High dynamic range surface reconstruction method in structured light measurement

Abstract

According to the high dynamic range surface reconstruction method in structured light measurement provided by the invention, the fringe projections of high and low amplitude intensities are fused, phases of high and low reflectivity areas are mutually compensated, the problem that the reflectivity difference of the object surface in the high dynamic range is relatively large is effectively solved, and finally, for a mirror reflection area which still cannot be recovered, the high dynamic range is reconstructed. According to the limiting continuity property, compensation is carried out after least square fitting of a phase curve is carried out, a mirror surface highlight area is effectively filled, the reconstruction integrity of a high dynamic range surface is improved, and the three-dimensional measurement precision is improved; and a complicated pre-calibration experiment is not needed to predict the optimal exposure time or projection fringe intensity template, so that the measurement efficiency is improved under the condition that the measurement precision is ensured, and the practicability is relatively high.

Description

Technical Field

[0001] This invention relates to the field of image recognition technology, specifically to a high dynamic range surface reconstruction method in structured light measurement. Background Technology

[0002] Structured light fringe projection is a typical optical 3D measurement method with advantages such as non-contact operation, high precision, and full-field coverage, making it widely used in machine vision, industrial measurement, and reverse engineering. This technique typically uses a digital projector to project multiple frequency sinusoidal fringes onto the object's surface to calculate the fringe order and unfolded phase map, mapping the object's height based on phase changes. However, the measurement accuracy of fringe projection is significantly reduced in areas with large differences in surface reflectivity or when disturbed by complex geometries. Therefore, high dynamic range (HDR) topography measurement has always been a challenge in structured light 3D measurement technology. In recent years, numerous scholars both domestically and internationally have proposed various HDR surface reconstruction methods, mainly including multiple exposure methods and fringe pattern intensity adjustment methods.

[0003] Multiple exposure methods primarily involve fusing fringe patterns acquired at different exposure times to obtain high-quality fringe or phase maps, which are then used for 3D reconstruction. While multiple exposure techniques can produce fringe images with a good signal-to-noise ratio and address high dynamic range issues, the large number of fringes used still results in low measurement efficiency. Fringe pattern intensity adjustment methods adjust the fringe intensity of different projection areas based on the surface reflectivity, or project and fuse multiple sets of appropriately intensified fringes to form the final 3D reconstructed fringe or phase. While this method improves measurement speed, it has limitations in morphology of highly reflective areas, leading to less accurate measurement results and higher hardware requirements. Therefore, achieving high measurement accuracy with less computation is a critical problem to be solved for surface reconstruction methods with high dynamic range. Summary of the Invention

[0004] To address the problem that existing surface reconstruction methods for high dynamic range cannot simultaneously achieve both measurement efficiency and accuracy, this invention provides a high dynamic range surface reconstruction method for structured light measurement. This method achieves high-precision measurement without complex calibration or a large number of projection fringes, thereby improving the completeness of high dynamic range surface reconstruction and increasing measurement efficiency.

[0005] The technical solution of this invention is as follows: a high dynamic range surface reconstruction method in structured light measurement, characterized by comprising the following steps: S1: Project a blank image with an amplitude of 255 onto a white diffuse surface, adjust the hardware so that the acquired grayscale is approximately 255 and fix the hardware parameters. S2: According to the preset high and low fringe projection amplitudes, use the K-step fringe pattern to project onto the measured surface and collect the fringe projection pattern; For each stripe image acquired, the modulation index S(x,y) is calculated according to the amplitude. The modulation index is calculated based on K high-amplitude fringe projection maps and denoted as: S h (x,y); The modulation index is calculated based on K low-amplitude fringe projection maps and denoted as: S l (x,y); Project the measured surface according to the preset blank amplitude, and acquire a blank image I. b； Based on blank graph I b Calculate the normalized reflectance r(x,y) from the corresponding projection plot; S3: Based on the reflectivity distribution and modulation index, divide the projection maps corresponding to the two amplitudes into regions to obtain the reflection regions to be processed included in the projection maps; The reflection areas to be processed include: specular reflection areas, high reflectivity areas, and low reflectivity areas; S4: Project the K-step fringe pattern onto the measured surface according to the preset high and low fringe projection amplitudes, and collect the fringe projection images. Calculate the phase diagrams corresponding to the two sets of fringe projection images for the high and low amplitudes respectively, denoted as: φ h (x,y) and φ l (x,y); For phase diagram φ h (x,y) and φ l Phase fusion is performed on the reflection region to be processed corresponding to (x,y), specifically including the following steps: a1: Projecting phase diagram φ with low amplitude l Based on (x,y), the low-amplitude projection phase diagram φ l The phase of the low-reflection region in (x,y) is represented by a high-amplitude projection phase diagram φ. h The phase of the corresponding region in (x,y) is replaced to obtain the fused projection phase map, which is denoted as φ. f (x,y); a2: Perform least-squares fitting phase compensation on the specular reflection region in the fused projection phase map to obtain the final phase φ(x,y); S5: Calibrate the acquisition system and use the system calibration parameters and the phase values ​​of each point in the final phase φ(x,y) to complete the reconstruction of the three-dimensional point cloud.

