table and / or the geometric compensation mapping relationship; the environmental interference component is used to adjust the noise intensity coefficient;

[0015] The updated parameters are fed back to the image pre-compensation processor, and pre-correction is re-performed to generate the final three-dimensional image.

[0016] Furthermore, the offset angle of the reflected light at different incident angles is calculated to generate a comparison table of the reflected path offset. The method is as follows:

[0017] The roughness parameters and refractive index parameters of the inner wall coating at the spatial position are collected synchronously. The roughness parameters are used to calculate the diffuse reflection component caused by micro-surface scattering, and the refractive index parameters are used to calculate the specular reflection component formed by Fresnel reflection.

[0018] Superimposing the diffuse reflection component and the specular reflection component according to energy weights to form a composite reflection model;

[0019] Based on the composite reflection model, the geometric optical reflection law is modified to establish a nonlinear mapping relationship between the incident angle and the actual reflection angle;

[0020] According to the nonlinear mapping relationship and the spatial coordinates of the cavity surface, a comparison table including the incident angle, the reflection angle deviation value and the three-dimensional coordinate offset vector is generated.

[0021] Furthermore, the process of correcting the geometrical optics reflection law based on the composite reflection model includes:

[0022] Under the target incident angle, the energy peak direction in the composite reflection model is determined as the actual main reflection direction;

[0023] Calculating the angle between the actual main reflection direction and the theoretical direction of the geometrical optics reflection law as the reflection deviation of the incident angle;

[0024] According to the reflection deviation and the local curvature radius of the cavity, the three-dimensional coordinate offset vector of the light on the imaging plane is derived.

[0025] Furthermore, the environmental disturbance parameters inside the cavity are monitored in real time and converted into noise intensity coefficients, including:

[0026] Continuously collect the original vibration acceleration signal output by the vibration sensor arranged on the key reflection area of ​​the cavity;

[0027] Performing a first-stage bandpass filtering process on the original vibration acceleration signal within a preset frequency range to extract a resonance frequency band signal related to the display frame rate of the projection image;

[0028] Calculating the average energy value of the resonant frequency band signal in a continuous time window to obtain a standardized vibration energy index;

[0029] Inputting the standardized vibration energy index into a preset vibration noise mapping table, wherein the vibration noise mapping table stores a discrete mapping relationship between vibration energy range and image noise intensity value;

[0030] Obtain the image noise intensity value corresponding to the current standardized vibration energy index by querying the vibration noise mapping table;

[0031] The acquired image noise intensity value is used as the noise intensity coefficient.

[0032] Furthermore, according to the reflection path offset comparison table, optical path compensation is performed on the original image, and the method is as follows:

[0033] Extracting the incident inclination angle value of the initial incident light direction vector of each pixel point of the target three-dimensional image in the local coordinate system of the cavity surface; traversing the matching entries in the reflection path offset comparison table using the incident inclination angle value as an index;

[0034] Reading a three-dimensional coordinate offset vector associated with the current incident tilt angle value from the matching entry; performing an inverse translation calculation on the display coordinates of the corresponding pixel point in the original image based on the component values ​​of the three-dimensional coordinate offset vector;

[0035] After completing the coordinate translation calculation pixel by pixel, the optically compensated image is output.

[0036] Furthermore, based on the geometric compensation mapping relationship, geometric distortion correction is performed, and the method is as follows:

[0037] The geometric compensation mapping relationship includes a coordinate error parameter set between the design reference surface coordinates of the spherical reflection cavity and the actual measured surface coordinates;

[0038] Mapping the projection coordinates of each pixel point of the optical compensation image to the design reference surface coordinate system;

[0039] According to the position of the projection coordinates of the target pixel point in the reference surface coordinate system, query the corresponding surface deformation deviation value in the coordinate error parameter set;

[0040] According to the direction and modulus of the queried surface deformation deviation value, the display coordinates of the target pixel point in the optical compensation image are adjusted for reverse compensation;

[0041] After completing the reverse compensation adjustment for all pixels, the image data is reconstructed and the geometric distortion corrected image is output.

[0042] Furthermore, the noise intensity coefficient is applied to suppress environmental noise in the following manner:

[0043] Loading the geometric distortion corrected image into the dynamic noise processing channel, and matching the preset filtering strategy according to the numerical range of the noise intensity coefficient;

[0044] When the coefficient value is less than the first threshold, the spatial domain mean filtering strategy is selected;

[0045] When the coefficient value is between the first threshold and the second threshold, selecting the temporal motion compensation filtering strategy;

[0046] When the coefficient value is greater than the second threshold, a multi-stage cascade noise reduction strategy is selected;

[0047] A noise suppression operation is performed on spatial pixel groups of the geometrically-distorted image according to a selected filtering strategy, where:

[0048] The spatial domain mean filtering strategy includes: calculating the weighted average of the color components in the neighborhood of the target pixel to replace the original pixel value;

[0049] The temporal motion compensation filtering strategy includes: compensating and correcting the motion pixel trajectory between consecutive frames to eliminate dynamic blur noise;

[0050] The multi-level cascade noise reduction strategy includes sequentially performing spatial noise suppression and temporal trajectory smoothing processing; and outputting the target three-dimensional image after noise suppression processing to a projection device.

[0051] Furthermore, the image sensor is used to capture the actual projected distorted image, and the difference data between the distorted image and the pre-corrected image is extracted. The difference data is then decomposed into a system error component and an environmental interference component. The method is as follows:

[0052] The actual projection image data after being reflected by the cavity is collected by an image sensor fixed on the projection receiving surface;

[0053] The target 3D image after noise suppression is used as the pre-correction image reference, and the color values ​​of the corresponding positions of the actual projection image and the pre-correction image are compared pixel by pixel to calculate the original difference data matrix;

[0054] Based on the spatial distribution of optical errors mapped by the reflection path offset comparison table, a subset of differences with fixed spatial patterns is separated from the original difference data matrix as the systematic error component.

[0055] According to the time-varying characteristics of the environmental noise associated with the noise intensity coefficient, the fluctuation component that conforms to the variation period of the noise intensity coefficient is filtered out from the original difference data as the environmental interference component.

[0056] Furthermore, the system error component is used to update the reflection path offset comparison table and / or the geometric compensation mapping relationship, and the method is:

[0057] Identify the spatial region position markers of the system error components on the cavity surface;

[0058] Associating the spatial region position mark with the incident angle distribution range recorded in the reflection path offset comparison table;

[0059] The statistical mean of the system error components of all pixels in the spatial area is used as the offset vector update amount corresponding to the incident angle range;

[0060] Iteratively correct the three-dimensional coordinate offset vector originally recorded in the reflection path offset comparison table using the offset vector update amount;

[0061] Synchronously associating the spatial region position mark with the surface coordinate node in the geometric compensation mapping relationship;

[0062] Extract the mean value of the directional vector component of the system error component as the deformation compensation increment of the coordinate node corresponding to the geometric compensation mapping relationship;

[0063] The deformation compensation increment is added to the original compensation value of the coordinate node in the geometric compensation mapping relationship to generate an updated geometric compensation mapping relationship.

[0064] Furthermore, the environmental interference component is used to adjust the noise intensity coefficient in the following manner:

[0065] Extract the time-varying fluctuation amplitude peak characteristics contained in the environmental interference component, and obtain the corresponding noise intensity reference value by querying the vibration noise mapping table;

[0066] The currently stored noise intensity coefficient value is called, and the noise intensity reference value is overwritten to the currently stored noise intensity coefficient value to generate a noise intensity coefficient update value.

