A dynamic three-dimensional reconstruction method based on high-speed plane structure light time domain analysis

Abstract

The application relates to the technical field of three-dimensional reconstruction, and discloses a dynamic three-dimensional reconstruction method based on high-speed surface structure light time domain analysis, which comprises the following steps: building a binocular measurement framework, pre-calibrating through a dynamic calibration plate, establishing a synchronization error correlation model, dynamically adjusting a trigger delay, and outputting a synchronization calibration coefficient; collecting dynamic coordinate data in a simulated motion state, introducing a distortion correction coefficient, adjusting and synchronously correcting distortion and parameter coupling error, and dynamically updating an internal parameter; projecting an initial multi-frequency phase shift grating, feeding back through stripe contrast, linkage adjusting core parameters, forming a closed-loop collection mechanism, and outputting an image sequence and core parameters suitable for a scene; combining a four-step phase shift method to solve wrapped phase, obtaining absolute phase through multi-frequency heterodyne, and generating a pseudo-judgment mask; separating pseudo-judgment and real phase, optimizing a triangulation model, removing residual pseudo-judgment, and completing point cloud registration through double-drive registration combined with dynamic precision index feedback optimization.

Description

Technical Field

[0001] This invention relates to the field of three-dimensional reconstruction technology, and specifically to a dynamic three-dimensional reconstruction method based on high-speed surface structured light temporal analysis. Background Technology

[0002] Existing high-speed surface structured light dynamic 3D reconstruction systems generally adopt a combination architecture of high-speed projectors and high-speed industrial cameras. Synchronous triggering relies on wired electrical triggering modules. By projecting grating patterns with preset fixed parameters, they use traditional multi-frequency heterodyne methods or phase-shifting methods to decompose the phase and complete 3D reconstruction by combining static calibration parameters. At the application level, they are initially used for medium- and low-speed dynamic scene measurement. However, in high-speed, complex lighting, and extreme environmental scenarios, their own technical defects make it difficult to achieve high-precision and stable reconstruction, which limits their industrialization and scenario-based promotion.

[0003] In existing technologies, wired electrical triggering is commonly used to synchronize the projector and camera via a trigger line. Wired transmission is prone to signal attenuation and delay, resulting in large synchronization errors. This fails to meet the sub-microsecond synchronization requirements of high-speed dynamic scenes. The trigger parameters are fixed and cannot be dynamically adjusted according to the object's motion, making it difficult to balance synchronization accuracy and measurement frame rate. Pursuing high accuracy requires reducing the frame rate, while increasing the frame rate exacerbates synchronization errors, creating a technical bottleneck. The wired triggering architecture increases system integration complexity and hardware deployment costs, hindering device miniaturization and flexible deployment. Existing technologies use static calibration boards for system calibration. The fixed feature point layout of the calibration board has low reflectivity, which can lead to blurred recognition under complex lighting conditions. The efficiency of inter-frame feature point matching is low, and the calibration process only collects static posture data, which cannot adapt to parameter changes in dynamic scenes. This results in the accumulation of subsequent reconstruction errors, limiting the industrialization and cross-scene application of visual inspection technology.

[0004] Therefore, there is a need to provide a dynamic 3D reconstruction method based on high-speed surface structured light temporal analysis. Summary of the Invention

[0005] The purpose of this invention is to provide a dynamic three-dimensional reconstruction method based on high-speed surface structured light temporal analysis. To solve the above-mentioned problems in the prior art, this invention achieves this through the following technical solution:

[0006] In a first aspect, the present invention provides a dynamic three-dimensional reconstruction method based on high-speed surface structured light temporal analysis, which specifically includes the following steps:

[0007] Step 1: Build a binocular measurement architecture, pre-calibrate using a dynamic calibration board, establish a synchronization error correlation model, dynamically adjust the trigger delay, and output the synchronization calibration coefficient;

[0008] Step 2: Based on the synchronous calibration coefficient, simulate motion state to collect dynamic coordinate data, introduce distortion correction coefficient, adjust and synchronously correct distortion and parameter coupling error, and dynamically update intrinsic parameters;

[0009] Step 3: Combine the initial multi-frequency phase-shift grating with dynamic intrinsic parameters, and adjust the core parameters in conjunction with the feedback of stripe contrast to form a closed-loop acquisition mechanism, and output image sequences and core parameters adapted to the scene.