[0006] Its further features are: In step S2, the projection amplitudes of the high and low fringes satisfy the following conditions: The exposure of the reflective areas in the high and low amplitude fringe projection maps is complementary; The blank amplitude is the same as the lower amplitude of the two fringe projections; The stripe projection amplitude is set to 255 and 200; the blank amplitude is set to 200. In step S2, the reflectivity r(x,y) is calculated as follows: r(x,y)=I b (x,y) / 255; In the formula, (x,y) are pixel coordinates, I b (x,y) represents a blank graph I. b The grayscale value of each pixel in the corresponding projection image; In step S2, the modulation S(x,y) is calculated as follows: ; In the formula, K is the total number of fringe projection images corresponding to each amplitude, and k is the k-th fringe projection image; I k '(x,y) is the gray value of the pixel with coordinates (x,y) in the i-th stripe projection image; The gray value I k The calculation method for (x,y) is as follows: I k '(x,y) =ρrtI k (x,y)+E; In the formula: ρ is the camera sensitivity; r is the surface reflectivity of the object being measured; t is the camera exposure time; E For random noise; I k (x,y) represents the grayscale value projected by the projector; k '(x,y) represents the grayscale values ​​actually captured by the camera; In step S3, the specific rules for dividing the reflection area to be processed are as follows: ; Among them, mask l (x,y), mask h (x,y) 、mask s (x,y) represent the low-reflection region, high-reflection region, and specular reflection region, respectively, and (x,y) are pixel coordinates; Step a2 specifically includes the following steps; a21: Perform phase adjustment n The least-squares fit of a polynomial curve of degree n is expressed as follows: ; In the formula, x i and y i Let a0~a1 be the coordinates of the i-th pixel in the row and column directions of the phase map. n and b0~b n Let C(x) be the coefficients of the polynomial. i ) and C(y i () represents the fitted curve; a22: By using the normal phase values ​​around the specular highlight area to perform curve fitting in the horizontal and vertical directions respectively, the phase values ​​corresponding to the central highlight saturated pixels are corrected; this process is transformed into solving... n Coefficients of the polynomial of degree: ; ; ; Among them, h For pixel (x) i ,y i The number of row pixels acquired around the specular highlight area; w For pixel (x) i ,y i The number of column pixels acquired around the specular highlight area; φ(x i ), φ(y i ) for pixel (x) i ,y i Normal phase values ​​acquired around the specular highlight area; J x J is the fitted curve in the horizontal direction. y The fitted curve is in the vertical direction; a23: Calculated n After obtaining the coefficients of the polynomial, the curve constructed based on the obtained coefficients is used to compensate and fill the saturated pixels in the specular highlight area, and φ is adjusted accordingly. f The final phase φ(x,y) is obtained by fitting and compensating the corresponding specular reflection region in (x,y). Step S5 specifically includes the following steps: b1: Use the reverse engineering method to calibrate the projector and establish the relationship between the camera pixel plane and the projector pixel plane by using the phase value of the projected fringes; b2: Generate a corresponding virtual chessboard image on the pixel plane of the projector based on the established correspondence between the chessboard images captured by the camera; b3: The projector is calibrated using the camera calibration method, and the relative positional relationship between the camera and the projector is calculated using stereo vision methods to achieve system calibration; b4: Reconstruct the 3D point cloud using the system calibration parameters and the final phase φ(x,y) obtained from fitting compensation.

