An adaptive point cloud fusion method based on error evaluation

Abstract

This invention discloses an adaptive point cloud fusion method based on error assessment, comprising the following steps: calibrating the camera and projector in the structured light system; projecting binary-coded and phase-shifted sinusoidal-coded fringes using the projector, and capturing structured light images using the camera; extracting the sub-pixel coordinates of the centerline of the structured light fringes, calculating the phase features of the phase-shifted coded fringes, and reconstructing two sets of 3D point clouds respectively; analyzing the incident angle features of the point clouds, the relationship between the pre-assessed point cloud error and the incident angle, and calculating the fusion weight; fusing matching points in the two sets of 3D point clouds, and interpolating the fusion vector of the matching points based on the pixel positions of the 3D points in the 2D image to fuse the non-matching points. This invention solves the problems of inaccurate extraction of the centerline of the fringes in the edge region and interference from noise in all regions on the phase calculation in the existing structured light 3D measurement process, which further leads to measurement errors.

Description

Technical Field

[0001] This invention relates to the field of three-dimensional measurement technology, and in particular to an adaptive point cloud fusion method based on error assessment. Background Technology

[0002] Structured light 3D measurement, as an active non-contact measurement method, has the advantages of being fast, real-time, easy to implement, high-precision, non-destructive, and flexible across the entire field. During measurement, structured light is first projected onto the object being measured. Then, a camera or sensor is used to capture the structured light information modulated on the object's surface. Finally, by combining the sensor's internal parameters and positional relationships, the deformed structured light pattern is analyzed to reconstruct the 3D information of the object's surface.

[0003] The fringe center method and the phase-shifting method are two approaches to achieve 3D structured light measurement. The former obtains the pixel coordinates of the projector corresponding to the fringe center by identifying its ordinal number. Extracting the sub-pixel coordinates of the fringe center is a key step in the fringe center method. A Taylor series expansion is performed on the grayscale values ​​of the light stripes along the fringe normal direction to identify the extreme positions of the center points. However, in areas with large projection angles, fringes with large grayscale jumps may form in local areas of the field of view during measurement, resulting in less than ideal measurement accuracy. The latter method requires identifying the fringe phase of the target point to obtain the pixel coordinates of the projector corresponding to that point, thus acquiring the 3D information of the target object. It has advantages such as high measurement efficiency and high measurement resolution, and is less affected by the measurement angle, but is more susceptible to noise.

[0004] A structured light 3D measurement method based on phase shift amount encoding fringe order is disclosed in Chinese patent document CN109974626B. This method includes six steps: encoding sinusoidal fringes, calibrating the system, estimating the imaging area, acquiring projected images, determining the fringe order, and establishing the system pixel correspondence. The method described in this invention utilizes phase shift amount to pre-encode the fringe order, and through additional... The phase shift of the frame sinusoidal fringes uniquely encodes the fringe order, reducing the number of images required to encode the fringe order of high-frequency fringes. This method uses phase information as the encoding unit, which has stronger noise resistance compared to the binary fringe encoding method that uses grayscale information as the encoding unit. Moreover, the calculation of the fringe order is highly independent, and it is accelerated by using a graphics processing unit (GPU), which ensures the timeliness of the measurement method. However, this structured light three-dimensional measurement method based on phase shift encoding of fringe order may accumulate errors in highly abrupt regions due to phase unfolding errors.

[0005] A single-camera, multi-projection, dual-axis structured light 3D reconstruction measurement system is disclosed in Chinese patent document CN119554995B. This system includes: projecting structured light onto the object to generate and acquire a dual-axis phase-coded fringe image using a structured light projector; reconstructing a height map of the object based on the image; removing noise from occluded areas based on the installation position of the structured light projector group and the acquired image to obtain a height map of the unobstructed true height region; fusing the true height region height maps to obtain a complete and accurate height map; and obtaining point cloud data of the object based on the height map and performing 3D reconstruction. This invention can project structured light at a 45-degree rotation using a structured light projector and present dual-axis phase-coded fringes on the object, eliminating the effects of shadows and high reflectivity, increasing the effective point cloud quantity, improving the system's measurement accuracy and precision, and obtaining complete and accurate measurement results for the object. However, this single-camera, multi-projection, dual-axis structured light 3D reconstruction measurement system requires four structured light projectors and a rotating support, increasing hardware complexity and calibration difficulty.