[0010] Step 4: Based on the obtained image sequence and core parameters, the wrapping phase is solved by the four-step phase shift method, the absolute phase is obtained by multi-frequency heterodyne, and the phase change gradient is calculated simultaneously to generate the artifact prediction mask;

[0011] Step 5: Use the artifact prediction mask to separate artifacts from the true phase. Optimize the triangulation model using a dynamic phase compensation factor and offset prediction matrix. Remove residual artifacts using the dynamic phase compensation factor. Optimize the point cloud registration using dual-drive registration combined with dynamic accuracy feedback.

[0012] Secondly, the present invention provides a dynamic three-dimensional reconstruction system based on high-speed surface structured light temporal analysis, which specifically includes the following modules:

[0013] Synchronous calibration module: Builds a binocular measurement architecture, pre-calibrates using a dynamic calibration board, establishes a synchronization error correlation model, dynamically adjusts the trigger delay, and outputs synchronization calibration coefficients;

[0014] Correction and update module: Based on the synchronous calibration coefficient, simulate motion state to collect dynamic coordinate data, introduce distortion correction coefficient, adjust and synchronously correct distortion and parameter coupling error, and dynamically update intrinsic parameters;

[0015] Linked acquisition module: Combines dynamic intrinsic parameters to project an initial multi-frequency phase-shift grating, and adjusts core parameters in a linked manner through feedback of stripe contrast to form a closed-loop acquisition mechanism, outputting image sequences and core parameters adapted to the scene;

[0016] Prediction and labeling module: Based on the obtained image sequence and core parameters, the module calculates the wrapping phase using a four-step phase shifting method, obtains the absolute phase through multi-frequency heterodyne, and simultaneously calculates the phase change gradient to generate an artifact prediction mask;

[0017] Feedback optimization module: It calls the artifact prediction mask to separate artifacts from the true phase, optimizes the triangulation model by combining the dynamic phase compensation factor with the offset prediction matrix, removes residual artifacts by combining the dynamic phase compensation factor, and completes point cloud registration by combining dual-drive registration with dynamic accuracy index feedback optimization.

[0018] The beneficial effects of this invention are:

[0019] 1. In terms of architecture, a multi-view adaptation architecture is constructed, consisting of a single high-speed DLP projector and multiple symmetrically arranged high-speed industrial cameras. An optical trigger lamp and a high-precision optical sensor are integrated to create an optical closed-loop synchronous link for projection acquisition, eliminating the transmission delay of wires and supporting dynamic adjustment of baseline distance and grating pattern encoding logic. A standard dynamic calibration board is used to simulate the motion state of the test scene for pre-calibration, establishing a correlation model between time difference and equipment operating parameters. A dynamic delay adjustment formula with adaptive synchronization calibration coefficient is designed to achieve a dynamic balance between synchronization accuracy and frame rate, replacing the traditional fixed delay setting. A scene-adaptive calibration board is designed to adapt to extreme industrial environments and complex lighting conditions. A dynamic distortion correction coefficient related to image size is introduced and pixel-level dynamic adjustment is achieved. A dynamic coupling relationship between the distortion correction coefficient and the intrinsic parameter matrix is ​​constructed to eliminate multi-parameter coupling interference. At the same time, a real-time update mechanism for intrinsic parameters that dynamically adjusts the frequency according to the intensity of motion is designed.

[0020] 2. By combining dynamic intrinsic parameters to achieve deep binding between grating projection and equipment calibration status, a fringe contrast quantization feedback mechanism is established. A nonlinear linkage adjustment formula for grating frequency and exposure time, along with a stepwise adaptation strategy for phase shift step size, is designed to form a contrast-driven closed-loop adaptive acquisition mechanism, achieving real-time dynamic adaptation of acquisition parameters to the measurement scene. A phase shift correction coefficient dynamically adjusted with fringe contrast is introduced, and combined with the dynamic phase shift step size to correct phase shift deviations caused by motion, the three-frequency heterodyne method absolute phase solution formula is optimized to balance solution accuracy and dynamic feature capture. This is achieved by calculating the absolute phase between frames. A mechanism for generating artifact prediction masks using varying gradients is proposed, marking potential artifact regions and recording phase change trends to enable early intervention in artifact processing. Temporal context interpolation is employed to accurately separate artifacts from the true phase. A dynamic phase compensation factor is introduced and combined with the offset prediction matrix to optimize the triangulation model, simultaneously correcting motion errors and point cloud offsets. An adaptive temporal filter bound to the dynamic phase compensation factor is designed to achieve differentiated removal of artifact residues. A dual-drive point cloud registration strategy is proposed, combined with a feedback mechanism for dynamic accuracy indicators, to achieve dynamic adjustment of registration weights and closed-loop readjustment of preceding parameters. Attached Figure Description