[0007] The high dynamic range surface reconstruction method in structured light measurement provided in this application uses two fringe projections of high and low amplitude intensities for fusion, mutually compensating for the phase of high and low reflectivity regions. This effectively solves the problem of large differences in surface reflectivity within the high dynamic range. Finally, for specular reflection regions that still cannot be recovered, compensation is performed by least-squares fitting of the phase curve based on the continuity property of the limit, effectively filling the specular highlight regions. This improves the reconstruction integrity of the high dynamic range surface and enhances the accuracy of three-dimensional measurement. It eliminates the need for complex pre-calibration experiments to predict the optimal exposure time or projected fringe intensity template, improving measurement efficiency while ensuring measurement accuracy, and is highly practical. Attached Figure Description

[0008] Figure 1 These are fringe patterns with different intensity amplitudes; Figure 2 It is a single-row grayscale curve of a stripe pattern with different intensity amplitudes; Figure 3 These are the results of fringe phase unfolding at different intensity amplitudes; Figure 4 These are fringes of different intensity amplitudes at a signal-to-noise ratio of 30dB. Figure 5 It represents the phase error of each intensity fringe under different noise levels; Figure 6 It is the standard deviation of the phase error of each intensity fringe under different noise conditions; Figure 7 This is an example of high dynamic range surface region partitioning; Figure 8 This is an example of a point cloud of a specular reflection region; Figure 9 This is a hardware diagram of the experimental system; Figure 10 These are the collected fringe patterns under different amplitude projections: (a) Maximum amplitude 200; (b) Maximum amplitude 255. Figure 11 The following are the modulation and reflectance diagrams: (a) the fringe modulation with a maximum amplitude of 200; (b) the fringe modulation with a maximum amplitude of 255; (c) the reflectance distribution diagram; and (d) the division of reflectance areas. Figure 12 These are the unfolded phase diagrams and final phase diagrams of fringes with different amplitudes; where: (a) the fringe phase with the highest amplitude of 200; (b) the fringe phase with the highest amplitude of 255; and (c) the final phase result after processing by the proposed method. Figure 13These are the point cloud reconstruction results of stripes with different amplitudes, including (a) the point cloud reconstructed from stripes with the highest amplitude of 200; (b) the point cloud reconstructed from stripes with the highest amplitude of 255; and (c) the point cloud result after processing by the proposed method. Detailed Implementation

[0009] Based on theoretical and simulation analysis, this method addresses the high dynamic range surface reconstruction problem in 3D fringe projection measurement. It evaluates the signal-to-noise ratio and phase error of images under fringe projections of different intensity amplitudes, and proposes a high dynamic range surface reconstruction method combining dual-amplitude fringe projection and phase curve fitting. First, phase fusion is performed using two sets of fringes with different intensity amplitudes to solve the problem of large differences in surface reflectivity. Finally, based on the phase continuity property, phase curve fitting is performed on the specular reflection regions that cannot be recovered during fusion, thereby achieving high-quality 3D reconstruction of surfaces with varying reflectivity. Specifically, the high dynamic range surface reconstruction method in structured light measurement of this application includes the following steps.

[0010] S1: Project a blank image with an amplitude of 255 onto a white diffuse surface, adjust the hardware so that the acquired grayscale is approximately 255 and fix the hardware parameters.

[0011] To ensure a good signal-to-noise ratio, the highest intensity amplitude of the stripes generated by the computer is typically 255. However, when there are significant differences in reflectivity on the surface of the object being measured, the pixels projected onto the highest intensity area are prone to saturation, forming saturated regions that hinder accurate 3D reconstruction. Currently, a widely used method is to identify saturated regions through pre-acquisition and other methods, then reduce the exposure time or the projection intensity of that region, and finally perform image fusion to complete the reconstruction. The image of the stripes projected by the projector onto the camera is as follows: I k '(x,y) =ρrtI k (x,y)+E; In the formula: (x,y) are pixel coordinates, ρ is camera sensitivity; r is the surface reflectivity of the object being measured; t is the camera exposure time; E is random noise; Ik(x,y) is the gray value projected by the projector; Ik'(x,y) is the gray value actually acquired by the camera.

[0012] In industrial measurements, E is much smaller than ρrtI. k (x, y) Therefore, when the camera sensitivity and exposure time are fixed, the intensity of the acquired fringe is proportional to the intensity of the projected fringe. When encountering a surface with high reflectivity, the appearance of image saturation regions can be prevented by reducing the amplitude of the projected fringe intensity. Therefore, a blank image with an amplitude of 255 is acquired by projecting onto a white diffuse reflective surface. Based on the aforementioned analysis, the hardware is adjusted to make the acquired grayscale approximately 255, and the hardware parameters are fixed.

