Monocular telecentric structured light three-dimensional reconstruction optimization and depth of field expansion method

Abstract

The application discloses a monocular telecentric structured light three-dimensional reconstruction optimization and depth of field expansion method, which comprises the following steps: controlling a displacement table to drive a standard ceramic flat plate to step along the height direction, collecting structured light stripes and solving unwrapping phase at each height; dividing the camera field of view into several regions to form a regional optimal reconstruction height lookup table; when measuring an object to be measured, firstly, preliminary three-dimensional reconstruction is carried out, then, the required compensation displacement of each region is calculated in combination with the lookup table, and the acquisition times of the displacement table are reduced through height layering clustering; local high-quality point clouds are collected and reconstructed at each compensation height, finally, the local point clouds are unified to the same coordinate system according to the actual displacement amount of the displacement table and the calibration correction amount of the standard flat plate, and the local point clouds are fused in the overlapping region according to the reconstruction quality weight. The application can compensate the regional optimal reconstruction height difference caused by the oblique incidence of the projector, improve the local phase stability and reconstruction accuracy, and effectively expand the depth of field range of the system.

Description

Technical Field

[0001] This invention belongs to the field of structured light 3D measurement technology, specifically relating to a monocular telecentric structured light 3D reconstruction optimization and depth-of-field extension method. Background Technology

[0002] Structured light 3D measurement technology projects coded stripes onto the surface of an object using a projector, and a camera captures images of the deformed stripes modulated by the object's surface topography. The 3D topography of the object's surface is then obtained through phase calculation, phase unwrapping, and system calibration. This technology offers advantages such as non-contact operation, high measurement speed, rich surface shape information, and suitability for automated inspection, and has wide applications in industrial parts inspection, precision manufacturing, microstructure measurement, and surface topography analysis.

[0003] In high-precision inspection scenarios, telecentric lenses are commonly used to construct monocular telecentric structured light measurement systems due to their near-constant magnification, low distortion, and insensitivity to changes in object distance. A typical arrangement involves the telecentric camera's optical axis being perpendicular to the plane of the displacement stage or the reference plane to be measured. The projector projects structured light fringes onto the surface at an angle to create the triangulation angles required for 3D measurement. This structure maintains stable imaging scale at the camera end; however, the oblique projection of the projector causes variations in the distance from the projector, the angle of incidence, and the focal plane state across different areas within the same field of view, leading to spatial differences in fringe sharpness, fringe modulation, and phase resolution quality.

[0004] Existing monocular structured light measurement methods typically perform calibration and reconstruction at a fixed working distance or measurement height. When the height of the object under test varies significantly, or when the projector is angled, resulting in different optimal projection sharpness heights at different positions in the field of view, a single measurement height is insufficient to guarantee high-quality reconstruction across the entire field of view simultaneously. In some areas, phase noise may increase due to projection defocus or decreased fringe modulation; in other areas, although the image brightness is high, the unwrapping phase may exhibit local nonlinearity, glitches, or unfolding errors, leading to unstable reconstructed surface shapes. Therefore, relying solely on a single reconstruction at a fixed height easily results in problems such as insufficient reconstruction accuracy in local areas, increased noise in edge regions, and limited effective depth of field.

[0005] To extend depth of field, existing technologies include multi-focal plane acquisition, mechanical scanning, image sharpness evaluation, and multi-exposure fusion. However, most of these methods rely on two-dimensional image gradients, brightness contrast, or edge sharpness as evaluation criteria, which cannot directly characterize the quality of phase resolution during structured light 3D reconstruction. For a standard plane, under ideal reconstruction conditions, the unwrapped phase should possess good linear or planar characteristics within a local region. Therefore, the straightness of the unwrapped phase is closer to the final 3D reconstruction accuracy than simple image sharpness.

[0006] Furthermore, even if optimal regions at different heights are obtained, if data is still acquired region by region and height by height, the number of stage movements and image acquisitions will increase significantly, affecting measurement efficiency. If the final point cloud stitching is simply based on the nominal displacement of the stage, it is also prone to step, misalignment, and overlapping noise due to stage repetitive positioning errors, local reconstruction noise, or hard switching of region boundaries. Therefore, it is necessary to further establish a more stable quality evaluation, hierarchical acquisition, and quality weight fusion mechanism without changing the basic hardware layout of the telecentric camera, projector, and stage. Summary of the Invention

[0007] The purpose of this invention is to provide a method for optimizing and extending the depth of field in monocular telecentric structured light 3D reconstruction. This method addresses problems in existing monocular telecentric structured light measurement systems under oblique projector illumination conditions, such as uneven reconstruction quality within the field of view, inability to achieve optimal reconstruction across the entire area at a fixed measurement height, limited depth of field, and discontinuous boundaries in multi-height point cloud stitching. Without altering the equipment or basic workflow, this method improves local phase stability and final 3D reconstruction accuracy through multi-height table construction on a standard ceramic plate, comprehensive quality evaluation, effective depth of field interval determination, coarse reconstruction guidance of the object under test, compensated height hierarchical clustering, and quality weighted point cloud fusion.