[0021] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0022] Figure 1 This is a flowchart of the steps of a dynamic three-dimensional reconstruction method based on high-speed surface structured light temporal analysis provided in Embodiment 1 of the present invention;

[0023] Figure 2This is a schematic diagram of a dynamic three-dimensional reconstruction system based on high-speed surface structured light temporal analysis provided in Embodiment 2 of the present invention. Detailed Implementation

[0024] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0025] Example 1: As Figure 1 As shown in the figure, the dynamic three-dimensional reconstruction method based on high-speed surface structured light temporal analysis provided by the embodiments of the present invention specifically includes the following steps:

[0026] Step 1: Build a binocular measurement architecture, pre-calibrate using a dynamic calibration board, establish a synchronization error correlation model, dynamically adjust the trigger delay, and output the synchronization calibration coefficient;

[0027] In a specific embodiment, a binocular measurement architecture is built using a single high-speed DLP projector and multiple high-speed industrial cameras. The cameras are symmetrically arranged on both sides of the projector. The baseline distance can be dynamically adjusted according to the measurement scenario to reduce blind spots. The integrity of the reconstruction of complex shapes is improved by complementary data from multiple perspectives. At the same time, the projector has the ability to dynamically generate multi-frequency phase-shift gratings and adjust the pattern encoding logic according to the requirements to adapt to the measured objects with different dynamic characteristics.

[0028] An optical trigger lamp is integrated at the output end of a high-speed DLP projector, and a high-precision optical sensor is installed at each high-speed industrial camera lens to build a closed-loop synchronous link that triggers optical sensing and starts data acquisition, eliminating signal attenuation and delay caused by wire transmission and simplifying system integration complexity.

[0029] Before data acquisition, a standard dynamic calibration board is used for pre-calibration. The calibration board is controlled to simulate the typical motion state of the scene under test, and the time difference between each projection trigger moment and the camera exposure start moment is monitored synchronously.

[0030] By combining the multi-frame data collected during pre-calibration, a correlation model between time difference and equipment operating parameters is established. The trigger delay is corrected in real time through dynamic delay adjustment, and the synchronization error is controlled within the preset range.

[0031] The camera's built-in high-precision timer and projector controller record the exposure start time and grating projection start time respectively. The time difference is calculated frame by frame. Multiple frames of pre-calibration data are collected, the average value is taken, and the trigger delay is adjusted for analysis to ensure the stability of the calibration results.

[0032] Specifically, through the dynamic delay adjustment formula: Analysis yielded the corrected trigger delay ,in, For the initial trigger delay, To record the time difference, The synchronization calibration coefficient is dynamically adapted according to the camera response speed. The more intense the scene, the larger the value of the synchronization calibration coefficient. The dynamic balance between synchronization accuracy and frame rate is achieved through coefficient adaptation. The calibration coefficient is dynamically updated according to the motion characteristics of the object being measured, avoiding the error accumulation of static calibration methods in high-speed scenes. The inherent delay of the device and the real-time measurement error are combined to replace the fixed delay setting. A single adjustment of the synchronization calibration coefficient optimizes both synchronization accuracy and frame rate adaptability.

[0033] Step 2: Based on the synchronous calibration coefficient, simulate motion state to collect dynamic coordinate data, introduce distortion correction coefficient, adjust and synchronously correct distortion and parameter coupling error, and dynamically update intrinsic parameters;

[0034] In a specific embodiment, a dynamic calibration board is made of scene-adaptive materials. High-temperature resistant materials are selected for high-temperature industrial scenarios to avoid deformation and failure of traditional calibration boards in extreme environments. A special high-reflectivity coating is sprayed on the surface of feature points to improve the recognition stability of feature points under complex lighting conditions, realize rapid matching of feature points between frames, and adapt to different measurement field of view requirements.

[0035] The control calibration board simulates the typical motion state of the object under test and collects dynamic coordinate data, including pixel coordinates and actual three-dimensional coordinates. A high-speed DLP projector projects a preset low-frequency phase-shift grating pattern, and a high-speed industrial camera simultaneously acquires multiple frames of calibration images. During the acquisition process, the pixel coordinates and corresponding actual three-dimensional coordinates of feature points in each frame are recorded in real time to construct a dynamic coordinate dataset. By simulating motion state, the calibration error in dynamic scenes is reduced.

[0036] By combining the collected dynamic coordinate dataset, a dynamic intrinsic parameter model is established, and distortion correction coefficients are introduced to specifically correct the nonlinear distortion of the camera and projector.