[0006] A method for 3D measurement of low-reflectivity workpieces based on structured light is disclosed in Chinese patent document CN118799410A. This method includes building a 3D measurement system for low-reflectivity workpieces based on structured light, capturing images of a calibration board using left and right industrial cameras, performing binocular joint calibration, obtaining the intrinsic and extrinsic parameters of the left and right cameras, capturing sinusoidal fringe images and Gray code encoded images projected onto the low-reflectivity workpiece using the left and right industrial cameras, fusing the conventional phase-shift image group and the high-exposure phase-shift image group obtained in step 2 with guided filtering to obtain a set of high dynamic phase-shift image groups, combining the camera intrinsic and extrinsic parameters, using the principle of triangulation to convert parallax into depth information, reconstructing 3D coordinates, and then obtaining 3D point clouds to complete 3D reconstruction. By using dual-guided phase-shift fusion to compensate for the correct phase of the low-reflectivity workpiece, it is not necessary to estimate the number of exposures and the optimal exposure time required for the measured object, thus improving the integrity of the 3D reconstruction of the low-reflectivity workpiece and speeding up the process. However, this method for 3D measurement of low-reflectivity workpieces based on structured light may have residual errors in the edge areas due to inaccurate phase fusion.

[0007] To address the shortcomings of the existing technologies, providing an adaptive point cloud fusion method based on error assessment is a worthwhile research topic. Summary of the Invention

[0008] The purpose of this invention is to overcome the shortcomings of large measurement errors in edge regions and large noise in phase-shifted point clouds, and to provide an adaptive point cloud fusion method based on error assessment, which achieves the technical effect of reducing measurement errors.

[0009] The objective of this invention is achieved through the following technical solution:

[0010] An adaptive point cloud fusion method based on error evaluation includes the following steps:

[0011] Step 1: Calibrate the camera and projector in the structured light system;

[0012] Step 2: Project binary-coded and phase-shifted sinusoidal-coded stripes using a projector, and capture structured light images using a camera;

[0013] Step 3: Extract the sub-pixel coordinates of the center line of the structured light stripes, calculate the phase features of the phase-shifted coded stripes, and reconstruct two sets of three-dimensional point clouds respectively.

[0014] Step 4: Analyze the incident angle characteristics of the point cloud, the relationship between the pre-evaluated point cloud error and the incident angle, and calculate the fusion weight;

[0015] Step 5: Merge the matching points in the two sets of 3D point clouds. Based on the pixel positions of the 3D points in the 2D image, interpolate the fusion vector of the matching points to merge the non-matching points, thus reducing measurement errors.

[0016] Optionally, the calibration method for the structured light system in step one includes the following steps:

[0017] S1. Project a circular array onto the circular calibration plate, and then simultaneously capture images of the calibration plate with features of both the actual and projected circles using a camera.

[0018] S2. Identify the center pixel coordinates of the circular features on the calibration board.

[0019] S3. Use the least squares method to obtain the parameter matrix of the camera and projector, optimize the calibration process of the structured light system, and ensure the accuracy of the 3D coordinate calculation.

[0020] Optionally, in step two, the width of the binary encoded bright stripes is... The total stripe width is A black and white alternating binary stripe with a 50% duty cycle has a grayscale value that matches:

[0021] ;

[0022] The width of the phase-shifted sinusoidal encoded stripes in each cycle is... The number of phase shift steps is Step, its grayscale conforms to:

[0023] This provides an optimized coding foundation for stripe center extraction and phase calculation, improving the reliability of feature extraction.

[0024] Optionally, the method for extracting the sub-pixel coordinates of the center line in step three includes the following steps:

[0025] S1. Calculate the eigenvalues ​​and eigenvectors of each pixel in the image using the Hessian matrix. The Hessian matrix for any point in the image is represented as:

[0026] ;

[0027] S2. Based on the calculated eigenvalues ​​and eigenvectors, the sub-pixel coordinates of the stripe center are obtained as follows: ,in These are the pixel coordinates at the extreme values ​​of the eigenvalues. For the corresponding feature vector, ;

[0028] S3. The phase calculation method used is as follows: ;

[0029] Ensure the accuracy of 3D point cloud reconstruction.