[0008] To achieve the above objectives, the present invention adopts the following technical solution: a method for monocular telecentric structured light 3D reconstruction optimization and depth-of-field extension, comprising the following steps:

[0009] Step 1: Build and calibrate a monocular telecentric structured light 3D measurement system. The system includes a telecentric camera, a projector, a displacement stage, a standard ceramic plate, and a data processing unit. The optical axis of the telecentric camera is perpendicular to the motion plane of the displacement stage. The projector projects structured light stripes onto the standard ceramic plate or the object to be measured on the displacement stage at an angle relative to the telecentric camera. Step 2: Place the standard ceramic plate on the displacement stage, control the displacement stage to move in steps along the height direction according to the preset step size, and acquire structured light stripe images at each height position. Step 3: Perform phase calculation and phase unwrapping on the structured light fringe images acquired at each height position to obtain the unwrapped phase distribution of the standard ceramic plate at different height positions; Step 4: Divide the field of view of the telecentric camera into multiple local regions, and calculate the reconstruction quality evaluation value of each local region at different height positions. The reconstruction quality evaluation value includes at least the unwrapped phase straightness evaluation value. Step 5: Based on the reconstruction quality evaluation values ​​of the same local area at different height positions, determine the optimal reconstruction height of the local area, determine the effective depth of field range of the local area, and establish a lookup table for the optimal reconstruction height of the area. Step 6: Place the object to be tested on the displacement stage and perform preliminary three-dimensional reconstruction at the initial height position to obtain the rough height distribution of the surface of the object to be tested. Step 7: Based on the rough height distribution and the regional optimal reconstruction height lookup table, calculate the displacement stage compensation amount required for different regions of the object under test, and perform height stratification or clustering on the displacement stage compensation amount to obtain several representative compensation heights. Step 8: Control the displacement stage to move sequentially to the representative compensation height, acquire structured light stripe images at each representative compensation height and perform local three-dimensional reconstruction to obtain multiple local high-quality point clouds; Step 9: Based on the actual displacement of the displacement stage and the displacement correction obtained from the standard ceramic plate calibration, the multiple local high-quality point clouds are transformed to a unified coordinate system, and the overlapping areas are fused according to the reconstruction quality weight to obtain the complete three-dimensional reconstruction result of the object under test.

[0010] Furthermore, the reconstruction quality evaluation value also includes one or more of the following: fringe modulation evaluation value, planar reconstruction residual evaluation value, local phase smoothness evaluation value, and repeated measurement stability evaluation value, which are combined by weighting to form a comprehensive quality evaluation function.

[0011] Furthermore, the comprehensive quality evaluation function is: ,in This indicates the phase straightness evaluation item. This indicates the evaluation item for the stripe tone system. This represents the residual evaluation term of the standard ceramic plate reconstruction result relative to the fitted plane. This indicates a local phase or high smoothness evaluation term. This represents the standard deviation of height or phase obtained from repeated acquisitions. , , , and These are the weighting coefficients.

[0012] Furthermore, the optimal reconstruction height lookup table for the region is a pixel-level lookup table, a block-level lookup table, or a region-level lookup table, and the local region is a rectangular grid region, a strip region divided along the oblique direction of the projector, or an irregular region adaptively generated according to the stripe quality distribution.

[0013] Furthermore, the effective depth of field range refers to the height range within which the reconstruction quality evaluation value of a local area is higher than a preset threshold or the decrease in reconstruction quality relative to the optimal value is less than a preset proportion. The effective depth of field range is used to determine whether the local surface of the object under test needs to be compensated by a displacement stage.

[0014] Furthermore, the high degree of hierarchical or clustering includes equal-spaced binning, K-means clustering, hierarchical clustering, or threshold-based merging of the required displacement stage compensation for different regions, in order to reduce the number of displacement stage movements and structured light image acquisitions.