[0037] A frame-by-frame feature point matching method combined with dynamic filtering and error correction mechanisms is adopted. By combining multiple frames of acquired calibration images, the pixel coordinates of feature points on the calibration board are extracted frame by frame. During the extraction process, feature point filtering criteria are constructed based on the gray-level gradient and neighborhood consistency of feature points to automatically remove fuzzy feature points. For each candidate feature point, the variance of gray-level values ​​in the neighborhood range of the candidate feature point is calculated. If the variance is lower than a set threshold, it is determined to be a fuzzy point and is directly removed. By using the geometric distribution characteristics of feature points, abnormal points that deviate from the layout of the calibration board array are removed to ensure that the retained feature points are clear and effective target points.

[0038] For the selected valid feature points, the deviation between their actual pixel coordinates and ideal pixel coordinates is calculated point by point. The corresponding deviation values ​​are substituted into the calculation formula of the dynamic distortion correction coefficient to generate the distortion correction coefficient for each frame in real time. The pixel-level dynamic adjustment of the distortion correction coefficient is realized for the deviation differences of different feature points.

[0039] By using dynamically generated distortion correction coefficients, the intrinsic parameter matrices of the camera and projector are simultaneously corrected. The distortion correction coefficients compensate for the nonlinear distortion error of the optical system and correct the distortion deviation of the feature point pixel coordinates. By combining the dynamic coupling relationship between the distortion correction coefficients and the intrinsic parameter matrices, the parameter coupling interference between the camera intrinsic parameters, the projector intrinsic parameters and the system extrinsic parameters is eliminated.

[0040] Specifically, a dynamic update mechanism is constructed to dynamically update the distortion correction coefficient by calculating the deviation between the actual pixel coordinates and the ideal pixel coordinates of the feature points in real time, thereby compensating for the dynamic distortion of optical components under high-speed motion.

[0041] Through formula Analysis yields distortion correction coefficients ,in, These are the actual pixel coordinates of the collected feature points. For the ideal pixel coordinates of the feature point, This represents the row and column number of the pixel in the image coordinate system. The distortion deviation is correlated with the image size to the camera image width, allowing the correction coefficient to adapt to cameras with different resolutions. Simultaneously, the distortion correction coefficient... Dynamic coupling with the intrinsic parameter matrix solves the problems of parameter coupling and dynamic distortion simultaneously, avoiding the accumulation of errors caused by the independence of distortion correction and parameter solution;

[0042] The distortion correction coefficient is used to correct the distortion terms in the camera intrinsic parameter matrix. At the same time, the intrinsic parameter update frequency is dynamically adjusted in combination with the movement speed of the calibration board. The more intense the movement, the higher the intrinsic parameter update frequency, ensuring that the intrinsic parameter parameters match the object's motion state in real time and solving the dynamic error caused by fixed parameters in static calibration. The dynamic update mechanism does not require interruption of the measurement process and completes the intrinsic parameter optimization simultaneously during the reconstruction process, balancing real-time performance and calibration accuracy.

[0043] Step 3: Combine the initial multi-frequency phase-shift grating with dynamic intrinsic parameters, and adjust the core parameters in conjunction with the feedback of stripe contrast to form a closed-loop acquisition mechanism, and output image sequences and core parameters adapted to the scene.

[0044] In a specific embodiment, the high-speed industrial camera continuously acquires image sequences according to the adjusted core parameters, periodically recalculates the contrast and dynamically updates the core parameters, forming a closed-loop adaptive mechanism for controlling acquisition. During the acquisition process, the grating frequency, phase shift step size and exposure time of each frame are recorded synchronously.

[0045] An initial multi-frequency phase-shift grating pattern is projected onto the surface of the object under test using a high-speed DLP projector. A multi-frequency combination design is adopted to balance the phase expansion range and accuracy. A high-speed industrial camera simultaneously acquires multiple frames of images with the initial exposure time. The initial core parameter settings are combined with general scene optimization to reduce the complexity of operation.

[0046] The stripe contrast is calculated frame by frame for the initial images being acquired. The contrast quantifies the stripe sharpness and the degree of environmental interference, using the formula: Analysis yielded stripe contrast ,in, This represents the maximum grayscale value of the stripes in a single frame of the image. The minimum grayscale value of the stripes in a single frame image is used for quantization feedback mechanism to improve adaptation accuracy;

[0047] Set a contrast threshold range, and trigger different control strategies based on the calculated relationship between the stripe contrast and the threshold range: if the stripe contrast is less than or equal to the preset lower limit of the contrast threshold, the grating frequency and exposure time are adjusted in conjunction with the formula, while the phase shift step size is reduced to improve stripe clarity; if the stripe contrast is greater than or equal to the upper limit of the contrast threshold, the grating frequency and exposure time are adjusted in the opposite direction, and the phase shift step size is increased to avoid overexposure; if the stripe contrast is less than the upper limit of the contrast threshold but greater than the lower limit of the contrast threshold, no adjustment is made.