[0067] Beneficial effects of the present invention:

[0068] The imaging quality and system adaptability are significantly improved. This method obtains the optical characteristic parameters of the cavity to construct a reflection path offset comparison table, and measures the actual shape deviation of the cavity to construct a geometric compensation mapping relationship. At the same time, it monitors the environmental disturbance in real time and converts it into a noise intensity coefficient. These key parameters are uniformly input into the image pre-compensation processor for integrated processing. In the pre-compensation stage, the original three-dimensional image is synchronously subjected to optical path compensation based on the offset comparison table, geometric distortion correction based on the geometric compensation mapping, and environmental noise suppression using the noise intensity coefficient, and a pre-corrected image is output. This integrated pre-compensation processing can effectively and collaboratively correct the three mutually coupled error sources of optical distortion, geometric deviation, and environmental noise, significantly improving the accuracy and clarity of the initial projected image.

[0069] More importantly, the present invention establishes a closed-loop feedback optimization mechanism. The image sensor captures the actual projected distorted image, and extracts the difference data between it and the pre-corrected image. These differences are then intelligently decomposed into components reflecting the inherent errors of the system and components reflecting random interference from the environment. The system error component is used to dynamically update the reflection path offset comparison table and the geometric compensation mapping relationship, so that the compensation model can continuously track and adapt to possible slow changes in the cavity itself; at the same time, the environmental interference component is used to adjust the noise intensity coefficient in real time, so that the noise suppression intensity accurately matches the current environmental disturbance level. Finally, the updated parameters are fed back to the pre-compensation processor to regenerate the final three-dimensional image.

[0070] This mechanism enables the system to continuously self-learn and self-optimize, effectively overcoming the limitations of static parameter configuration. The system can automatically compensate for the effects of cavity aging, minor installation changes, or fluctuating environmental conditions, maintaining high imaging quality over the long term. This significantly enhances the system's robustness and environmental adaptability, ultimately achieving high-precision, high-stability, and high-fidelity 3D image generation. BRIEF DESCRIPTION OF THE DRAWINGS

[0071] Other features, objects and advantages of the present application will become more apparent upon reading the detailed description of non-limiting embodiments made with reference to the following drawings:

[0072] Figure 1 It is a flow chart of a three-dimensional image generation method based on spherical reflection cavity parameter optimization of the present invention;

[0073] Figure 2 This is a flow chart of generating a reflection path offset comparison table of a three-dimensional image generation method based on spherical reflection cavity parameter optimization of the present invention;

[0074] Figure 3 This is a flow chart of noise intensity coefficient generation of a three-dimensional image generation method based on spherical reflection cavity parameter optimization of the present invention;

[0075] Figure 4 This is a flow chart of optical path compensation of a three-dimensional image generation method based on spherical reflection cavity parameter optimization according to the present invention;

[0076] Figure 5 This is a flow chart of geometric distortion correction of a three-dimensional image generation method based on spherical reflection cavity parameter optimization according to the present invention;

[0077] Figure 6 This is a flow chart for calculating system error components and environmental interference components of a three-dimensional image generation method based on spherical reflection cavity parameter optimization of the present invention. DETAILED DESCRIPTION

[0078] The present application will be further described in detail below with reference to the accompanying drawings and examples. It should be understood that the specific embodiments described herein are intended only to illustrate the relevant invention and are not intended to limit the invention. It should also be noted that, for ease of description, only portions relevant to the relevant invention are shown in the accompanying drawings.

[0079] It should be noted that, in the absence of conflict, the embodiments and features of the embodiments in this application can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0080] A first embodiment of the present invention provides a method for generating a three-dimensional image based on parameter optimization of a spherical reflection cavity, the method comprising:

[0081] Step S10: Obtain optical characteristic parameters of the inner wall of the spherical reflection cavity to calculate the offset angle of the reflected light at different incident angles and generate a comparison table of reflection path offsets; measure the deviation between the actual shape parameters of the cavity and the design reference shape parameters to construct a geometric compensation mapping relationship;

[0082] Step S20, real-time monitoring of environmental disturbance parameters inside the cavity and converting them into noise intensity coefficients; inputting the reflection path offset comparison table, geometric compensation mapping relationship and noise intensity coefficient into an image pre-compensation processor;

[0083] Step S30: Input the target 3D image into the image pre-compensation processor, and perform the following operations:

[0084] Performing optical path compensation on the original image according to the reflection path offset comparison table; performing geometric distortion correction based on the geometric compensation mapping relationship; applying the noise intensity coefficient to suppress environmental noise; and outputting the pre-corrected image to a projection device;

[0085] Step S40, capturing the actual projected distorted image through an image sensor, extracting difference data between the actual projected distorted image and the pre-corrected image, and decomposing the difference data into a system error component and an environmental interference component;

[0086] The system error component is used to update the reflection path offset comparison table and / or the geometric compensation mapping relationship; the environmental interference component is used to adjust the noise intensity coefficient;

[0087] Step S50: Feedback the updated parameters to the image pre-compensation processor, and re-perform pre-correction to generate a final three-dimensional image.

[0088] In order to more clearly illustrate the three-dimensional image generation method based on spherical reflection cavity parameter optimization of the present invention, the following is combined with Figure 1 Each step in the embodiment of the present invention is described in detail, including step S10 to step S50, and each step is described in detail as follows:

[0089] Step S10: Obtain optical characteristic parameters of the inner wall of the spherical reflection cavity to calculate the offset angle of the reflected light at different incident angles and generate a comparison table of reflection path offsets; measure the deviation between the actual shape parameters of the cavity and the design reference shape parameters to construct a geometric compensation mapping relationship;

[0090] like Figure 2 As shown, in this embodiment, the offset angle of the reflected light at different incident angles is calculated to generate a comparison table of the reflected path offset, and the method is as follows:

[0091] Step S11, synchronously collecting roughness parameters and refractive index parameters of the inner wall coating associated with the spatial position, wherein the roughness parameter is used to calculate the diffuse reflection component caused by micro-surface scattering, and the refractive index parameter is used to calculate the specular reflection component formed by Fresnel reflection;

[0092] Step S12, superimposing the diffuse reflection component and the specular reflection component according to energy weights to form a composite reflection model;

[0093] Step S13, modifying the geometrical optics reflection law based on the composite reflection model to establish a nonlinear mapping relationship between the incident angle and the actual reflection angle;

[0094] Step S14: generating a comparison table including incident angle, reflection angle deviation value and three-dimensional coordinate offset vector according to the nonlinear mapping relationship and the cavity curved surface space coordinates.

[0095] During the specific implementation of step S10, it is first necessary to accurately obtain the optical characteristic parameters of the inner wall of the spherical reflection cavity. The surface roughness parameters and material refractive index parameters associated with specific spatial position points on the inner wall of the cavity are synchronously collected by a high-precision optical measurement device. In this embodiment, the high-precision optical measurement device is preferably a confocal laser microscope system. The surface roughness parameters are used to quantitatively calculate the intensity distribution of the diffuse reflection component caused by the scattering effect of the coating microstructure, while the refractive index parameters are used to accurately solve the intensity distribution of the specular reflection component based on the Fresnel reflection law. These two types of parameters need to be calibrated and measured on the discretized grid points of the cavity surface to ensure strict correspondence in spatial position.

[0096] After obtaining the above basic parameters, a physical fusion calculation of the diffuse and specular reflection components is performed. Based on the law of conservation of light energy, the diffuse and specular reflection components measured at the same spatial location are vector-superimposed according to their respective weighted contributions to the total reflected energy, constructing a composite reflection physical model that characterizes the actual reflective properties at that location. This model comprehensively considers the dual effects of material microstructural scattering and macroscopic optical interfaces, breaking through the limitations of ideal mirror reflection in traditional geometric optics. The weight ratio is determined by experimental calibration: the total reflectivity of the coating is measured under a standard light source, and a weighted lookup table is established by separating the energy contributions of the diffuse and specular reflection components. Ultimately, a composite bidirectional reflectance distribution function (BRDF) model is generated that integrates the two reflection mechanisms.

[0097] Using this composite BRDF model, an incident ray is set in a ray tracing simulation platform (such as Zemax or CODEV). For each incident angle, the ray propagation equation of the composite reflection model is iteratively solved to calculate the angular deviation (Δθ) between the actual reflection direction and the ideal geometric reflection direction. This deviation is determined by both the offset in the specular reflection direction and the directional dispersion caused by diffuse scattering. Ultimately, a nonlinear mapping function is formed, which takes the incident angle and surface normal as input and outputs the reflection angle deviation. This function is stored as a discrete data table, covering the discrete grid points on the entire cavity surface.