[0015] Furthermore, when performing local 3D reconstruction at each representative compensation height, a region mask is generated based on the region's optimal reconstruction height lookup table, the preliminary reconstruction height of the object to be measured, and the current displacement stage compensation height, retaining only the local point cloud within the optimal reconstruction height or effective depth of field range.

[0016] Furthermore, when the local high-quality point cloud is transformed to a unified coordinate system, the actual displacement is obtained based on the encoder reading of the displacement stage, the nominal step displacement, or the feedback from the displacement sensor, and coordinate compensation is performed by combining the displacement correction obtained by the multi-height reconstruction plane fitting of the standard ceramic plate.

[0017] Furthermore, the fusion methods for the overlapping regions include weighted averaging based on reconstruction quality weights, selective retention based on the maximum reconstruction quality, gradual fusion by setting an overlapping transition zone at the region boundary, or outlier removal, hole filling, and smoothing of the fused point cloud.

[0018] A monocular telecentric structured light 3D reconstruction optimization and depth-of-field extension system based on regional optimal reconstruction height includes a telecentric camera, a projector, a displacement stage, a standard ceramic plate, and a processor, wherein the processor is configured to execute the above-described monocular telecentric structured light 3D reconstruction optimization and depth-of-field extension method.

[0019] Compared with the prior art, the present invention has the following beneficial effects:

[0020] (1) This invention directly uses the unwrapping phase straightness in the structured light reconstruction process to evaluate the local reconstruction state. Compared with the simple two-dimensional image sharpness evaluation, it can better reflect the phase unfolding and three-dimensional reconstruction quality.

[0021] (2) In view of the problem of uneven stripe quality in the field of view caused by oblique projection of the projector, the present invention establishes a lookup table for different optimal reconstruction heights for different regions, so as to avoid the decline in reconstruction quality of local areas caused by a single fixed height.

[0022] (3) This invention not only records the optimal reconstruction height of each region, but also records the effective depth range, which can determine whether the local area of ​​the object to be measured needs to be compensated for acquisition, thereby reducing unnecessary displacement stage movement and image acquisition.

[0023] (4) The present invention obtains the approximate height of the object to be measured through preliminary reconstruction, and performs height stratification or clustering on the compensation displacement, so that multiple regions can be collected synchronously under the same representative compensation height, thereby improving measurement efficiency.

[0024] (5) In the point cloud stitching stage, the present invention introduces the actual displacement of the displacement stage and the calibration correction of the flat plate, and uses mass weight fusion in the overlapping area, which can reduce displacement error and misalignment, step and noise caused by hard switching of the area boundary.

[0025] (6) This invention does not change the basic equipment structure of the telecentric camera vertical arrangement, the projector oblique projection and the displacement stage carrying the object. It only completes the reconstruction optimization through calibration, acquisition strategy and data processing method. The implementation cost is low and it is suitable for the transformation of the existing monocular telecentric structured light detection platform. Attached Figure Description

[0026] Figure 1 This is a flowchart of the monocular telecentric structured light 3D reconstruction optimization and depth-of-field extension method based on the optimal reconstruction height of different regions, according to the present invention.

[0027] Figure 2 This is a schematic diagram of the monocular telecentric structured light measurement system of the present invention.

[0028] Figure 3 This is a schematic diagram of the standard ceramic plate multi-height data acquisition and comprehensive quality evaluation process.

[0029] Figure 4 This diagram illustrates the division of the camera's field of view, the optimal reconstruction height, and the effective depth of field range.

[0030] Figure 5 This is a schematic diagram of the preliminary reconstruction of the point cloud of the object under test. The left image is the original image acquired at the initial height, and the right image is a rough point cloud map at the initial height.

[0031] Figure 6 A schematic diagram of the initial point cloud height layering and local region extraction for the object under test.

[0032] Figure 7 The diagram shows the coordinate unification and stitching of local point clouds at multiple heights. (a) shows local point clouds at different compensation heights, (b) shows local point cloud 1 after coordinate unification, and (c) shows local point cloud 2 after coordinate unification.