[0048] Through the formula: Analysis yielded the grating adjustment frequency With exposure adjustment time ,in, The initial grating frequency, The standard contrast threshold, For stripe contrast, This is the initial exposure time;

[0049] A non-linear relationship between stripe contrast and core parameters is established, and the adjustment range of core parameters is matched with the contrast deviation to avoid stripe distortion caused by over-adjustment. At the same time, through the coordinated control of grating frequency and exposure time, the problem of ambient light interference and reflection adaptation is solved simultaneously in a single adjustment. The core parameters include: grating frequency, exposure time and phase shift step.

[0050] For example, simultaneously reducing the phase shift step size to To improve stripe clarity; if the stripe contrast is greater than 0.7, it indicates that the object is highly reflective, so reduce the raster adjustment frequency to 0.8 times the original value, shorten the exposure adjustment time to 0.6 times the original value, and increase the phase shift step size to... To avoid overexposure; the adjusted grating frequency range is [12 line pairs / inch, 25 line pairs / inch], and the exposure time range is [20μs, 80μs];

[0051] Step 4: Based on the obtained image sequence and core parameters, the wrapping phase is solved by the four-step phase shift method, the absolute phase is obtained by multi-frequency heterodyne, and the phase change gradient is calculated simultaneously to generate the artifact prediction mask;

[0052] In a specific embodiment, the acquired multi-frequency phase-shift image sequence is combined to calculate the wrapping phase and absolute phase, remove abnormal phase-jump pixels, and use adjacent valid pixels for interpolation to complete the image; high-risk areas are marked by artifact prediction masks to distinguish between real phase changes and phase distortion caused by motion artifacts.

[0053] For each frame of image, a four-step phase shift method is used to calculate the wrapping phase. A preset phase shift correction coefficient is introduced, and combined with a dynamically adjusted phase shift step size, the phase shift deviation caused by the object's motion is corrected in real time. The correction intensity is dynamically adjusted according to the image contrast, and the wrapping phase calculation is combined to adapt to changes in the object's motion state, reducing phase distortion caused by motion.

[0054] Specifically, the formula is calculated using the four-step phase-shifting method: Analysis yields the package phase ,in, These are the grayscale values ​​of the corresponding pixels in the four-step phase-shifted image. For the phase shift step size, The phase shift correction factor is a preset value. The phase shift correction factor is determined based on the fringe contrast. The smaller the fringe contrast, the larger the phase shift correction factor.

[0055] The three-frequency heterodyne method is adopted. The equivalent frequency is calculated using the three adjusted grating frequencies. The phase expansion range is narrowed by the equivalent frequency, the absolute phase solution formula is optimized, phase ambiguity is eliminated, and the inter-frame phase change characteristics are preserved.

[0056] Specifically, through the formula: Analysis yields continuous absolute phase values ,in, For frequency The envelope phase obtained after grating projection. Pi For the floor function, For frequency The enclosed phase obtained after grating projection. The grating frequency of the original grating 1, The grating frequency of the original grating 2, The equivalent frequencies of the two frequencies;

[0057] By combining the continuous absolute phase values ​​of the obtained consecutive frames, the phase change gradient between the current frame and the previous frame is calculated. ,in, This is the inter-frame time interval. Set a gradient threshold for the continuous absolute phase values ​​of the previous frame. If the phase change gradient is greater than the gradient threshold, the corresponding region is determined to be a high motion risk region and marked as a potential artifact region. At the same time, the phase change trend of the corresponding region is recorded to form an artifact prediction mask, which is transmitted synchronously to realize early intervention in artifact processing and avoid deep mixing of artifacts with the real shape. Otherwise, it is not marked.

[0058] Step 5: Use the artifact prediction mask to separate artifacts from the true phase, optimize the triangulation model by combining the dynamic phase compensation factor with the offset prediction matrix, remove residual artifacts by combining the dynamic phase compensation factor, and complete the point cloud registration by combining dual-drive registration with dynamic accuracy index feedback optimization.

[0059] In a specific implementation, the generated artifact prediction mask is invoked to determine the absolute phase of the current frame. The process involves partitioning the area and using temporal context interpolation to construct a phase recovery model by combining the normal phase data from the previous two frames and the next frame. This model removes the phase components corresponding to artifacts and restores the true phase information.

[0060] For normal regions, absolute phase data is directly retained to achieve accurate separation of artifacts from the real shape, avoiding the problem of traditional filtering methods accidentally deleting real features or leaving artifacts.