[0098] Based on the nonlinear mapping function, the cavity surface is spatially meshed (grid resolution ≤ 1° × 1°). For each grid cell, the normal vector at its center point coordinates is calculated, and the incident angles from 0° to 90° (step size ≤ 0.5°) are traversed, and the corresponding reflection angle deviation value Δθ is recorded. Further combined with the ray tracing algorithm, the angle deviation is converted into an offset vector (Δx, Δy, Δz) of three-dimensional spatial coordinates. Finally, a structured comparison table containing reflection angle deviation values ​​and three-dimensional offset vectors is generated with the incident angle-surface coordinates as the index key and stored in the non-volatile memory of the pre-compensation processor.

[0099] In this embodiment, the process of correcting the geometrical optics reflection law based on the composite reflection model includes:

[0100] Step S131, under the target incident angle, determining the energy peak direction in the composite reflection model as the actual main reflection direction;

[0101] Step S132, calculating the angle between the actual main reflection direction and the theoretical direction of the geometrical optics reflection law as the reflection deviation of the incident angle;

[0102] Step S133 , deriving the three-dimensional coordinate offset vector of the light on the imaging plane according to the reflection deviation and the local curvature radius of the cavity.

[0103] When modifying the reflection law based on the composite reflection model, the actual principal reflection direction must first be determined. Based on the target incident angle, a gradient descent algorithm is used to search for the global peak of the reflected energy distribution in the spherical coordinate system within the pre-established composite bidirectional reflectance distribution function model.

[0104] The specific operation is to fix the incident direction and iteratively calculate the reflected energy values ​​corresponding to different outgoing directions in a hemispherical space with a step size of 0.1°. When the energy change rate between three consecutive iterations is less than 1%, the outgoing direction vector corresponding to the energy peak is determined to be the actual main reflection direction under the current incident conditions. This direction combines the superposition effects of specular reflection and diffuse scattering, breaking through the limitations of the ideal reflection law.

[0105] After determining the actual main reflection direction, the reflection deviation is accurately calculated. According to the law of geometric optics reflection, under the same conditions of incident angle and surface normal, the theoretical reflection direction vector is calculated. Then, the spatial angle θ between the actual main reflection direction vector and the theoretical reflection direction vector is solved by the vector dot product formula. d This angle is the reflection deviation corresponding to the current angle of incidence. The calculation process is implemented using a floating-point processor with an angular resolution of 0.01°, ensuring that the accuracy of the deviation data meets the requirements of high-fidelity imaging.

[0106] Finally, the three-dimensional coordinate offset vector is derived. Based on the reflection deviation θ obtained above d , combining the physical laws of light propagation to perform spatial geometric transformation calculations. The specific steps include: obtaining the local curvature radius R of the cavity at the measurement point (obtained by fitting the 3D laser scanning point cloud data), setting the propagation distance L of the light from the reflection point to the imaging plane (determined by the system optical design parameters). Establishing a two-dimensional coordinate system in the plane containing the incident surface, normal and reflected light, and converting the angle deviation θ d Converted to lateral offset Δs=L·tan(θ d ). Further, according to the curvature radius R and the direction of the tangent plane of the reflection point, Δs is decomposed into a three-dimensional offset vector (Δx, Δy, Δz) in the imaging plane coordinate system, where the offset component Δz in the direction perpendicular to the imaging plane is given by the formula Δz=R×(1-cos(θ d The resulting three-dimensional coordinate offset vector is written into the reflection path offset comparison table, providing spatial deformation parameters for optical path compensation.

[0107] In this embodiment, the deviation between the actual shape parameters of the cavity and the design reference shape parameters is measured to construct a geometric compensation mapping relationship. The method is as follows:

[0108] During the measurement phase of the cavity's actual shape parameters, data acquisition is performed using a portable three-dimensional coordinate measuring arm. Its equipped laser scanning head moves along a preset path along the cavity's inner wall. Three reference spheres are fixed to the cavity base to establish a measurement coordinate system. The cavity's curved surface is then scanned at a 0.5 mm point spacing, covering the entire inner surface at 10-degree intervals along the meridian direction. Ultimately, a point cloud dataset of no fewer than 5,000 spatial sampling points is acquired. This point cloud data is recorded in real time by the device's built-in software, containing the 3D coordinate information for each sampling point, with measurement accuracy maintained within ±0.03 mm.

[0109] The deviation calculation process is performed within the design software environment. The original CAD design model of the cavity is imported into Geomagic Control X measurement and analysis software, where the measured point cloud data is automatically aligned with the theoretical model using the software's best fit alignment function. The software automatically projects each measured point, mapping it perpendicularly to the nearest point on the CAD surface. The normal distance between the two points is then calculated as the deviation value for that position. For example, the deviation value measured at coordinates (102.3, 55.1, -20.8) is +0.12 mm, representing the offset of the actual surface relative to the designed surface along the normal direction. The coordinates and deviation values ​​of all measured points are exported as a structured data table.

[0110] The construction of the geometric compensation mapping relationship adopts the discretized grid processing method. The surface of the cavity CAD model is divided into 10 mm × 10 mm grid units, and each grid unit covers an area of ​​approximately 100 square millimeters. The arithmetic mean of the deviation values ​​of all measurement points falling into the same unit is taken, and the average value is used as the geometric compensation amount of the grid unit. For example, a grid unit contains 12 measurement points, and their deviation values ​​are +0.10, +0.11, +0.15..., respectively. The average compensation amount is calculated to be +0.12 mm. Finally, a two-dimensional compensation mapping table is generated. Each record in the table contains the coordinates of the grid center point and the corresponding compensation value, forming a direct mapping relationship from spatial position to compensation amount.

[0111] Real-time compensation is implemented with rapid response through spatial positioning. When the coordinates of the projected light point are transmitted to the pre-compensation processor, the system first locates the grid cell where the point is located, and then calls the compensation value stored in the mapping table. When compensation is performed, the original coordinates are translated along the surface normal. For example, if the original coordinates of a point are (105.0, 50.0, -20.0), the compensation amount is +0.12 mm, and the normal vector of the point is (0.12, 0.30, -0.95). After normalization, the corrected coordinates are calculated as (105.0 + 0.12 × 0.12, 50.0 + 0.12 × 0.30, -20.0 + 0.12 × -0.95). The compensation process is completed within 5 milliseconds, ensuring real-time imaging requirements.

[0112] The system uses a dynamic update mechanism to ensure long-term accuracy. When the temperature sensor detects a change in the cavity's ambient temperature exceeding 5 degrees Celsius, a re-measurement process is automatically triggered. The updated compensation mapping table is loaded into the processor memory via hot swap, while retaining historical versions for rollback. Accuracy verification uses a feature point review method. Three feature points equally spaced along the cavity's equatorial circle are selected and reflective markers are affixed. A laser tracker is used to measure the difference between the actual position and the compensated coordinates. The mapping table is considered valid when the maximum deviation is ≤0.05 mm.

[0113] Step S20, real-time monitoring of environmental disturbance parameters inside the cavity and converting them into noise intensity coefficients; inputting the reflection path offset comparison table, geometric compensation mapping relationship and noise intensity coefficient into an image pre-compensation processor;

[0114] like Figure 3 As shown, in this embodiment, the environmental disturbance parameters inside the cavity are monitored in real time and converted into noise intensity coefficients, specifically including:

[0115] Step S21, continuously collecting the original vibration acceleration signal output by the vibration sensor arranged on the key reflection area of ​​the cavity;

[0116] Step S22, performing a first-stage band-pass filtering process within a preset frequency range on the original vibration acceleration signal to extract a resonance frequency band signal related to the display frame rate of the projection image;

[0117] Step S23, calculating the average energy value of the resonant frequency band signal in a continuous time window to obtain a standardized vibration energy index;

[0118] Step S24: inputting the standardized vibration energy index into a preset vibration noise mapping table, wherein the vibration noise mapping table stores a discrete mapping relationship between vibration energy range and image noise intensity value;

[0119] Step S25 , obtaining an image noise intensity value corresponding to the current standardized vibration energy index by querying a vibration noise mapping table, and using the obtained image noise intensity value as the noise intensity coefficient.