[0013] S2: Project the fringe pattern onto the surface under test using K steps according to the preset high and low fringe projection amplitudes, and collect the fringe projection images; project the fringe pattern onto the surface under test according to the preset blank amplitude, and collect a blank image I. b .

[0014] The projection amplitudes of the high and low fringes satisfy the following conditions: The exposure of the reflective areas in the two amplitude fringe projection maps is complementary.

[0015] That is, if a high-reflectance area in a high-amplitude projection image is overexposed, it will not be overexposed in a low-amplitude projection image; similarly, a low-reflectance area in a low-amplitude projection image may be too dark and have poor phase quality, but the phase quality is better in a high-amplitude projection image.

[0016] At the same time, the blank amplitude is set to be the same as the lower amplitude of the two fringe projections. In this embodiment, the high amplitude value is set to 255, the low amplitude value is set to 200, and the blank amplitude value is set to 200.

[0017] Project the measured surface and acquire K-step fringe patterns with amplitudes of 255 and 200, and a blank pattern with an amplitude of 200. b ; The value of K is set according to actual needs. In this embodiment, K is set to 8.

[0018] Based on blank graph I b The corresponding projection image is used to calculate the normalized reflectance r(x,y) based on the mean image of the stripe pattern; the calculation method for reflectance r(x,y) is as follows: r(x,y)=I b (x,y) / 255; In the formula, (x,y) are pixel coordinates, I b (x,y) represents a blank graph I. b The grayscale value of each pixel in the corresponding projection image.

[0019] For each stripe image acquired, the modulation index S(x,y) is calculated based on the amplitude of the acquired stripe pattern. The modulation index was calculated based on K fringe projections with an amplitude of 255, and denoted as: S h (x,y); The modulation index, calculated based on K fringe projection maps with an amplitude of 200, is denoted as: S l (x,y).

[0020] Specifically, the modulation index S(x,y) is calculated as follows: ; In the formula, K is the total number of fringe projection images corresponding to each amplitude, and k is the k-th fringe projection image; I k '(x,y) is the gray value of the pixel with coordinates (x,y) in the i-th stripe projection image.

[0021] In actual measurements, to avoid pixel saturation, stripes with smaller amplitudes are typically selected for projection and acquisition. However, this selection results in four typical regions on the surface of the object being measured: normal diffuse reflection region, high diffuse reflection region, low diffuse reflection region, and specular reflection region.

[0022] Normal diffuse reflectance regions provide clear fringe information, enabling accurate 3D reconstruction. Low diffuse reflectance regions, due to their low signal-to-noise ratio, may introduce some phase error, but the impact is minor. High diffuse reflectance regions are typically located around specular reflectance regions; when projecting large-amplitude fringes, they can lead to severe phase errors. As for specular reflectance regions, due to the surface imaging angle, regardless of how low the amplitude of the projected fringes is, the fringe information of this region cannot be fully recovered, resulting in invalid phase data, which manifests as flying spots or holes in the point cloud.

[0023] S3: Based on the reflectivity distribution and modulation value, divide the projection maps corresponding to the two amplitudes into regions respectively.

[0024] In this method, regions are divided based on reflectivity distribution and modulation index. Regions that remain saturated under low-amplitude projection are considered specular reflection regions, and regions with high reflectivity are considered high-reflectivity regions, with the phase calculated directly using low-amplitude fringes. Regions with excessively low reflectivity and modulation index less than that of high-amplitude fringes are considered low-reflectivity regions, with the phase calculated using high-amplitude fringes. The effective threshold for fringe modulation index is empirically set to 0.2. This yields the reflection regions to be processed included in the projection map for each amplitude; these regions include specular reflection regions, high-reflectivity regions, and low-reflectivity regions.

[0025] The specific rules for dividing the reflection area to be processed are as follows: ; Among them, mask l (x,y), mask h (x,y) 、mask s (x,y) represent the low-reflection region, high-reflection region, and specular reflection region, respectively, and (x,y) are pixel coordinates.

[0026] S4: Calculate the phase diagrams corresponding to the two sets of fringe projections for high and low amplitude values, respectively, denoted as: φ h (x,y) and φ l (x,y); In this embodiment, the acquisition amplitudes are set to 255 and 200 respectively. A K-step fringe pattern is used, and three frequencies are set. Projection patterns in the horizontal and vertical directions are acquired respectively. The phase pattern corresponding to the fringe projection with an amplitude of 255 and the phase pattern corresponding to the fringe projection with an amplitude of 200 are calculated. The specific calculation method is based on existing technology.