[0030] Optionally, the pre-evaluation method for the relationship between point cloud error and incident angle in step four includes the following steps:

[0031] S1. Simulate the incident angle α, reflection angle β, and fringe width in a simulation environment. Errors in fringe center extraction under individual influences, among which These are real values, derived from simulation presets. The measured value comes from center extraction of the stripes captured by the camera. The error is the difference between the true value and the measured value, i.e.:

[0032] ;

[0033] S2. The evaluation method for the relationship between phase-shift method error and noise is as follows: Evaluate the impact of noise on the phase-shift method measurement error under the measurement environment, whereby... Indicates signal strength. Indicates noise intensity. To represent the impact of error, in pixels, we have:

[0034] ;

[0035] ;

[0036] ;

[0037] S3, The weight is calculated as follows:

[0038] ;

[0039] The fusion strategy is dynamically adjusted to minimize measurement error.

[0040] Optionally, the method for calculating the matching point in step five is as follows:

[0041] Using the sub-pixel coordinates at the fringe center point, the phase value at the fringe center point is calculated, corresponding to its three-dimensional coordinates under the two measurement methods. and These points are recorded as a set of matching points, providing an accurate correspondence basis for point cloud fusion.

[0042] Optionally, the method for fusing matching points in step five is as follows:

[0043] The matching points between the positions of the fringe center point and the phase-shift point are fused according to weights, that is:

[0044] ,

[0045] The weights are derived from the error prediction based on the pre-error assessment results, using the incident angle.

[0046] Where the matching point fusion vector is ;

[0047] Image pixel coordinates of matching points The index is the fusion vector of the matching points. The coordinates of the point to be interpolated are Interpolation is performed on the fusion vector to obtain the fusion vector of the non-matching points. ;

[0048] The interpolation along a single direction is calculated by the following formula:

[0049] ;

[0050] Extending this to the three directions of three-dimensional coordinates, we can represent them as follows:

[0051] The fusion method for non-matching points is as follows: This solves the problem of merging non-matching points and improves the integrity and accuracy of point clouds.

[0052] Positive and beneficial effects: 1. This adaptive point cloud fusion method based on error assessment significantly reduces measurement errors in edge regions and overall point cloud noise through adaptive weight allocation of errors, thereby improving the accuracy and robustness of 3D measurement. It solves the problems of inaccurate extraction of the stripe center line in edge regions and interference of noise in all regions on phase calculation in the existing structured light 3D measurement process, which further leads to measurement errors.

[0053] 2. This adaptive point cloud fusion method based on error assessment achieves adaptive weight allocation by quantizing the correlation between error and incident angle. This allows the fusion process to prioritize the use of phase shift method in edge regions and fringe center method in low-noise regions, thereby optimizing the overall point cloud quality and dynamically adjusting the fusion strategy to minimize measurement error.

[0054] 3. The adaptive point cloud fusion method based on error evaluation adopts a weighted fusion formula and performs bilinear interpolation on the fusion vector based on the image pixel coordinates of the matching points, extending to the non-matching points, thus achieving seamless fusion of the entire point cloud. It retains the high-resolution advantage of the phase-shifting method and utilizes the edge stability of the fringe center method. Experiments show that the point cloud error is significantly reduced after fusion, solving the fusion problem of non-matching points and improving the integrity and accuracy of the point cloud. Attached Figure Description

[0055] Figure 1 is an overall flowchart of the present invention;

[0056] Figure 2 is an external view of the measurement system of the present invention;

[0057] Figure 3 is a schematic diagram of the system calibration of the present invention;

[0058] Figure 4 is a schematic diagram of the stripe center method of the present invention;

[0059] Figure 5 is a schematic diagram of the phase shift method measurement principle of the present invention;

[0060] Figure 6 is a schematic diagram of the matching points of the present invention;

[0061] Figure 7 is a schematic diagram of the stripe center extraction error of the present invention;

[0062] Figure 8 is a schematic diagram of the interpolation method of the present invention;

[0063] Figure 9 shows the point cloud and error distribution of the fringe center method of the present invention;

[0064] Figure 10 shows the point cloud and error distribution of the phase-shifting method of the present invention;

[0065] Figure 11 shows the point cloud and error distribution after fusion according to the present invention;

[0066] Figure 12 is a comparison of the histograms of error distribution in the fringe center method of the present invention.

[0067] Figure 13 is a comparison of the histograms of the error distribution of the phase-shifting method of the present invention;

[0068] Figure 14 is a histogram of the fusion error distribution of the present invention. Detailed Implementation

[0069] Embodiments of the present invention are described in detail below, examples of which are illustrated in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain the present invention, and should not be construed as limiting the present invention.