[0120] During the real-time monitoring phase, a three-axis MEMS accelerometer is fixedly installed in the key reflection area of ​​the spherical reflection cavity as a vibration sensor. Specifically, a model with a bandwidth ≥1kHz and a resolution ≤100μg is selected. Each sensor continuously outputs the original vibration acceleration signal at a sampling rate of 2000Hz. The signal contains time domain waveform data of the three axes X / Y / Z. The key reflection area is defined as six equally divided areas within the longitude range of ±30° on the equatorial plane of the cavity. A sensor is arranged at the center point of each area. The sensor base is rigidly connected to the cavity wall by vacuum adsorption to ensure the efficiency of vibration transmission. All sensors are connected to the 16-bit ADC data acquisition card through shielded cables, and a continuously updated circular data buffer is established in the pre-compensation processor. The buffer depth maintains the original waveform data of the last 2 seconds.

[0121] The signal processing link performs two-stage filtering and energy extraction. First, the original vibration signal is subjected to a first-stage bandpass filter of 8-12Hz. Experimental measurements show that this frequency band has a harmonic resonance relationship with the 60Hz display frame rate of the projection equipment. The filtering is implemented using a zero-phase FIR filter with an order of 100 to ensure that the passband fluctuation is ≤0.1dB. The filtered signal is then segmented in the time domain, with a time window length of 0.5 seconds (corresponding to a 30-frame image period), and the effective value (RMS) of the signal in each window is calculated as the average energy value. . This value is calculated using the formula:

[0122] ;

[0123] Where N is the number of sampling points in the window (N=1000), x k 、y k 、z k The three-axis filtered acceleration values ​​are obtained respectively. Finally, the Evib values ​​of the six sensors are arithmetic averaged to obtain the standardized vibration energy index.

[0124] The vibration-noise mapping relationship is established based on pre-calibrated experimental data. In a temperature-controlled laboratory environment, a standard vibration table is used to apply controlled vibrations of 0.1g to 1.0g (in steps of 0.1g) to the cavity, and projected test images are simultaneously collected. By calculating the decrease in the image structure similarity index (SSIM), a quantitative correspondence between vibration energy and image quality loss is established. For example, when Evib = 0.3g, the SSIM drops to 0.85, and the noise intensity coefficient is defined as 0.7 (normalized value range 0-1.0). Finally, a discrete mapping table is generated:

[0125] Vibration energy range (g) Noise intensity factor [0.0，0.2) 0.3 [0.2，0.4) 0.7 [0.4，0.6) 0.9

[0126] This table is burned into the processor's read-only memory in binary form.

[0127] The real-time conversion process is achieved by table interpolation. When the current standardized vibration energy index Evib (such as 0.35g) is obtained, the mapping table is searched to determine whether it is in the range [0.2g, 0.4g]. The linear interpolation algorithm is used to calculate the accurate noise intensity coefficient. K noise :

[0128] ;

[0129] The calculated result is rounded to two decimal places and output as the current noise intensity coefficient. The system updates the coefficient value every 0.5 seconds, with an update delay of less than 10ms to ensure synchronization with the image generation frame rate.

[0130] Step S30: Input the target 3D image into the image pre-compensation processor, and perform the following operations:

[0131] Performing optical path compensation on the original image according to the reflection path offset comparison table; performing geometric distortion correction based on the geometric compensation mapping relationship; applying the noise intensity coefficient to suppress environmental noise; and outputting the pre-corrected image to a projection device;

[0132] like Figure 4 As shown, in this embodiment, according to the reflection path offset comparison table, the optical path compensation is performed on the original image, and the method is as follows:

[0133] Step S31, extracting the incident inclination angle value of the initial incident light direction vector of each pixel point of the target three-dimensional image in the local coordinate system of the cavity surface; using the incident inclination angle value as an index, traversing the matching entries in the reflection path offset comparison table;

[0134] Step S32: reading a three-dimensional coordinate offset vector associated with the current incident tilt angle value from the matching entry; performing a reverse translation calculation on the display coordinates of the corresponding pixel point in the original image based on the component values ​​of the three-dimensional coordinate offset vector;

[0135] Step S33: Outputting an optically compensated image after completing coordinate translation calculation pixel by pixel.

[0136] Optical path compensation starts with the analysis of the direction of the incident light. The pre-compensation processor receives the original pixel data stream of the target three-dimensional image and calculates the spatial direction of the initial incident light in the local coordinate system of the cavity for each pixel. It is specifically implemented through the following steps: first, a global coordinate system with the center of the cavity as the origin is established, and the coordinates of the emission source of the light corresponding to each pixel are determined in combination with the optical parameters of the projection equipment; then, based on the theoretical intersection position of the light and the cavity surface, a local coordinate system with the normal of the intersection as the Z axis is constructed; finally, the incident light vector is converted to the local coordinate system through the coordinate transformation matrix, and its angle with the Z axis is calculated as the incident inclination angle α For example, if the incident vector of a pixel is (0.12, 0.30, -0.95), the calculation result is α =arccos(0.95)≈18.19°. The inclination angle is stored as a floating point number with an accuracy of 0.01°.

[0137] The offset matching query uses a hierarchical index mechanism. The reflection path offset comparison table is stored in the memory according to the incident inclination angle. α Sorted storage, index structure is B+ tree. αThe value (e.g. 18.19°) is used as the query key, and a binary search is performed in the lookup table: first locate the storage block of 18.0°-18.5° (corresponding to an angle step of 0.5°), and then linearly scan the matching entries within the block. The matching condition is | θ table -θ query |≤0.25°, when the 18.20° entry is found, the match is considered successful. If there is no exact match (probability <0.1%), the two nearest neighbor entries are taken for linear interpolation. Each matching record contains a three-dimensional coordinate offset vector ΔV=( Δx, Δy, Δz ), for example in α =18.20° when ΔV=(0.12, -0.08, 0.05) mm is stored.

[0138] The reverse translation of coordinates is achieved through affine transformation. After obtaining the offset vector ΔV, the reverse compensation calculation is performed on the display coordinates of the current pixel: Let the display coordinates of pixel P in the original image be ( X p ，Y p ，Z p ), the coordinate P' after compensation is calculated as follows: P'= ( X p -Δx×k x ，Y p -Δy×k y ，Z p -Δz×k z );

[0139] in k x , k y , k z is the axial scale factor, determined by the projection system calibration (typical value k x = k y =1.2, k z=0.8). For example, the original coordinates (105.0, 50.0, -20.0) are scaled by ΔV = (0.12, -0.08, 0.05) to produce the output coordinates (105.0 - 0.12 × 1.2, 50.0 - (-0.08) × 1.2, -20.0 - 0.05 × 0.8) = (104.856, 50.096, -20.04). This calculation is performed in parallel in the GPU shader, with each pixel processed independently.

[0140] Boundary processing is performed before outputting the compensation results. When the compensation coordinates exceed the display buffer range: if it is a single point out-of-bounds, the nearest valid pixel value is used for filling; if it is a continuous area out-of-bounds, the surface continuation algorithm is activated to predict the pixel value based on the curvature of the adjacent valid area. The final optically compensated image is transmitted to the projection device at a frame rate of 120fps, and the processing latency is measured to be 3.2ms / frame (at 4K resolution).