[0033] Figure 8 The images show a comparison between depth-of-field fusion reconstruction and direct reconstruction results. (a) shows the overall point cloud after depth-of-field fusion, (b) shows the enlarged boundary region after depth-of-field fusion, (c) shows the enlarged texture region after depth-of-field fusion, (d) shows the overall point cloud after direct reconstruction, (e) shows the enlarged boundary region after direct reconstruction, and (f) shows the enlarged texture region after direct reconstruction. Detailed Implementation

[0034] Combination Figure 1This invention proposes a method for optimizing and extending the depth of field in monocular telecentric structured light 3D reconstruction based on the optimal reconstruction height for different regions. This method addresses the issues of inconsistent fringe sharpness, phase quality, and reconstruction accuracy in different regions of the same field of view caused by oblique projection of the projector in monocular telecentric structured light systems, and the difficulty of achieving effective depth of field across the entire field of view using traditional fixed measurement heights. It proposes a reconstruction scheme involving standard ceramic plate table creation, coarse reconstruction of the object under test, layered compensation acquisition by the displacement stage, and multi-height quality weight fusion. The system maintains the basic structure of the telecentric camera perpendicular to the displacement stage, the projector obliquely projecting structured light, and the displacement stage supporting the standard ceramic plate or the object under test. First, the displacement stage is controlled to move the standard ceramic plate stepwise along the height direction, acquiring structured light fringes and calculating the unwrapped phase at each height. Then, the camera's field of view is divided into several regions, and the phase straightness, fringe modulation, planar reconstruction residual, and repeatability stability are calculated for each region. A comprehensive quality evaluation function is constructed to determine the optimal reconstruction height and effective depth of field range for each region, forming a lookup table of optimal reconstruction heights for each region. When measuring the object under test, a preliminary 3D reconstruction is first performed to obtain a rough height distribution. Then, a lookup table is used to calculate the required compensation displacement for each region, and height-level hierarchical clustering is used to reduce the number of timers for the displacement stage. High-quality local point clouds are collected and reconstructed at each compensation height. Finally, the local point clouds are unified to the same coordinate system based on the actual displacement of the displacement stage and the calibration correction of the standard ceramic plate. In overlapping areas, the points are fused according to the reconstruction quality weights. This method utilizes the unwrapping phase straightness, fringe modulation, planar reconstruction residual, and repeatability stability of the standard ceramic plate at different heights on the displacement stage to establish a lookup table for the optimal reconstruction height for different regions of the same field of view. During the measurement of the object under test, the reconstruction accuracy is improved and the depth of field is expanded through displacement stage hierarchical compensation and multi-height point cloud fusion.

[0035] This invention does not require changes to the existing monocular telecentric structured light hardware layout, can compensate for the difference in optimal reconstruction height in the region caused by oblique projection of the projector, improve local phase stability and reconstruction accuracy, and effectively expand the system depth range.

[0036] The technical solution of the present invention will be further described below with reference to the accompanying drawings and specific embodiments. It should be understood that the following embodiments are only used to explain the present invention and are not intended to limit the scope of protection of the present invention. Equivalent substitutions or combinations made to the structured light coding method, region partitioning method, quality evaluation function, clustering method, displacement compensation method, and point cloud fusion strategy without departing from the core idea of ​​the present invention shall all fall within the scope of protection of the present invention.

[0037] (1) System structure and measurement principle

[0038] like Figure 2As shown, the monocular telecentric structured light 3D measurement system in this embodiment includes a telecentric camera 1, a projector 2, a displacement stage 4, a standard ceramic plate, the object to be measured, and a data processing unit 5. The telecentric camera 1 is mounted above the displacement stage 4, with its optical axis substantially perpendicular to the plane of motion of the displacement stage. The projector 2 is positioned to the side of the telecentric camera 1, projecting structured light stripes, including sinusoidal stripes, Gray code stripes, multi-frequency stripes, or other structured light coded patterns, onto the standard ceramic plate or the object to be measured on the displacement stage 4 at an inclined angle. Figure 2 In the diagram, 3 represents a standard ceramic plate or the object to be measured. The arrow pointing from the telecentric camera 1 to the data processing unit 5 indicates the direction of image data flow. The arrow pointing from the data processing unit 5 to the projector 2 indicates the projection control signal. The bidirectional arrow between the data processing unit 5 and the displacement stage 4 indicates the displacement control and displacement feedback signals. The bidirectional arrow on the left side of the displacement stage 4 indicates the lifting and lowering movement of the displacement stage along the height direction.

[0039] The coordinate relationship between the telecentric camera, projector, and displacement stage is obtained by using a calibration plate or standard plane, and the correspondence between the displacement stage's height movement direction, nominal step size, and actual displacement is determined.