[0070] Example 1

[0071] As shown in Figures 1 to 14, an adaptive point cloud fusion method based on error assessment includes the following steps:

[0072] Step 1: Calibrate the camera and projector in the structured light system;

[0073] Step 2: Project binary-coded and phase-shifted sinusoidal-coded stripes using a projector, and capture structured light images using a camera;

[0074] Step 3: Extract the sub-pixel coordinates of the center line of the structured light stripes, calculate the phase features of the phase-shifted coded stripes, and reconstruct two sets of three-dimensional point clouds respectively.

[0075] Step 4: Analyze the incident angle characteristics of the point cloud, the relationship between the pre-evaluated point cloud error and the incident angle, and calculate the fusion weight;

[0076] Step 5: Merge the matching points in the two sets of 3D point clouds. Based on the pixel positions of the 3D points in the 2D image, interpolate the fusion vector of the matching points to merge the non-matching points. Through error adaptive weight allocation, the measurement error in the edge region and the overall point cloud noise are significantly reduced, improving the accuracy and robustness of 3D measurement. This solves the problem of inaccurate extraction of the stripe center line in the edge region and the interference of noise in all regions on the phase calculation in the existing structured light 3D measurement process, which further leads to measurement errors.

[0077] The calibration method for the structured light system in step one includes the following steps:

[0078] S1. Project a circular array onto the circular calibration plate, and then simultaneously capture images of the calibration plate with features of both the actual and projected circles using a camera.

[0079] S2. Identify the center pixel coordinates of the circular features on the calibration board.

[0080] S3. Use the least squares method to obtain the parameter matrices of the camera and the projector;

[0081] Among them, structured light measurement systems, such as Figure 2 As shown, a structured light 3D measurement system is composed of a camera and a projector, which are fixed on a movable support and controlled by a host computer.

[0082] like Figure 3 As shown, in step one of the calibration process, a circular array is first projected onto the circular calibration plate. The projected circular array must not intersect with the actual circle on the calibration plate. Then, an image of the calibration plate is captured. The captured image contains both the actual circle on the calibration plate and the projected circle. The center features of the actual circle and the projected circle are extracted. Finally, based on the pixel position and 3D position of the actual circle, the intrinsic and extrinsic parameters of the camera are obtained. Based on the pixel position and 3D position of the projected circle, the intrinsic and extrinsic parameters of the projector are obtained.

[0083] It improves calibration accuracy and efficiency, provides an accurate system parameter basis for subsequent point cloud reconstruction, reduces cumulative measurement errors caused by calibration errors, optimizes the calibration process of the structured light system, and ensures the accuracy of 3D coordinate calculation.

[0084] In step two, the width of the binary encoded bright stripes is The total stripe width is A black and white alternating binary stripe with a 50% duty cycle has a grayscale value that matches:

[0085] ;

[0086] The width of the phase-shifted sinusoidal encoded fringes in each cycle is... The number of phase shift steps is Step, its grayscale conforms to:

[0087] ,

[0088] Adjust the modulation of the stripes, which includes a base grayscale A and a modulated grayscale B, so that the dark stripes are not underexposed and the bright stripes are not overexposed.

[0089] The pattern is deformed by the shape modulation of the surface of the object under test, and the deformed image is captured by a camera.

[0090] Binary encoding provides clear fringe ordinal numbers, while phase-shift encoding enables high-resolution phase extraction. The combination of the two enhances the robustness of fringe features, reduces the impact of noise during image acquisition, provides an optimized coding foundation for fringe center extraction and phase calculation, and improves the reliability of feature extraction.

[0091] Example 2

[0092] The method for extracting the sub-pixel coordinates of the center line in step three includes the following steps:

[0093] S1. Calculate the eigenvalues ​​and eigenvectors of each pixel in the image using the Hessian matrix. The Hessian matrix for any point in the image is represented as:

[0094] ;

[0095] S2. Based on the calculated eigenvalues ​​and eigenvectors, the sub-pixel coordinates of the stripe center are obtained as follows: ,in These are the pixel coordinates at the extreme values ​​of the eigenvalues. For the corresponding features vector, S3. The phase calculation method used is as follows:

[0096] ;

[0097] Sine-coded stripes are used to extract the center and phase features of the stripes, while binary coding is used to obtain the ordinal numbers of the center and phase features of the stripes.