[0061] For the absolute phase data after separating residual artifacts, recalculate the inter-frame phase change. By combining the motion characteristics of objects, the correlation between phase change and point cloud offset is established, the offset direction and magnitude of the current frame point cloud relative to the previous frame are predicted, and an offset prediction matrix is ​​generated.

[0062] Introducing dynamic phase compensation factor ,in, The preset phase change threshold is negatively correlated with the amount of phase change between frames. The more intense the motion, the smaller the dynamic phase compensation factor and the greater the compensation force. At the same time, combined with the offset prediction matrix, the triangulation model is improved to simultaneously complete motion error compensation and point cloud offset correction.

[0063] Specifically, the calculation formula is based on three-dimensional coordinates: , , ,in, The baseline distance of the binocular cameras. , The pixel coordinates of the current frame. , Let these be the coordinates of the camera's principal point. This refers to the projection angle of a high-speed DLP projector. , These are the three-dimensional coordinates of the corresponding pixel in the previous frame. , , The corresponding offset in the offset prediction matrix is ​​used to synchronously correct motion errors and point cloud offsets using formulas.

[0064] By combining the 3D point cloud data after separating artifacts, a time-domain filter dynamically bound to the dynamic phase compensation factor is designed for adaptive adjustment. A larger window is used to smooth artifact residue in areas with violent motion, while a smaller window is used to preserve details and remove residual artifacts in normal areas.

[0065] A dual-drive strategy of phase features and geometric constraints is adopted to extract the absolute phase features corresponding to the corrected point cloud, construct an inter-frame phase feature correlation map, and combine the spatiotemporal continuity of phase features to initially establish the point cloud correspondence and eliminate obvious mismatches.

[0066] By introducing geometric constraints and combining the dynamic intrinsic and extrinsic parameters of the camera and projector with the spatial distribution of the three-dimensional coordinates of the point cloud, the preliminary matching results are verified and optimized. Erroneous corresponding points that deviate from the geometric constraints are eliminated, the matching relationship between point clouds between frames is accurately determined, and the point cloud registration is completed.

[0067] Establish dynamic accuracy indicators, combine standard reference coordinates and registration errors, quantify reconstruction and registration accuracy, and use formulas. Analysis yields dynamic accuracy values ,in, Reconstruct the Z coordinate for the current frame. As the standard reference coordinate, The registration error value is set. If the dynamic accuracy value exceeds the set threshold, the core parameter readjustment mechanism is triggered to optimize the grating projection, phase resolution, and compensation core parameters. The weights of the dual-drive strategy are dynamically adjusted according to the registration error, adjusting the geometric constraint weights and phase feature weights to ensure a balance between registration accuracy and efficiency. For example, the larger the registration error, the larger the geometric constraint weights, and the smaller the registration error, the smaller the phase feature weights, and the value does not exceed the preset adjustment range.

[0068] Example 2: Figure 2 As shown in the figure, the dynamic three-dimensional reconstruction system based on high-speed surface structured light temporal analysis provided by the embodiment of the present invention specifically includes the following modules:

[0069] Synchronous calibration module: Builds a binocular measurement architecture, pre-calibrates using a dynamic calibration board, establishes a synchronization error correlation model, dynamically adjusts the trigger delay, and outputs synchronization calibration coefficients;

[0070] Correction and update module: Based on the synchronous calibration coefficient, simulate motion state to collect dynamic coordinate data, introduce distortion correction coefficient, adjust and synchronously correct distortion and parameter coupling error, and dynamically update intrinsic parameters;

[0071] Linked acquisition module: Combines dynamic intrinsic parameters to project an initial multi-frequency phase-shift grating, and adjusts core parameters in a linked manner through feedback of stripe contrast to form a closed-loop acquisition mechanism, outputting image sequences and core parameters adapted to the scene;

[0072] Prediction and labeling module: Based on the obtained image sequence and core parameters, the module calculates the wrapping phase using a four-step phase shifting method, obtains the absolute phase through multi-frequency heterodyne, and simultaneously calculates the phase change gradient to generate an artifact prediction mask;

[0073] Feedback optimization module: It calls the artifact prediction mask to separate artifacts from the true phase, optimizes the triangulation model by combining the dynamic phase compensation factor with the offset prediction matrix, removes residual artifacts by combining the dynamic phase compensation factor, and completes point cloud registration by combining dual-drive registration with dynamic accuracy index feedback optimization.