[0040] Because telecentric cameras have good magnification stability, the imaging scale changes little at different heights, which is beneficial for unified coordinate compensation of multi-height point clouds based on the motion of the displacement stage. Since the projector is obliquely positioned, the distance and incident angle from the projector vary between different regions within the same field of view, causing the fringe modulation, fringe width, and phase quality of different regions to change with height. This invention utilizes the phase quality variation of a standard ceramic plate at different heights to determine the optimal reconstruction height for each region.

[0041] (2) Structured light phase calculation

[0042] At each height position, the projector projects N-step phase-shifted sinusoidal fringes onto the surface of a standard ceramic plate or the object under test. The nth fringe image acquired by the camera can be represented as:

[0043]

[0044] in, Represents camera pixel coordinates, Indicates background light intensity. Indicates the stripe tone system, This represents the desired wrapper phase. The wrapper phase is obtained using a phase-shifting algorithm. The unwrapped phase is obtained by combining Gray code, multi-frequency phase expansion, or time phase expansion methods. :

[0045]

[0046] in, This indicates the phase order. The unwrapped phase is a core intermediate quantity in 3D reconstruction, and its continuity, linearity, and stability directly affect the final reconstruction accuracy.

[0047] (3) Standard ceramic flat plate multi-height table building

[0048] like Figure 3 As shown, a standard ceramic plate is placed on the displacement stage, and the displacement stage is controlled to move from... Move to ,form height position At each height position, the system acquires structured light fringe images and calculates the unwrapping phase. Because standard ceramic plates have high flatness and stable reflection characteristics, their local unwrapped phase should ideally exhibit a stable linear or planar change.

[0049] Divide the camera's field of view into Local area , For the first The local region and the first At each height position, perform planar fitting on the local unwrapped phase:

[0050]

[0051] in, and Let represent the slope (phase gradient) coefficients of the fitted phase plane along the camera pixel coordinate u and v directions at the i-th local region and the k-th height position, respectively. This represents the constant term of the fitted plane, i.e., the phase bias.

[0052] Calculate the local phase linear residual:

[0053]

[0054] in, The smaller the value, the closer the phase is to the ideal planar distribution at that height, and the better the local reconstruction quality. Accordingly, the phase straightness evaluation term can be defined as:

[0055]

[0056] in, To prevent constants with zero denominators, in addition to phase straightness, this invention further introduces fringe modulation, reconstruction plane residuals, local smoothness, and repeatability stability to form a comprehensive quality evaluation function:

[0057]

[0058] in, This indicates the evaluation item for the stripe tone system. This represents the residual evaluation term of the standard ceramic plate reconstruction result relative to the fitted plane. This indicates a local phase or high smoothness evaluation term. This represents the standard deviation of phase or height obtained from repeated acquisitions. , , , and The weighting coefficients are used. This comprehensive evaluation function avoids misjudgments caused by relying solely on phase straightness, ensuring that the optimal height selection simultaneously considers clear fringes, stable phase, small planar reconstruction error, and good repeatability.

[0059] For each local region The height that maximizes the comprehensive quality evaluation function is selected as the optimal reconstruction height.

[0060]

[0061] At the same time, set a quality threshold. Or relative decrease The effective depth of field range for this region is obtained as follows:

[0062]

[0063] in, For the first The maximum quality rating value for a given region across all altitudes. and This allows for the creation of a lookup table for the optimal reconstruction height of a region. This lookup table can be pixel-level, block-level, or region-level. In practical applications, block-level or region-level lookup tables offer better noise resistance and lower storage requirements; pixel-level lookup tables can be used in more detailed scenarios.

[0064] (4) Lookup table smoothing and outlier handling

[0065] like Figure 4 As shown, the oblique projection of the projector causes the optimal reconstruction height to vary continuously along the projection direction. To avoid abnormal optimal heights in individual areas due to noise, reflections, or phase unfolding errors, this embodiment smooths the lookup table for the optimal reconstruction height of a region. Neighborhood median filtering, bilateral filtering, low-order surface fitting, or a one-dimensional smoothing method based on the projection direction can be used to smooth the optimal reconstruction height. Make corrections.

[0066] When the overall quality curve of a certain region is below the quality threshold across the entire height range, the region can be marked as an invalid region. When a region has multiple near-optimal heights, the middle position of the effective depth of field interval can be selected as the target compensation height to improve the fault tolerance capability under the positioning error of the displacement stage.

[0067] (5) Preliminary reconstruction and compensation height calculation of the object under test

[0068] like Figure 5 As shown, the object to be measured is placed on a displacement stage, and a rapid 3D reconstruction is performed at the initial height position to obtain a rough height distribution. This preliminary reconstruction is used to estimate the height range of the surface of the object being measured, and is not required to achieve the final measurement accuracy.