[0141] like Figure 5 As shown, in this embodiment, based on the geometric compensation mapping relationship, geometric distortion correction is performed, and the method is as follows:

[0142] Step S34, the geometric compensation mapping relationship includes a coordinate error parameter set between the design reference surface coordinates of the spherical reflection cavity and the actual measured surface coordinates;

[0143] Step S35, mapping the projection coordinates of each pixel point of the optical compensation image to the design reference curved surface coordinate system;

[0144] Step S36, querying the corresponding surface deformation deviation value in the coordinate error parameter set according to the position of the projection coordinates of the target pixel point in the reference surface coordinate system;

[0145] Step S37, performing reverse compensation adjustment on the display coordinates of the target pixel point in the optical compensation image according to the direction and modulus of the queried surface deformation deviation value;

[0146] Step S38, after completing reverse compensation adjustment for all pixels, reconstruct the image data and output a geometric distortion corrected image.

[0147] The geometric distortion correction operation begins with coordinate mapping transformation. The pre-compensation processor takes the image that has completed optical path compensation as input and performs coordinate system transformation on the projection coordinates of each pixel in the image. The specific process is as follows: let the display coordinates of the pixel point in the optically compensated image be (u, v), and combine the intrinsic parameter matrix of the projection device and the cavity posture matrix to calculate its three-dimensional space coordinates in the design reference surface coordinate system through the inverse perspective projection transformation. For example, the position of a pixel point (u=320, v=240) in the reference coordinate system is calculated to be (102.3 mm, 55.1 mm, -20.8 mm), which corresponds to the parametric coordinates of the surface in the CAD model (s=0.35, t=0.72).

[0148] The coordinate error parameter set is stored using an octree spatial index structure. This dataset contains 100,000 discrete sampling points, each of which records the deviation vector between the designed coordinate position and the actual measured coordinate position. The spatial index is divided into 1mm grids, and each octree node stores the designed coordinates and corresponding 3D deviations for all sampling points within its spatial bounding box. When the reference coordinates of a target pixel are passed in, the system calls the spatial query interface: first, the 3D coordinates are converted to surface parameter coordinates (s, t); then, the leaf node containing the parameter coordinates is located in the octree index (the index depth is 8 levels); the deviation vector data of the four nearest sampling points within the node is extracted; the Euclidean distance from the target point to each sampling point is calculated; and finally, the precise deviation vector at the target point is calculated using an inverse distance weighted interpolation algorithm.

[0149] Deviation compensation performs reverse spatial displacement. After obtaining the deviation vector of the current pixel, the display coordinates are adjusted inversely: assuming the deviation vector obtained is (Δx=0.12mm, Δy=-0.08mm, Δz=0.05mm), and the original display coordinates are (X=105.0, Y=50.0, Z=-20.0), the new coordinates after correction are calculated according to the formula:

[0150] X'=X-Δx×a x ; Y'=Y-Δy×a y ; Z'=Z-Δz×a z ;

[0151] Among them, the scale factor a x =1.2, a y =1.2, a z = 0.8, which is determined by the projection system calibration. Substituting these values ​​yields the actual output coordinates (105.0 - 0.12 × 1.2, 50.0 - (-0.08) × 1.2, -20.0 - 0.05 × 0.8) = (104.856, 50.096, -20.04). This calculation is performed in parallel on the graphics processor, with each pixel processed independently.

[0152] Special boundary conditions are handled during full image reconstruction. When the compensated coordinates exceed the valid display area, if the pixel is isolated and out of bounds, the nearest valid neighbor pixel value is used to fill the gap. If a continuous area is out of bounds, a surface continuation algorithm is activated to predict the current pixel value based on the curvature trend of adjacent valid pixels. After all pixels are processed, the data is reassembled into a complete frame image and output to the projection device.

[0153] In this embodiment, the noise intensity coefficient is applied to suppress environmental noise in the following manner:

[0154] Loading the geometric distortion corrected image into the dynamic noise processing channel, and matching the preset filtering strategy according to the numerical range of the noise intensity coefficient;

[0155] When the coefficient value is less than the first threshold, the spatial domain mean filtering strategy is selected;

[0156] When the coefficient value is between the first threshold and the second threshold, selecting the temporal motion compensation filtering strategy;

[0157] When the coefficient value is greater than the second threshold, a multi-stage cascade noise reduction strategy is selected;

[0158] A noise suppression operation is performed on spatial pixel groups of the geometrically-distorted image according to a selected filtering strategy, where:

[0159] The spatial domain mean filtering strategy includes: calculating the weighted average of the color components in the neighborhood of the target pixel to replace the original pixel value;

[0160] The temporal motion compensation filtering strategy includes: compensating and correcting the motion pixel trajectory between consecutive frames to eliminate dynamic blur noise;

[0161] The multi-level cascade noise reduction strategy includes sequentially performing spatial noise suppression and temporal trajectory smoothing processing; and outputting the target three-dimensional image after noise suppression processing to a projection device.

[0162] The environmental noise suppression operation is started in the dynamic noise processing channel. The pre-compensation processor first loads the image data that has completed geometric distortion correction into the video memory buffer, and reads the real-time updated noise intensity coefficient from the register. K noise (The coefficient range is 0 to 1.0). Automatically select the filtering strategy based on the coefficient value: When it is detected K noise When the coefficient is less than 0.3, spatial mean filtering mode is activated; when the coefficient value is between 0.3 and 0.7, temporal motion compensation filtering mode is enabled; when the coefficient value is greater than or equal to 0.7, multi-stage cascade noise reduction mode is enabled. Strategy switching is completed within 0.1 milliseconds through the hardware state machine, and the switching signal directly controls the enable terminal of the filtering algorithm module.

[0163] Spatial domain mean filtering uses weighted averaging. A 5×5 neighborhood window is created centered around the target pixel, and each color channel is processed independently. Convolution is performed within the window using a preset Gaussian weight matrix, with weights distributed as high at the center and low at the edges: 0.20 for the center pixel, 0.18 for adjacent pixels, 0.14 for diagonal pixels, 0.08 for edge pixels, and 0.02 for corner pixels. For example, if the target pixel's original red channel value is R=120, and the pixel above it has a weight of R=115 (weight 0.18), and the pixel to its left has a weight of R=118 (weight 0.18), the weighted summation yields a new value, R'=120×0.20+115×0.18+118×0.18+..., with the final result rounded to the nearest integer. Processing skips two rows and two columns around the image edge to avoid boundary anomalies.

[0164] Temporal motion compensation filtering requires caching three consecutive frames. The current frame is segmented into 16×16 pixel macroblocks, and the minimum absolute difference (MSAD) is calculated within a ±8-pixel search range of the reference frames (previous and next frames). A SAD threshold of 500 is set. When a macroblock's matching value falls below this threshold, it is considered a moving block, and its motion vector (Vx, Vy) is recorded. Trajectory compensation is performed on each pixel within the moving macroblock: the pixel value in the current frame is weighted 60%, the pixel value at the corresponding position offset (X+Vx, Y+Vy) in the previous frame is weighted 30%, and the pixel value at the corresponding position offset (X-Vx, Y-Vy) in the next frame is weighted 10%. The sum of these three values ​​is used as the new compensated value. For static areas, the median value of the pixels at the same position in the three frames is directly taken.

[0165] Multi-level cascade noise reduction is performed in two stages. The first stage uses non-local mean spatial filtering: a 21×21 pixel search window and a 7×7 pixel similarity window are set, and the filter parameter h=12× K noise (when K noise =0.8 when h=9.6). Calculate the Gaussian weighted Euclidean distance of all similar windows in the search window for the target pixel, and pixels with weights greater than 0.7 participate in the weighted average. The second stage is to connect the time domain Kalman filter: establish a state equation with the pixel RGB value as the state variable, and set the process noise variance to 0.1× K noise The observation noise variance is fixed at 0.05. The output of each frame is a linear fusion of the state prediction value and the current observation value, and the fusion coefficient is dynamically adjusted according to the noise intensity.