[0074] The above provides a detailed description of one embodiment of the present invention, but the content described is only a preferred embodiment of the present invention and should not be considered as limiting the scope of the present invention. The above formulas are all dimensionless numerical calculations, and the formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world situation. The preset parameters in the formulas are set by those skilled in the art based on actual conditions and historical experience, and can be adjusted according to actual conditions. The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. All equivalent changes and improvements made in accordance with the scope of the present invention should still fall within the patent coverage of the present invention.

Claims

1. A dynamic three-dimensional reconstruction method based on high-speed surface structured light temporal analysis, characterized in that, Includes the following steps: A binocular measurement architecture is built, and a synchronization error correlation model is established through pre-calibration using a dynamic calibration board. The trigger delay is dynamically adjusted, and the synchronization calibration coefficient is output. Based on the synchronous calibration coefficient, dynamic coordinate data is collected by simulating motion state. A distortion correction coefficient is introduced to adjust and synchronously correct the distortion and parameter coupling error, and the intrinsic parameters are dynamically updated. By combining the dynamic intrinsic parameter projection of the initial multi-frequency phase-shift grating and adjusting the core parameters in conjunction with the feedback of stripe contrast, a closed-loop acquisition mechanism is formed, and the image sequence and core parameters adapted to the scene are output. Based on the obtained image sequence and core parameters, the wrapping phase is solved by the four-step phase shift method, the absolute phase is obtained by multi-frequency heterodyne, and the phase change gradient is calculated simultaneously to generate the artifact prediction mask. The artifact prediction mask is called to separate the artifacts from the true phase. The triangulation model is optimized by combining the dynamic phase compensation factor with the offset prediction matrix. The residual artifacts are removed by combining the dynamic phase compensation factor. The point cloud registration is completed by combining dual-drive registration with dynamic accuracy index feedback optimization.

2. The dynamic three-dimensional reconstruction method based on high-speed surface structured light temporal analysis according to claim 1, characterized in that, The method for dynamically adjusting the trigger delay is as follows: Before data acquisition, a standard dynamic calibration board is used for pre-calibration. The calibration board is controlled to simulate the typical motion state of the scene under test, and the time difference between each projection trigger moment and the camera exposure start moment is monitored synchronously. By combining the multi-frame data collected during pre-calibration, a correlation model between time difference and equipment operating parameters is established. The trigger delay is corrected in real time through dynamic delay adjustment, and the synchronization error is controlled within the preset range. The camera's built-in high-precision timer and projector controller record the exposure start time and grating projection start time respectively. The time difference is calculated frame by frame, and the average value of multiple frames of pre-calibration data is used for analysis to adjust the trigger delay.

3. The dynamic three-dimensional reconstruction method based on high-speed surface structured light temporal analysis according to claim 1, characterized in that, The method for obtaining the distortion correction coefficient is as follows: A frame-by-frame feature point matching method combined with dynamic filtering and error correction mechanism is adopted. By combining the acquired multi-frame calibration images, the pixel coordinates of the feature points of the calibration board are extracted frame by frame. During the extraction process, based on the gray-level gradient and neighborhood consistency of the feature points, a feature point filtering criterion is constructed to automatically remove fuzzy feature points. For each of the selected valid feature points, the deviation between its actual pixel coordinates and ideal pixel coordinates is calculated. The corresponding deviation value is then substituted into the calculation formula of the dynamic distortion correction coefficient to generate the distortion correction coefficient for each frame in real time. Based on the deviation differences of different feature points, the distortion correction coefficient is dynamically adjusted at the pixel level.

4. The dynamic three-dimensional reconstruction method based on high-speed surface structured light temporal analysis according to claim 1, characterized in that, The method for dynamically updating internal parameters is as follows: The distortion term in the camera intrinsic parameter matrix is ​​corrected by using distortion correction coefficients. At the same time, the intrinsic parameter update frequency is dynamically adjusted in combination with the calibration board's motion speed. The intrinsic parameters are matched with the object's motion state in real time, eliminating the dynamic error caused by fixed parameters in static calibration. Intrinsic parameter optimization is completed synchronously during the reconstruction process.