[0069] For pixels or region units The surface to be measured in the middle, based on the rough height and the optimal reconstruction height for the corresponding area Calculate the required displacement stage compensation. If the approximate height of the region is represented by the average height of the region, then combine... Figure 4 The table shown is the optimal reconstruction height lookup table for the region. Each area can be reconstructed at its optimal height. The displacement stage compensation amount corresponding to the rough height calculation of the object under test in this region. :

[0070]

[0071]

[0072] in, Indicates the first The displacement stage compensation amount corresponding to each region. In practical applications, The sign is determined by the definitions of the displacement stage coordinate system and the camera coordinate system. If a certain region... It has fallen into the effective depth of field range In this case, compensation data collection need not be performed separately for that area.

[0073] (6) Compensated height hierarchical clustering and multi-height acquisition

[0074] Moving the displacement stage separately for each region would result in a large number of data acquisitions. To improve efficiency, this invention compensates for the amount of data collected in all regions. By performing high-level stratification or clustering, similar compensation values ​​are merged into a single representative compensation height. The representative compensation height can be represented as:

[0075]

[0076] in, This represents the displacement compensation amount calculated for the i-th local region based on the approximate height and the lookup table. This indicates a highly hierarchical or clustering operator applied to the compensation amounts across all regions. Represents category index The corresponding representative compensation height, among which , Let M represent the number of compensation heights after clustering, and M be the number of local regions participating in the clustering. Clustering methods can include equally spaced height binning, K-means clustering, hierarchical clustering, or threshold-based merging strategies. Through this step, the displacement stage only needs to move at least a few representative compensation heights to cover the optimal or near-optimal reconstruction state of multiple regions.

[0077] like Figure 6 As shown, based on the Z-axis height distribution of the initially reconstructed point cloud, several height layers are divided, and the corresponding local regions of each layer are extracted. Figure 6 The horizontal axis represents the complete Z-axis height range; The calculation uses a sliding statistical window centered at candidate height z with a half-width of 0.1 mm. Local density peaks correspond to candidate height layers and do not represent the stage step size or system accuracy. Compensation amounts for each region are stratified or clustered, and similar compensation amounts are merged into representative compensation heights. Local point clouds are collected and reconstructed at each representative height, and masks are generated based on the effective depth of field range for each region, retaining only high-quality point clouds to avoid including areas with blurred fringes or unstable phases in the final result.

[0078] (7) Coordinate unification based on the correction amount of the displacement stage and the plate

[0079] like Figure 7 As shown, the local point clouds obtained at different compensation heights are located at different displacement stage positions. Directly superimposing the point clouds would cause a height shift. Therefore, it is necessary to unify the local point clouds to the initial coordinate system based on the actual displacement of the displacement stage. Let the first... The local point cloud acquired during the secondary compensation was The actual displacement of the displacement stage is Then we have:

[0080]

[0081] in, Indicates the first The local point cloud obtained from the secondary compensation acquisition is uniformly expressed in the initial displacement stage coordinate system after actual displacement compensation; This represents the point cloud in the local measurement coordinate system acquired during this acquisition; Indicates the direction of motion of the displacement stage and the first The homogeneous transformation matrix is ​​constructed from the actual displacement. If the displacement stage moves only along the camera's optical axis, the transformation is mainly a translation in the Z direction.

[0082] Considering the repeatability error of the displacement stage and the system reconstruction error, this invention further utilizes the plate fitting results obtained during the multi-height table building stage of the standard ceramic plate to correct the displacement. Let the correction amount be... Then, the point cloud in the unified coordinate system can be represented as:

[0083]

[0084] in, The error can be obtained by fitting the distance deviation, plane normal deviation, or displacement stage encoder error between reconstructed planes at different heights using a standard ceramic plate. This correction reduces the stitching error caused by relying solely on nominal displacement. It unifies the local point clouds obtained at different compensation heights to the same coordinate system, providing a foundation for subsequent point cloud fusion.