[0166] The final processing results are written to the output buffer. The system monitors output quality in real time, calculating the peak signal-to-noise ratio every 10 frames. When the PSNR falls below 35dB, the filtering strength is automatically increased (either by expanding the spatial window to 7×7 or increasing the temporal memory depth to 7 frames). All processing is performed in a dedicated image processing chip. In the worst-case scenario, processing a single 4K resolution frame takes 3.5 milliseconds, meeting the requirements for real-time projection at 120 frames per second. A processing log records the noise figure, usage strategy, and processing time for each frame, and outputs it via the serial port for later analysis and optimization.

[0167] Step S40, capturing the actual projected distorted image through an image sensor, extracting difference data between the actual projected distorted image and the pre-corrected image, and decomposing the difference data into a system error component and an environmental interference component;

[0168] The system error component is used to update the reflection path offset comparison table and / or the geometric compensation mapping relationship; the environmental interference component is used to adjust the noise intensity coefficient;

[0169] like Figure 6 As shown, in this embodiment, the image sensor is used to capture the actual projected distorted image, and the difference data between the actual projected distorted image and the pre-corrected image is extracted. The difference data is then decomposed into a system error component and an environmental interference component. The method is as follows:

[0170] Step S41, collecting actual projection image data after being reflected by the cavity through an image sensor fixed on the projection receiving surface;

[0171] Step S42: Using the noise-suppressed target 3D image as a pre-corrected image reference, pixel by pixel comparison is performed between the color values ​​of corresponding positions of the actual projection image and the pre-corrected image to calculate an original difference data matrix.

[0172] Step S43, based on the optical error spatial distribution law mapped by the reflection path offset comparison table, separating the difference subset with a fixed spatial pattern from the original difference data matrix as the systematic error component;

[0173] Step S44 , based on the time-varying characteristics of the environmental noise associated with the noise intensity coefficient, the fluctuation component that conforms to the variation period of the noise intensity coefficient is filtered out from the original difference data as the environmental interference component.

[0174] In this embodiment, image acquisition is achieved by a high-resolution industrial camera. A 20-megapixel CMOS image sensor is fixedly installed at the center of the projection receiving surface, and its spectral response range covers the color gamut of the projection device. The camera captures the actual image projected by the spherical reflection cavity at a frame rate of 120fps and transmits the 12-bit RAW data to the processing unit in real time via the CameraLink interface. The synchronization signal generator ensures that the acquisition frame is strictly aligned with the pre-corrected image output frame, and the time deviation is controlled within ±100μs. A spatial coordinate label is attached to each frame of the image to record its absolute position on the receiving surface.

[0175] The difference calculation performs pixel-level comparison. The pre-corrected image reference data is read from the dual-port RAM and spatially aligned with the actual projected image: First, an affine transformation matrix is ​​established through SIFT feature matching to eliminate the overall offset caused by mechanical displacement; then, the absolute difference value ΔE(x, y) of the three RGB channels is calculated pixel by pixel. The difference calculation formula is:

[0176] ΔE(x, y)=|R act(x，y) -R pre(x，y) |+|G act(x，y) -G pre(x，y) |+|B act(x，y) -B pre(x，y) |;

[0177] For example, if the measured actual R = 120 and the reference R = 115 at coordinates (100, 200) results in a delta R of 5. A 1920 × 1080 disparity matrix is ​​generated for the full frame, and the data is stored in 32-bit floating-point format. The 3% edge region is marked as invalid due to significant distortion and is not included in subsequent analysis.

[0178] Systematic error separation is based on spatial pattern matching. An optical error distribution template mapped by a reflection path offset comparison table is preloaded. This template contains the theoretical error spatial pattern of each region of the cavity surface. The difference matrix is ​​convolved with the template:

[0179] C(i,j)=∑∑ΔE(m,n)×T(mi,nj);

[0180] Correlation coefficients were calculated within a sliding 5×5 window. When the local correlation coefficient C > 0.85 and the spatial gradient direction was consistent, the difference in that region was considered a systematic error. Pixel sets from all eligible regions were extracted and fitted using the least squares method to generate a surface for the systematic error components. For example, a ring-shaped difference band was isolated at the cavity equator, with an amplitude of approximately 8-12 grayscale levels, which is consistent with the deviation predicted by the optical model.

[0181] The extraction of environmental interference components relies on time-frequency analysis. A sliding window of difference data with a length of 2 seconds (240 frames) is established, and a fast Fourier transform is performed on each pixel sequence in the window.K noise Historical records to determine the current main interference frequency f noise (Typical value 8-12Hz). f noise The energy in the ±1Hz frequency band centered at ( ) is inversely transformed to the time domain to obtain the periodic fluctuation component. For example, a 10Hz sinusoidal fluctuation with an amplitude of ±3 grayscale is detected at coordinates (300, 400). This component is synchronized with the 10.2Hz disturbance recorded by the ambient vibration sensor. Finally, the full frame of periodic fluctuation pixels is output as the environmental interference component.

[0182] In this embodiment, the system error component is used to update the reflection path offset comparison table and / or the geometric compensation mapping relationship, and the method is:

[0183] Step S431, identifying the spatial region position mark of the system error component on the cavity surface;

[0184] Step S432, associating the spatial region position mark with the incident angle distribution range recorded in the reflection path offset comparison table;

[0185] Step S433: taking the statistical mean of the system error components of all pixels in the spatial region as the offset vector update amount corresponding to the incident angle range;

[0186] Step S434, iteratively correcting the three-dimensional coordinate offset vector originally recorded in the reflection path offset comparison table using the offset vector update amount;

[0187] Step S435, synchronously associating the spatial region position mark with the surface coordinate node in the geometric compensation mapping relationship;

[0188] Step S436 , extracting the mean of the directional vector components of the system error components as the deformation compensation increment of the coordinate node corresponding to the geometric compensation mapping relationship;

[0189] Step S437 : adding the deformation compensation increment to the original compensation value of the coordinate node in the geometric compensation mapping relationship to generate an updated geometric compensation mapping relationship.

[0190] The spatial positioning of the system error components is achieved through coordinate mapping. The pixel coordinates of the significant area (error value > threshold 15 grayscale) in the system error component matrix are extracted and converted to the cavity surface coordinate system in combination with the image sensor calibration parameters. The conversion matrix established by the laser tracker is as follows:

[0191] Let pixel coordinates (x p ,y p ), and the three-dimensional coordinates of the surface are obtained through matrix operation (x c,y c , z c ), with a positioning accuracy of ±0.1 mm. For example, a circular area with a radius of 20 mm and a center of (102.3, 55.1, -20.8) was detected to have a systematic deviation, and this area was marked as update area A.

[0192] The incident angle range is associated using reverse ray tracing. For each 3D coordinate point in the marked area A, the incident ray vector is calculated based on the projection device posture parameters. A reverse query is performed in the reflection path offset comparison table:

[0193] Using the normal vector at the coordinate point as a reference, calculate the angle ε between the incident light and the normal, and match all entries within the ±1° range in the lookup table index. For example, if the ε distribution in area A is 17.5°-22.3°, then associate the data records in the lookup table with the five angle intervals of 18°, 19°, 20°, 21°, and 22°.

[0194] The offset vector update is generated by statistical calculation. For all pixels in area A, the projection values ​​of the system error components in the three axes X, Y, and Z are extracted. The mean of the axial deviation is calculated:

[0195] Δx avg =Σ(Δx i ) / N;Δy avg =Σ(Δy i ) / N;Δz avg =Σ(Δz i ) / N;

[0196] Where N is the number of valid pixels, and outliers outside ±3σ are excluded;

[0197] For example, we can calculate Δx avg =+0.13mm, Δy avg =-0.07mm, Δz avg = +0.04mm, forming the update vector ΔV update =(0.13, -0.07, 0.04);

[0198] The weighted fusion strategy is used for iterative correction of the comparison table. For each incident angle entry associated (such as the original offset ΔV of the 20° entry), old =(0.12, -0.08, 0.05)), update according to the formula:

[0199] ΔV new =η×ΔV old +β×ΔV update , where the weight coefficient η=0.8, β=0.2; Substituting the value into ΔV new=0.8×(0.12,-0.08,0.05)+0.2×(0.13,-0.07,0.04)=(0.122,-0.078,0.048).