[0027] For phase diagram φ h (x,y) and φ l Phase fusion is performed on the corresponding reflection regions to be processed in (x,y), specifically including the following steps.

[0028] a1: Projecting phase diagram φ with low amplitude l Based on (x,y), the low-amplitude projection phase diagram φ l The phase of the low-reflection region in (x,y) is represented by a high-amplitude projection phase diagram φ. h The phase of the corresponding region in (x,y) is replaced to obtain the fused projection phase map, which is denoted as φ. f (x,y).

[0029] When highly reflective areas appear on the surface of an object being measured, pixel saturation is typically avoided by reducing camera exposure time or decreasing projected fringe intensity. However, due to the characteristics of the material surface and the influence of the imaging angle, some bright areas may still exist even after adjustments. These areas are usually located in smooth and uniform parts of the surface, prone to specular reflection, leading to incomplete phase information and thus causing significant phase errors, which appear as holes or invalid points in the point cloud. Nevertheless, the phase of these areas is usually continuous. Therefore, after mitigating the bright areas by adjusting the fringe intensity, the remaining areas can be compensated and filled using phase least squares fitting, effectively restoring the fringe information of these areas and improving the accuracy and completeness of the 3D reconstruction.

[0030] a2: Perform least-squares fitting phase compensation on the specular reflection region in the fused projection phase map to obtain the final phase φ(x,y). This includes the following steps.

[0031] a21: Perform phase adjustment n The least-squares fit of a polynomial curve of degree n is expressed as follows: ; In the formula, x i and y i Let a0~a1 be the coordinates of the i-th pixel in the row and column directions of the phase map. n and b0~b n Let C(x) be the coefficients of the polynomial. i ) and C(y i () represents the fitted curve; a22: By using the normal phase values ​​around the specular highlight area to perform curve fitting in the horizontal and vertical directions respectively, the phase values ​​corresponding to the central highlight saturated pixels are corrected; this process is transformed into solving... n Coefficients of the polynomial of degree: ; ; ; Among them, h For pixel (x) i ,y i The number of row pixels acquired around the specular highlight area; w For pixel (x) i ,y i The number of column pixels acquired around the specular highlight area; φ(x i ), φ(y i ) for pixel (x) i ,y i Normal phase values ​​acquired around the specular highlight area; J x J is the fitted curve in the horizontal direction. y The fitted curve is in the vertical direction; a23: Calculate the coefficients {a*} and {b*}. Based on the curve constructed from the obtained coefficients, compensate and fill the saturated pixels in the specular highlight area, and adjust φ. f The final phase φ(x,y) is obtained by fitting and compensating the corresponding specular reflection region in (x,y).

[0032] S5: Calibrate the acquisition system and use the system calibration parameters and the phase values ​​of each point in the final phase φ(x,y) to complete the reconstruction of the three-dimensional point cloud.

[0033] Step S5 specifically includes the following steps: b1: Use the reverse engineering method to calibrate the projector and establish the relationship between the camera pixel plane and the projector pixel plane by using the phase value of the projected fringes; b2: Generate a corresponding virtual chessboard image on the pixel plane of the projector based on the established correspondence between the chessboard images captured by the camera; b3: The projector is calibrated using the camera calibration method, and the relative positional relationship between the camera and the projector is calculated using stereo vision methods to achieve system calibration; b4: Reconstruct the 3D point cloud using the system calibration parameters and the final phase φ(x,y) obtained from fitting compensation.

[0034] This method analyzes the phase calculation principle in structured light 3D measurement and simulates the impact of different noise conditions and different amplitude fringe projections on phase error. It proves that different fringe amplitude intensities have little impact on phase error under the same noise conditions. Based on this, a high dynamic range surface reconstruction method based on the combination of dual amplitude projection and phase curve fitting is proposed. That is, fringe projections with high and low amplitude intensities are fused to compensate for the phase of high and low reflectivity regions. Finally, least squares fitting phase compensation is performed on the specular reflection regions that still cannot be recovered.

[0035] To verify the effectiveness of this method, a verification experiment was constructed. The method is explained in detail below, with reference to accompanying figures and examples.