[0085] (8) Quality weight fusion

[0086] like Figure 8 As shown, after coordinate unification, the local point clouds at multiple heights are fused and compared with the direct reconstruction results. For overlapping regions, if multiple heights obtain 3D points at the same pixel or adjacent spatial locations, weights are assigned based on their respective reconstruction quality evaluation values. The fused height can be expressed as:

[0087]

[0088] in, Indicates the first Secondary compensation is collected at camera pixels. The fusion weight at the point, This indicates the local reconstruction height corresponding to this acquisition. This represents the height value after fusion. The fusion weight can be determined by the overall quality evaluation value of the current region, the deviation between the actual reconstructed height and the optimal reconstructed height, the fringe modulation degree, or the phase residual. For some regions, a strategy of retaining the region with the highest quality can also be adopted. For region boundaries, an overlapping fusion band of a certain width is set to allow the weight to change smoothly from one region to adjacent regions, thereby reducing steps and breaks caused by hard switching. Compared with direct reconstruction, the fused point cloud shows improvements in both boundary continuity and local surface continuity.

[0089] After fusion, outlier removal, hole filling, and surface smoothing can be performed to obtain a complete 3D reconstruction of the object under test.

[0090] Example

[0091] In this embodiment, a telecentric camera is vertically mounted above the displacement stage, and a projector is mounted to one side of the telecentric camera, projecting 15-step phase-shifted sinusoidal fringes onto the displacement stage at an inclined angle. The displacement stage initially supports a standard ceramic plate and moves within a set height range in fixed steps. A set of fringe images is acquired at each height position, and the unwrapped phase is calculated.

[0092] The camera's field of view is divided into several striped or rectangular grid regions along the oblique direction of the projector. For each region, the plane fitting residual of the unwrapped phase, the fringe modulation, the standard flat reconstructed plane residual, and the standard deviation of repeated acquisitions are calculated. A comprehensive quality evaluation function is then used to obtain the quality score for that region at each height. The location with the highest quality score is recorded as the optimal reconstruction height; the range of locations with quality scores above a threshold is recorded as the effective depth of field interval.

[0093] Subsequently, the object under test is placed on a displacement stage, and a rapid reconstruction is performed at the initial height position to obtain a coarse height distribution. The system calculates the required displacement stage compensation for each region based on the coarse height distribution and an optimal reconstruction height lookup table, and clusters similar compensation amounts into a few representative heights. The displacement stage is then moved sequentially to these representative heights, and the system acquires fringe images and reconstructs local point clouds at each representative height.

[0094] At each representative height, the data processing unit generates a mask based on whether the current height falls within the effective depth range of the corresponding area, retaining only the high-quality local point cloud at the current height. Then, based on the displacement correction amount obtained from the displacement stage encoder readings and the standard ceramic plate calibration, each local point cloud is transformed to the initial object coordinate system. For overlapping areas, fusion weights are set according to the comprehensive quality evaluation value, prioritizing or weighting the retention of point cloud data with higher fusion quality.

[0095] Through the above process, this embodiment can enable different regions of the object under test to be acquired near their respective optimal reconstruction heights without changing the hardware layout of the telecentric camera, projector, and displacement stage. Compared with single reconstruction at a fixed height, this method can reduce regional phase quality differences caused by oblique projection, improve local reconstruction stability, and obtain a larger effective depth of field range through multi-height fusion.

[0096] In this embodiment, the standard ceramic plate is only used for preliminary table building, displacement correction, and quality evaluation, and does not participate in the definition of the final morphology of the object under test. In practical applications, the standard ceramic plate can be replaced with other standard planar components with known flatness and stable reflection characteristics; the structured light coding method can adopt multi-step phase shifting, Gray code-assisted phase shifting, multi-frequency phase shifting, or other coding methods; the region division can be adaptively determined according to the oblique direction of the projector, the image grid, or the reconstruction quality distribution.

[0097] The above-described specific embodiments are merely preferred embodiments of the present invention and do not constitute a limitation on the scope of protection of the present invention. Any modifications, substitutions, and improvements made by those skilled in the art to the comprehensive quality evaluation function, effective depth-of-field threshold, clustering hierarchical strategy, displacement correction method, and point cloud fusion method based on the technical concept of the present invention should fall within the scope of protection of the present invention.