[0200] The geometric compensation mapping update performs node-level adjustments. The spatial coordinates of region A are mapped to the parameter space of the geometric compensation mapping table, locating the affected surface coordinate nodes. For example, region A corresponds to 32 nodes within the parameter coordinate range (s, t) ∈ [0.30-0.40] × [0.70-0.75]. The direction vectors of the system error components at these nodes are extracted, and the mean normal component (e.g., +0.11 mm) and the mean tangential offset (e.g., -0.05 mm) are calculated.

[0201] The incremental superposition of compensation amount adopts channel processing:

[0202] Normal compensation update:

[0203] Assume the original node normal compensation value δ old =-0.02mm;

[0204] New value δ new =δ old +γ×Δδ mean (γ=0.3); Δδ mean It is the regional statistical average of the projection values ​​of the system error component in the direction of the surface normal.

[0205] Get δ new =-0.02+0.3×0.11=+0.013mm

[0206] Tangential compensation update:

[0207] Original tangent vector T old =(0.01, -0.03) mm;

[0208] New vector T new =T old +γ×ΔT mean =(0.01,-0.03)+0.3×(-0.05,0)=(-0.005,-0.03)mm, ΔT mean is the regional average of the projection vectors of the system error components in the tangent plane of the surface.

[0209] The updated geometry compensation mapping table takes effect after verification, and the spatial interpolation algorithm automatically smooths the transition between nodes. The system records change logs containing region locations, update vectors, version numbers, and timestamps, supporting rollback of historical versions.

[0210] In this embodiment, the environmental interference component is used to adjust the noise intensity coefficient in the following manner:

[0211] Step S441: extracting the time-varying fluctuation amplitude peak feature contained in the environmental interference component, and obtaining the corresponding noise intensity reference value by querying the vibration noise mapping table;

[0212] Step S442 : calling the currently stored noise intensity coefficient value, overwriting the noise intensity reference value with the currently stored noise intensity coefficient value, and generating a noise intensity coefficient update value.

[0213] This embodiment uses an accelerometer mounted on the side wall of the cavity to collect vibration data every 0.1 seconds. When the vibration frequency is detected within the range of 8-12 Hz (the sensitive frequency range of the projection equipment), the vibration amplitude value (unit: g) is recorded.

[0214] Take 50 vibration data within the last 5 seconds, filter out the 5 largest values, calculate the average value as the fluctuation peak, query the corresponding noise intensity reference value, directly overwrite the currently stored noise intensity coefficient value, and generate an updated noise intensity coefficient value.

[0215] Step S50: Feedback the updated parameters to the image pre-compensation processor, and re-perform pre-correction to generate a final three-dimensional image.

[0216] When the system completes updating the reflection path offset comparison table, the geometric compensation mapping relationship, and the noise intensity coefficient, the pre-compensation processor immediately performs parameter reloading.

[0217] Although the various steps in the above embodiment are described in the above-mentioned order, those skilled in the art will understand that in order to achieve the effect of this embodiment, different steps do not have to be executed in such an order. They can be executed simultaneously (in parallel) or in a reverse order. These simple changes are within the scope of protection of the present invention.

[0218] A three-dimensional image generation system based on spherical reflection cavity parameter optimization according to a second embodiment of the present invention is used to implement a three-dimensional image generation method based on spherical reflection cavity parameter optimization. The system includes:

[0219] A module for constructing a reflection path offset comparison table and geometric compensation mapping relationship is configured to obtain optical characteristic parameters of the inner wall of the spherical reflection cavity, calculate the offset angle of the reflected light at different incident angles, and generate a reflection path offset comparison table; measure the deviation between the actual shape parameters of the cavity and the design reference shape parameters, and construct a geometric compensation mapping relationship;

[0220] A data input module is configured to monitor the environmental disturbance parameters inside the cavity in real time and convert them into noise intensity coefficients; input the reflection path offset comparison table, geometric compensation mapping relationship and noise intensity coefficient into an image pre-compensation processor;

[0221] The correction module is configured to input the target three-dimensional image into the image pre-compensation processor and perform the following operations:

[0222] Performing optical path compensation on the original image according to the reflection path offset comparison table; performing geometric distortion correction based on the geometric compensation mapping relationship; applying the noise intensity coefficient to suppress environmental noise; and outputting the pre-corrected image to a projection device;

[0223] a correction module configured to capture the actual projected distorted image through an image sensor, extract difference data between the actual projected distorted image and the pre-corrected image, and decompose the difference data into a system error component and an environmental interference component;

[0224] The system error component is used to update the reflection path offset comparison table and / or the geometric compensation mapping relationship; the environmental interference component is used to adjust the noise intensity coefficient;

[0225] The output module is configured to feed back the updated parameters to the image pre-compensation processor to re-perform pre-correction to generate the final three-dimensional image.

[0226] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working process and related instructions of the system described above can refer to the corresponding process in the aforementioned method embodiment and will not be repeated here.

[0227] It should be noted that the three-dimensional image generation system based on spherical reflection cavity parameter optimization provided in the above embodiment is only illustrated by the division of the above functional modules. In actual applications, the above functions can be assigned to different functional modules as needed, that is, the modules or steps in the embodiments of the present invention can be further decomposed or combined. For example, the modules in the above embodiment can be combined into one module, or further divided into multiple sub-modules to complete all or part of the functions described above. The names of the modules and steps involved in the embodiments of the present invention are only for the purpose of distinguishing the modules or steps and are not to be regarded as improper limitations of the present invention.

[0228] An electronic device according to a third embodiment of the present invention includes:

[0229] at least one processor; and

[0230] a memory communicatively connected to at least one of the processors; wherein,

[0231] The memory stores instructions that can be executed by the processor, and the instructions are used to be executed by the processor to implement the above-mentioned three-dimensional image generation method based on spherical reflection cavity parameter optimization.

[0232] A fourth embodiment of the present invention provides a computer-readable storage medium, wherein the computer-readable storage medium stores computer instructions, and the computer instructions are used to be executed by the computer to implement the above-mentioned three-dimensional image generation method based on spherical reflection cavity parameter optimization.

[0233] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes and related instructions of the storage device and processing device described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0234] Those skilled in the art should be able to appreciate that the modules and method steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of the two, and the programs corresponding to the software modules and method steps can be placed in random access memory (RAM), internal memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disks, removable disks, CD-ROMs, or any other form of storage medium known in the art. In order to clearly illustrate the interchangeability of electronic hardware and software, the above description has generally described the components and steps of each example according to their functions. Whether these functions are performed in electronic hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art may use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0235] The terms "first", "second", etc. are used to distinguish similar objects, rather than to describe or indicate a particular order or sequence.

[0236] The term "comprise" or any other similar term is intended to cover non-exclusive inclusion such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed or inherent to such process, method, article, or apparatus.

[0237] Thus far, the technical solutions of the present invention have been described in conjunction with the preferred embodiments shown in the accompanying drawings. However, it will be readily understood by those skilled in the art that the scope of protection of the present invention is obviously not limited to these specific embodiments. Without departing from the principles of the present invention, those skilled in the art may make equivalent changes or substitutions to the relevant technical features, and the technical solutions after such changes or substitutions will fall within the scope of protection of the present invention.