[0036] First, to verify whether fringes of different amplitudes affect phase unfolding, fringe sequences with maximum intensities of 255, 200, 150, and 100 were generated by computer, as follows: Figure 1 As shown. The grayscale curve of a certain row of stripes is as follows. Figure 2 As shown, Figure 2 The vertical axis represents the gray value, and the horizontal axis represents the pixel coordinate.

[0037] Phase calculation and unfolding are performed when the maximum intensity of the projected fringes are different, and the unfolded phase is as follows: Figure 3 As shown, Figure 3 In the diagram, the Z-axis represents the phase, and the x and y axes represent pixel coordinates. Simulation results show that under ideal conditions, the phase calculations for fringes of different intensities completely overlap, and the accuracy of the unfolded phase is not affected by the fringe intensity.

[0038] In actual imaging processes, noise from the camera and the environment is unavoidable. The signal-to-noise ratio of industrial cameras is typically above 40 dB, and with the development and improvement of manufacturing processes in recent years, it can even reach 60 dB. Therefore, this study uses a computer to generate Gaussian noise of 25, 30, 35, and 40 dB respectively and adds it to the fringe pattern for simulation analysis, comparing the phase error generated by different amplitude fringes under different noise environments. Figure 4 These are the intensity fringes under a signal-to-noise ratio of 30 dB. Figure 5 The diagram shows the phase error curves for a specific row of fringes at different intensities under varying noise conditions. Figure 5 The vertical axis of each graph represents the unwrapped phase, and the horizontal axis represents the pixel.

[0039] Figure 6 The standard deviation of the phase error for each intensity fringe under different noise conditions is given. As the signal-to-noise ratio (SNR) decreases, the phase error gradually increases, but the phase error for fringes of different intensities at the same SNR does not change significantly. Figure 6 The vertical axis represents the phase standard deviation, and the horizontal axis represents the fringe amplitude. This demonstrates that almost the same measurement accuracy can be achieved using fringe patterns with any set of intensity amplitudes. Therefore, the reconstruction using fringe projection patterns corresponding to two sets of amplitude values ​​in this application can meet practical needs.

[0040] In actual measurement, to avoid pixel saturation, stripes with smaller amplitudes are usually selected for projection and acquisition, resulting in four typical regions on the surface of the measured object: normal diffuse reflection region, high diffuse reflection region, low diffuse reflection region, and specular reflection region. For example... Figure 7 The image shown is obtained by projecting low-amplitude fringes onto two metal workpieces, with a white diffuse reflective background. Normal diffuse reflective areas provide clear fringe information, enabling accurate 3D reconstruction. Low diffuse reflective areas, due to their low signal-to-noise ratio, may introduce some phase error, but the impact is minor. High diffuse reflective areas are typically located around specular reflective areas; when projecting large-amplitude fringes, they can lead to severe phase errors. As for specular reflective areas, due to the surface imaging angle, regardless of how low the amplitude of the projected fringes is, the fringe information in this area cannot be fully recovered, resulting in invalid phase data, which manifests as flying spots or holes in the point cloud, such as... Figure 8 As shown, it is necessary to use the normal phase values ​​around the specular highlight area to perform curve fitting in the horizontal and vertical directions respectively, in order to correct the phase values ​​corresponding to the highlight saturation pixels in the middle.

[0041] To verify the effectiveness of the method presented in this paper, a grating projection 3D measurement system consisting of a projector and a camera was constructed, such as... Figure 9 As shown in the figure, the system includes a digital projector (marked as 1 in the figure) with a resolution of 1280×720 pixels; and a Daheng MER-504-10GM-P camera (marked as 2 in the figure) with a resolution of 2448×2048 pixels and a signal-to-noise ratio of 38.5 dB. For the algorithm, a three-frequency, eight-step phase-shifting method is used for phase extraction and unfolding, with the number of fringe periods corresponding to the three frequencies selected as 64, 63, and 56, respectively. The experiment used a metal workpiece with a high dynamic range surface, exhibiting high surface reflectivity and partial specular reflection.

[0042] Using the constructed experimental system, two amplitude fringes of 255 and 200 were used for projection and acquisition. Only one fringe was needed for projection; the other fringe could be expanded using the calibration parameters and limit constraints for final phase fitting. Figure 10As shown in the figure, under high amplitude fringe projection, areas with high reflectivity on the workpiece surface exhibit pixel saturation, resulting in significant errors in fringe information. Under low fringe intensity amplitude projection, underexposure occurs on the surface of the measured object with low reflectivity, while overexposure still occurs in specular reflection areas, leading to missing fringe information.