Claims

1. A method for optimizing and extending the depth of field in monocular telecentric structured light 3D reconstruction, characterized in that, Includes the following steps: Step 1: Build and calibrate a monocular telecentric structured light 3D measurement system. The system includes a telecentric camera, a projector, a displacement stage, a standard ceramic plate, and a data processing unit. The optical axis of the telecentric camera is perpendicular to the motion plane of the displacement stage. The projector projects structured light stripes onto the standard ceramic plate or the object to be measured on the displacement stage at an angle relative to the telecentric camera. Step 2: Place the standard ceramic plate on the displacement stage, control the displacement stage to move in steps along the height direction according to the preset step size, and acquire structured light stripe images at each height position. Step 3: Perform phase calculation and phase unwrapping on the structured light fringe images acquired at each height position to obtain the unwrapped phase distribution of the standard ceramic plate at different height positions; Step 4: Divide the field of view of the telecentric camera into multiple local regions, and calculate the reconstruction quality evaluation value of each local region at different height positions. The reconstruction quality evaluation value includes at least the unwrapped phase straightness evaluation value. Step 5: Based on the reconstruction quality evaluation values ​​of the same local area at different height positions, determine the optimal reconstruction height of the local area, determine the effective depth of field range of the local area, and establish a lookup table for the optimal reconstruction height of the area. Step 6: Place the object to be tested on the displacement stage and perform preliminary three-dimensional reconstruction at the initial height position to obtain the rough height distribution of the surface of the object to be tested. Step 7: Based on the rough height distribution and the regional optimal reconstruction height lookup table, calculate the displacement stage compensation amount required for different regions of the object under test, and perform height stratification or clustering on the displacement stage compensation amount to obtain several representative compensation heights. Step 8: Control the displacement stage to move sequentially to the representative compensation height, acquire structured light stripe images at each representative compensation height and perform local three-dimensional reconstruction to obtain multiple local high-quality point clouds; Step 9: Based on the actual displacement of the displacement stage and the displacement correction obtained from the standard ceramic plate calibration, the multiple local high-quality point clouds are transformed to a unified coordinate system, and the overlapping areas are fused according to the reconstruction quality weight to obtain the complete three-dimensional reconstruction result of the object under test.

2. The method of claim 1, wherein, The reconstruction quality evaluation value also includes one or more of the following: fringe modulation evaluation value, planar reconstruction residual evaluation value, local phase smoothness evaluation value, and repeated measurement stability evaluation value, which are combined by weighting to form a comprehensive quality evaluation function.

3. The method according to claim 2, characterized in that, The comprehensive quality evaluation function is: ,in This indicates the phase straightness evaluation item. This indicates the evaluation item for the stripe tone system. This represents the residual evaluation term of the standard ceramic plate reconstruction result relative to the fitted plane. This indicates a local phase or high smoothness evaluation term. This represents the standard deviation of height or phase obtained from repeated acquisitions. , , , and These are the weighting coefficients.

4. The method according to claim 1, characterized in that, The optimal reconstruction height lookup table for the region is a pixel-level lookup table, a block-level lookup table, or a region-level lookup table. The local region is a rectangular grid region, a strip region divided along the oblique direction of the projector, or an irregular region adaptively generated according to the stripe quality distribution.

5. The method according to claim 1, characterized in that, The effective depth of field range refers to the height range within which the reconstruction quality evaluation value of a local area is higher than a preset threshold or the decrease in reconstruction quality relative to the best is less than a preset proportion. The effective depth of field range is used to determine whether the local surface of the object under test needs to be compensated by a displacement stage.

6. The method according to claim 1, characterized in that, The high degree of hierarchical or clustering includes equal-space binning, K-means clustering, hierarchical clustering, or threshold-based merging of the required displacement stage compensation for different regions, in order to reduce the number of displacement stage movements and structured light image acquisitions.

7. The method according to claim 1, characterized in that, When performing local 3D reconstruction at various representative compensation heights, a region mask is generated based on the region's optimal reconstruction height lookup table, the preliminary reconstruction height of the object to be measured, and the current displacement stage compensation height, retaining only the local point cloud within the optimal reconstruction height or effective depth of field range.

8. The method according to claim 1, characterized in that, When the local high-quality point cloud is transformed to a unified coordinate system, the actual displacement is obtained based on the encoder reading of the displacement stage, the nominal step displacement, or the feedback from the displacement sensor, and coordinate compensation is performed by combining the displacement correction obtained by the multi-height reconstruction plane fitting of the standard ceramic plate.

9. The method according to claim 1, characterized in that, The fusion methods for the overlapping regions include weighted averaging based on reconstruction quality weights, selective retention based on maximum reconstruction quality, gradual fusion by setting an overlapping transition zone at the region boundary, or outlier removal, hole filling, and smoothing of the fused point cloud.

10. A monocular telecentric structured light 3D reconstruction optimization and depth-of-field extension system based on regional optimal reconstruction height, characterized in that, It includes a telecentric camera, a projector, a displacement stage, a standard ceramic plate, and a processor, the processor being configured to perform the method of any one of claims 1 to 9.

Images