5. The dynamic three-dimensional reconstruction method based on high-speed surface structured light temporal analysis according to claim 1, characterized in that, The method for adjusting the core parameters in a coordinated manner is as follows: The stripe contrast is calculated frame by frame for the initial image being acquired. The stripe sharpness and the degree of environmental interference are quantified by the contrast, and the stripe contrast is analyzed to obtain the result. Set a contrast threshold range, and trigger different control strategies based on the calculated relationship between the stripe contrast and the threshold range: if the stripe contrast is less than or equal to the preset lower limit of the contrast threshold, the grating frequency and exposure time are adjusted in conjunction with the formula, while the phase shift step size is reduced to improve stripe clarity. If the stripe contrast is greater than or equal to the upper limit of the contrast threshold, the grating frequency and exposure time are adjusted in reverse, and the phase shift step is increased to avoid overexposure. If the stripe contrast is less than the upper limit of the contrast threshold but greater than the lower limit of the contrast threshold, no adjustment is made. Through the formula: Analysis yielded the grating adjustment frequency With exposure adjustment time ,in, The initial grating frequency, The standard contrast threshold, For stripe contrast, This is the initial exposure time; A non-linear relationship between stripe contrast and core parameters is established, and the adjustment range of core parameters is adapted to the contrast deviation through the coordinated control of grating frequency and exposure time.

6. The dynamic three-dimensional reconstruction method based on high-speed surface structured light temporal analysis according to claim 1, characterized in that, The method for forming a closed-loop acquisition mechanism is as follows: An initial multi-frequency phase-shifting grating is projected onto the surface of the object under test using dynamic intrinsic parameters, and a high-speed industrial camera synchronously acquires image sequences according to the initial parameters. During continuous acquisition by the camera, the stripe contrast is periodically recalculated, and the core parameters are dynamically updated and adjusted in conjunction with the contrast using the above method. During the acquisition process, the grating frequency, phase shift step size, and exposure time of each frame are recorded simultaneously, forming a closed-loop adaptive mechanism for contrast-driven parameter control and image acquisition.

7. The dynamic three-dimensional reconstruction method based on high-speed surface structured light temporal analysis according to claim 1, characterized in that, The method for obtaining the absolute phase is as follows: The formula is calculated using the four-step phase-shifting method: Analysis yields the package phase ,in, These are the grayscale values ​​of the corresponding pixels in the four-step phase-shifted image. For the phase shift step size, This is the preset phase shift correction coefficient; The three-frequency heterodyne method is employed. Using the adjusted three grating frequencies, the equivalent frequency is calculated. The phase expansion range is then narrowed using the equivalent frequency, as shown by the formula: Analysis yields continuous absolute phase values ,in, For frequency The enclosed phase obtained after grating projection. Pi For the floor function, For frequency The enclosed phase obtained after grating projection. The grating frequency of the original grating 1, The grating frequency of the original grating 2, The equivalent frequencies of the two frequencies are used to eliminate phase ambiguity while preserving the inter-frame phase change characteristics.

8. The dynamic three-dimensional reconstruction method based on high-speed surface structured light temporal analysis according to claim 1, characterized in that, The method for generating the artifact prediction mask is as follows: By combining the continuous absolute phase values ​​of the obtained consecutive frames, the phase change gradient between the current frame and the previous frame is calculated. ,in, This is the inter-frame time interval. Set a gradient threshold for the continuous absolute phase values ​​of the previous frame. If the phase change gradient is greater than the gradient threshold, the corresponding region is determined to be a high motion risk region and marked as a potential artifact region. The phase change trend of the corresponding region is recorded to form an artifact prediction mask.

9. The dynamic three-dimensional reconstruction method based on high-speed surface structured light temporal analysis according to claim 1, characterized in that, The method for optimizing the triangulation model is as follows: The absolute phase change between frames is calculated, a dynamic phase compensation factor is introduced, and a displacement prediction matrix containing the offset is generated by combining the correlation between phase change and point cloud offset. The dynamic phase compensation factor and the displacement prediction matrix are then substituted into the three-dimensional coordinate calculation formula. , , Optimize the triangulation model, where, The baseline distance of the binocular cameras. , The pixel coordinates of the current frame. , Let these be the coordinates of the camera's principal point. This refers to the projection angle of a high-speed DLP projector. , These are the three-dimensional coordinates of the corresponding pixel in the previous frame. , , The corresponding offset in the offset prediction matrix is ​​used to simultaneously correct motion errors and point cloud offsets.

10. The dynamic three-dimensional reconstruction method based on high-speed surface structured light temporal analysis according to claim 1, characterized in that, The feedback optimization method is as follows: Extract the absolute phase features of the corrected point cloud, combine them with geometric constraints to complete point cloud registration, and then use the formula... Calculate dynamic accuracy value ,in Reconstruct the Z coordinate for the current frame. As the standard reference coordinate, The registration error value, This is the dynamic phase compensation factor; If the dynamic accuracy value exceeds the set threshold, the pre-processor grating projection, de-phase, and compensation parameter readjustment mechanism is triggered; the weights of the dual-drive registration strategy of phase characteristics and geometric constraints are dynamically adjusted according to the registration error to balance registration accuracy and efficiency.

Images