[0043] The modulation indices of the two amplitude stripes calculated from the stripe diagram are as follows: Figure 11 As shown in (a) and 11(b), the reflectance map calculated from the blank map is as follows: Figure 11 As shown in (c). It can be seen from the figure that in the normal reflection area (corresponding to...) Figure 11 (d) In the gray area, the modulation of the two amplitude fringes is not significantly different. In the specular reflection area, the modulation of the low amplitude fringes remains very small, while in the surrounding high-reflection area, the modulation is alleviated compared to the high amplitude fringes, and the modulation value is higher. Therefore, it is necessary to combine the two characteristic indicators to divide the fringe pattern into regions. The result of the region division according to equation (13) is as follows: Figure 11 As shown in (d), in addition to the normal reflection area, the specular reflection area, high reflection area, and low reflection area are marked in red, green, and blue, respectively, providing a basis for further phase fusion.

[0044] Phase calculations can yield the unfolded phase diagrams corresponding to the two sets of fringe patterns, such as... Figure 12 As shown in (a) and 12(b), it can be seen from the figures that the phase after fusion and phase fitting using this method is as follows: Figure 12 As shown in (c), in the high-reflectivity and specular-reflectivity regions, the phase loss and error conditions in the low-amplitude fringe phase are significantly better than those in the high-amplitude fringe phase, but a small number of holes still exist. Therefore, by fusing the phases with optimal reflectivity and modulation and fitting and compensating for the remaining holes, the final phase result is more accurate and complete, the high-reflectivity and specular-reflectivity regions are compensated, and the phase continuity is better.

[0045] Finally, based on the system calibration results, the 3D point cloud is calculated, and the point clouds corresponding to the two amplitude fringes are as follows: Figure 13 (a) Figure 13 As shown in (b), the reconstructed point cloud exhibits significant errors and large void areas when projecting high-amplitude fringes. While the reconstructed point cloud shows some improvement with low-amplitude projection, some voids and errors still remain. This is because these areas are specular reflection regions, making it difficult for even low-amplitude fringes to form clear images. Finally, the point cloud after phase curve fitting compensation is shown in [image / description]. Figure 13As shown in (c), the point cloud of the highlighted area on the left was restored, and the axial highlight band error of the metal tool on the right was also compensated, resulting in a more accurate cylindrical point cloud result, which improved the integrity and accuracy of the 3D point cloud reconstruction.

[0046] The above experiments demonstrate that more complete and accurate reconstruction and measurement of high dynamic range surfaces can be achieved without the need for complex pre-calibration experiments to predict the optimal exposure time or projected fringe intensity template. This improves measurement efficiency while ensuring measurement accuracy and has strong practicality.

Claims

1. A method for high dynamic range surface reconstruction in structured light measurement, characterized in that, It comprises the following steps: S1: projecting and collecting a blank graph with an amplitude of 255 to a white diffuse reflection surface, adjusting the hardware to make the collected gray scale approximate to 255 and fixing the hardware parameters; S2: projecting and collecting the fringe projection graphs according to the preset high and low fringe projection amplitudes using K-step fringe graphs; For each collected fringe image, the modulation S(x, y) is calculated according to the amplitude; Based on the high amplitude of K stripe projection map calculation, the modulation is recorded as: S h (x, y); Based on low amplitude K fringe projection image, the modulation is calculated and recorded as: S l (x, y); According to the preset blank amplitude, the measured surface is projected, and a blank image I is collected b； Based on the blank map I b The corresponding projection map, the normalized reflectance r(x, y) is calculated; S3: dividing the projection graphs corresponding to the two amplitudes into regions according to the reflectivity distribution and the modulation value, to obtain the to-be-processed reflection regions included in the projection graphs; The to-be-processed reflection regions include: a specular reflection region, a high reflectivity region and a low reflectivity region; S4: according to the preset high and low two fringe projection amplitudes, using K-step fringe pattern to project the measured surface, and collecting the fringe projection pattern, respectively calculating the phase diagram corresponding to the high and low amplitude two fringe projection patterns, denoted as: φ h (x,y) and φ l (x,y); performing phase fusion on the to-be-processed reflection region corresponding to the phase map φ h (x,y) and φ l (x,y) and φ a1: project the low-amplitude phase map φ l (x,y) with the low-amplitude phase map φ l (x,y) with the low-amplitude phase map φ h (x,y) with the low-amplitude phase map φ f (x,y) ; a2: performing least square fitting phase compensation on the specular reflection region in the fused projection phase graph to obtain the final phase φ(x, y); S5: calibrating the collection system, and completing the reconstruction of the three-dimensional point cloud by using the system calibration parameters and the phase values of each point in the final phase φ(x, y).