Claims

1. A three-dimensional image generation method based on spherical reflection cavity parameter optimization, characterized in that: The method includes: Obtain the optical characteristic parameters of the inner wall of the spherical reflection cavity to calculate the deviation angle of the reflected light under different incident angles and generate a comparison table of the reflection path deviation; measure the deviation between the actual shape parameters of the cavity and the design reference shape parameters, and construct a geometric compensation mapping relationship; Real-time monitoring of environmental disturbance parameters inside the cavity and conversion into noise intensity coefficients; inputting the reflection path offset comparison table, geometric compensation mapping relationship and noise intensity coefficient into an image pre-compensation processor; The target three-dimensional image is input into the image pre-compensation processor, and the following operations are performed: Performing optical path compensation on the original image according to the reflection path offset comparison table; performing geometric distortion correction based on the geometric compensation mapping relationship; applying the noise intensity coefficient to suppress environmental noise; and outputting the pre-corrected image to a projection device; The image sensor is used to capture the actual projected distorted image, extract the difference data between the distorted image and the pre-corrected image, and decompose the difference data into a system error component and an environmental interference component; The system error component is used to update the reflection path offset comparison table and / or the geometric compensation mapping relationship; the environmental interference component is used to adjust the noise intensity coefficient; Feeding back the updated parameters to the image pre-compensation processor to re-perform pre-correction to generate the final three-dimensional image; Calculate the offset angle of the reflected light at different incident angles and generate a comparison table of the reflected path offset, which is as follows: The roughness parameters and refractive index parameters of the inner wall coating at the spatial position are collected synchronously. The roughness parameters are used to calculate the diffuse reflection component caused by micro-surface scattering, and the refractive index parameters are used to calculate the specular reflection component formed by Fresnel reflection. Superimposing the diffuse reflection component and the specular reflection component according to energy weights to form a composite reflection model; Based on the composite reflection model, the geometric optical reflection law is modified to establish a nonlinear mapping relationship between the incident angle and the actual reflection angle; Generate a comparison table including incident angle, reflection angle deviation value and three-dimensional coordinate offset vector according to the nonlinear mapping relationship and the cavity surface spatial coordinates; The geometric optics reflection law is modified based on the composite reflection model, including: Under the target incident angle, the energy peak direction in the composite reflection model is determined as the actual main reflection direction; Calculating the angle between the actual main reflection direction and the theoretical direction of the geometrical optics reflection law as the reflection deviation of the incident angle; According to the reflection deviation and the local curvature radius of the cavity, the three-dimensional coordinate offset vector of the light on the imaging plane is derived.

2. The three-dimensional image generation method based on spherical reflection cavity parameter optimization according to claim 1, characterized in that: Real-time monitoring of the environmental disturbance parameters inside the cavity and conversion into noise intensity coefficients, including: Continuously collect the original vibration acceleration signal output by the vibration sensor arranged on the key reflection area of ​​the cavity; Performing a first-stage bandpass filtering process on the original vibration acceleration signal within a preset frequency range to extract a resonance frequency band signal related to the display frame rate of the projection image; Calculating the average energy value of the resonant frequency band signal in a continuous time window to obtain a standardized vibration energy index; Inputting the standardized vibration energy index into a preset vibration noise mapping table, wherein the vibration noise mapping table stores a discrete mapping relationship between vibration energy range and image noise intensity value; Obtain the image noise intensity value corresponding to the current standardized vibration energy index by querying the vibration noise mapping table; The acquired image noise intensity value is used as the noise intensity coefficient.

3. The three-dimensional image generation method based on spherical reflection cavity parameter optimization according to claim 1, characterized in that: According to the reflection path offset comparison table, optical path compensation is performed on the original image, which is specifically as follows: Extracting the incident inclination angle value of the initial incident light direction vector of each pixel point of the target three-dimensional image in the local coordinate system of the cavity surface; traversing the matching entries in the reflection path offset comparison table using the incident inclination angle value as an index; Reading a three-dimensional coordinate offset vector associated with the current incident tilt angle value from the matching entry; performing an inverse translation calculation on the display coordinates of the corresponding pixel point in the original image based on the component values ​​of the three-dimensional coordinate offset vector; After completing the coordinate translation calculation pixel by pixel, the optically compensated image is output.

4. The three-dimensional image generation method based on spherical reflection cavity parameter optimization according to claim 3, characterized in that: Based on the geometric compensation mapping relationship, geometric distortion correction is performed, which is specifically as follows: The geometric compensation mapping relationship includes a coordinate error parameter set between the design reference surface coordinates of the spherical reflection cavity and the actual measured surface coordinates; Mapping the projection coordinates of each pixel point of the optical compensation image to the design reference surface coordinate system; According to the position of the projection coordinates of the target pixel point in the reference surface coordinate system, query the corresponding surface deformation deviation value in the coordinate error parameter set; According to the direction and modulus of the queried surface deformation deviation value, the display coordinates of the target pixel point in the optical compensation image are adjusted for reverse compensation; After completing the reverse compensation adjustment for all pixels, the image data is reconstructed and the geometric distortion corrected image is output.

5. The three-dimensional image generation method based on spherical reflection cavity parameter optimization according to claim 1, characterized in that: The noise intensity coefficient is applied to suppress environmental noise, specifically: Loading the geometric distortion corrected image into the dynamic noise processing channel, and matching the preset filtering strategy according to the numerical range of the noise intensity coefficient; When the coefficient value is less than the first threshold, the spatial domain mean filtering strategy is selected; When the coefficient value is between the first threshold and the second threshold, selecting the temporal motion compensation filtering strategy; When the coefficient value is greater than the second threshold, a multi-stage cascade noise reduction strategy is selected; A noise suppression operation is performed on spatial pixel groups of the geometrically-distorted image according to a selected filtering strategy, where: The spatial domain mean filtering strategy includes: calculating the weighted average of the color components in the neighborhood of the target pixel to replace the original pixel value; The temporal motion compensation filtering strategy includes: compensating and correcting the motion pixel trajectory between consecutive frames to eliminate dynamic blur noise; The multi-level cascade noise reduction strategy includes sequentially performing spatial noise suppression and temporal trajectory smoothing processing; and outputting the target three-dimensional image after noise suppression processing to a projection device.

6. The three-dimensional image generation method based on spherical reflection cavity parameter optimization according to claim 5, characterized in that: The image sensor captures the actual projected distorted image, extracts the difference data between it and the pre-corrected image, and decomposes the difference data into the system error component and the environmental interference component, which are specifically: The actual projection image data after being reflected by the cavity is collected by an image sensor fixed on the projection receiving surface; The target 3D image after noise suppression is used as the pre-correction image reference, and the color values ​​of the corresponding positions of the actual projection image and the pre-correction image are compared pixel by pixel to calculate the original difference data matrix; Based on the spatial distribution of optical errors mapped by the reflection path offset comparison table, a subset of differences with fixed spatial patterns is separated from the original difference data matrix as the systematic error component. According to the time-varying characteristics of the environmental noise associated with the noise intensity coefficient, the fluctuation component that conforms to the variation period of the noise intensity coefficient is filtered out from the original difference data as the environmental interference component.

7. The three-dimensional image generation method based on spherical reflection cavity parameter optimization according to claim 6, characterized in that: The system error component is used to update the reflection path offset comparison table and / or the geometric compensation mapping relationship, which is specifically: Identify the spatial region position markers of the system error components on the cavity surface; Associating the spatial region position mark with the incident angle distribution range recorded in the reflection path offset comparison table; The statistical mean of the system error components of all pixels in the spatial area is used as the offset vector update amount corresponding to the incident angle range; Iteratively correct the three-dimensional coordinate offset vector originally recorded in the reflection path offset comparison table using the offset vector update amount; Synchronously associating the spatial region position mark with the surface coordinate node in the geometric compensation mapping relationship; Extract the mean value of the directional vector component of the system error component as the deformation compensation increment of the coordinate node corresponding to the geometric compensation mapping relationship; The deformation compensation increment is added to the original compensation value of the coordinate node in the geometric compensation mapping relationship to generate an updated geometric compensation mapping relationship.

8. The three-dimensional image generation method based on spherical reflection cavity parameter optimization according to claim 2, characterized in that: The environmental interference component is used to adjust the noise intensity coefficient, which is specifically: Extract the time-varying fluctuation amplitude peak characteristics contained in the environmental interference component, and obtain the corresponding noise intensity reference value by querying the vibration noise mapping table; The currently stored noise intensity coefficient value is called, and the noise intensity reference value is overwritten to the currently stored noise intensity coefficient value to generate a noise intensity coefficient update value.

Images