2. The structured light measurement mid-high dynamic range surface reconstruction method according to claim 1, wherein: In step S2, the high and low fringe projection amplitudes satisfy the following conditions: The exposure conditions of the reflection regions in the high and low amplitude fringe projection graphs are complementary; The blank amplitude is the same as the lower amplitude in the two fringe projections.

3. The structured light measurement mid-high dynamic range surface reconstruction method according to claim 1, wherein: The fringe projection amplitudes are set to 255 and 200, and the blank amplitude is set to 200.

4. The structured light measurement mid-high dynamic range surface reconstruction method of claim 1, wherein: In step S2, the calculation method of the reflectivity r(x, y) is: r(x,y) = I b (x,y) / 255; where (x, y) is the pixel coordinate, I b (x, y) is the blank image b the gray value of each pixel in the corresponding projection image.

5. The structured light measurement mid-high dynamic range surface reconstruction method of claim 1, wherein: In step S2, the calculation method of the modulation S(x, y) is: ； In the formula, K is the total number of the fringe projection images corresponding to each amplitude, k is the kth fringe projection image; I k (x,y) is the gray value corresponding to the pixel point with coordinates (x,y) in the ith fringe projection image.

6. The structured light measurement mid-high dynamic range surface reconstruction method according to claim 5, wherein: The gray value I k The calculation method of (x, y) is: I k '(x,y) =ρrtI k (x,y)+E; where p is the camera sensitivity; r is the surface reflectance of the object being measured; and t is the camera exposure time; E is the random noise; I k (x,y) is the gray value projected by the projector; I k '(x,y) is the gray value actually captured by the camera.

7. The structured light measurement mid-high dynamic range surface reconstruction method of claim 1, wherein: In step S3, the specific division rules of the to-be-processed reflection regions are as follows: ； Wherein, mask l (x,y), mask h (x,y), mask s (x,y) are low reflection area, high reflection area, mirror reflection area respectively, and (x,y) is pixel coordinate.

8. The structured light measurement mid-high dynamic range surface reconstruction method of claim 1, wherein: In step a2, the following steps are specifically included: a21: performing a phase on n least squares fitting of a polynomial curve of degree two, expressed as: ； In the formula, x i and y i are the coordinates of the i-th pixel point in the row and column directions of the phase map, a0~a n and b0~b n are the coefficients of the polynomial, and C(x i ) and C(y i ) are the fitted curves. a22: use normal phase values around the mirror highlight area to do curve fitting in horizontal and vertical directions respectively to correct the phase value corresponding to the middle highlight saturated pixels; this process is converted to solve n coefficients of the sub-polynomial: ； ； ； wherein, wherein h is the number of rows of pixels acquired around the specular highlight region for the pixel point (x i ,y i ). w The number of column pixels acquired around the specular highlight region for the pixel point (x i ,y i ); φ(x i ,y i ) are normal phase values acquired around the mirror highlight region for the pixel point (x i ,y i ); J x is a fitting curve in the horizontal direction, J y is a fitting curve in the vertical direction; a23: calculate n After the coefficients of the sub-polynomial are calculated, the saturated pixels in the mirror highlight area are compensated and filled according to the curve constructed by the obtained coefficients, and φ f (x, y) is fitted and compensated to obtain the final phase φ(x, y).

9. The method of claim 1, wherein: In step S5, the following steps are specifically included: b1: calibrating the projector using the inverse machine method, and establishing the relationship between the camera pixel plane and the projector pixel plane through the phase values of the projected fringe; b2: generating a corresponding virtual checkerboard image on the projector pixel plane through the established corresponding relationship of the camera collected checkerboard image; b3: calibrating the projector by the method of calibrating the camera, so as to calculate the relative position relationship between the camera and the projector by using the stereo vision method, and realize the system calibration; b4: completing the reconstruction of the three-dimensional point cloud by using the system calibration parameters and the fitted and compensated final phase φ(x, y).

Images