[0098] like Figure 4 The pixel coordinates at the center of the stripes are obtained by using the stripe center extraction algorithm. Combined with the stripe ordinal number, a set of three-dimensional point clouds reconstructed by the stripe center method can be calculated.

[0099] like Figure 5 The phase extraction algorithm can be used to obtain the phase at each pixel, and then mapped to the projector pixel coordinates at each pixel, so as to calculate another set of three-dimensional point clouds reconstructed by the phase shift method.

[0100] This method improves the accuracy of fringe center localization and phase calculation, reduces the discrete error of reconstructed point clouds, provides high-quality input for subsequent fusion, solves the problems of inaccurate fringe center extraction and phase noise in traditional methods, and ensures the accuracy of 3D point cloud reconstruction.

[0101] like Figure 6 As shown, the pre-evaluation method for the relationship between point cloud error and incident angle in step four includes the following steps:

[0102] S1. Simulate the incident angle α, reflection angle β, and fringe width in a simulation environment. Errors in fringe center extraction under individual influences, among which These are real values, derived from simulation presets. The measured value comes from center extraction of the stripes captured by the camera. The error is the difference between the true value and the measured value, i.e.:

[0103] ;

[0104] S2. The evaluation method for the relationship between phase-shift method error and noise is as follows: Evaluate the impact of noise on the phase-shift method measurement error under the measurement environment, whereby... Indicates signal strength. Indicates noise intensity. To represent the impact of error, in pixels, we have:

[0105] ;

[0106] ;

[0107] ;

[0108] S3, The weight is calculated as follows:

[0109] By correlating quantization error with incident angle, adaptive weight allocation is achieved, enabling the fusion process to prioritize the use of phase shift method in edge regions and fringe center method in low-noise regions, thereby optimizing the overall point cloud quality and dynamically adjusting the fusion strategy to minimize measurement error.

[0110] Example 3

[0111] like Figure 7 As shown, the method for calculating the matching points in step five is as follows:

[0112] Using the sub-pixel coordinates at the fringe center point, the phase value at the fringe center point is calculated, corresponding to its three-dimensional coordinates under the two measurement methods. and These points are recorded as a set of matching points, ensuring the consistency of the fusion basis, avoiding point cloud misalignment problems, improving fusion efficiency, and providing an accurate correspondence basis for point cloud fusion.

[0113] The method for merging matching points in step five is as follows:

[0114] The matching points between the positions of the fringe center point and the phase-shift point are fused according to weights, that is:

[0115] ,

[0116] The weights are derived from the error prediction based on the pre-error assessment results, using the incident angle.

[0117] Where the matching point fusion vector is ;

[0118] like Figure 8 As shown, the image pixel coordinates of the matching points The index is the fusion vector of the matching points. The coordinates of the point to be interpolated are Interpolation is performed on the fusion vector to obtain the fusion vector of the non-matching points. ;

[0119] The interpolation along a single direction is calculated by the following formula:

[0120] Extending this to the three directions of three-dimensional coordinates, we can represent them as follows:

[0121]

[0122] The fusion method for non-matching points is as follows: It employs a weighted fusion formula and is based on the image pixel coordinates of the matching points. Bilinear interpolation is applied to the fusion vector to extend it to non-matching points, achieving seamless fusion of the entire point cloud. This retains the high-resolution advantage of the phase-shifting method while utilizing the edge stability of the fringe center method. Experiments show that the point cloud error is significantly reduced after fusion, solving the problem of fusion of non-matching points and improving the integrity and accuracy of the point cloud.

[0123] Example 4

[0124] To verify the effectiveness of the proposed method, the radius of a bearing steel ball conforming to the G10 standard in GB / T 308.1-2013 was measured. The process error of the radius of the steel ball of this specification is within 0.0125mm.

[0125] The point cloud obtained by fringing center method, phase shift method, and fusion with this method is shown below. Figure 9 Figure 11 shows that the point cloud error comes from the distance of each point in the point cloud to the fitted sphere, and each point is represented by a different color according to the error.

[0126] As can be seen in Figure 9 The area outlined in the middle represents the region of large error in the fringe center method at large angles, while... Figure 10 There are no significant error points at the same positions in the middle;

[0127] Figure 10 shows error points of different colors at various locations, indicating that all locations are affected by noise.

[0128] Figure 11 shows that point cloud fusion using this method reduces both the error at large incident angles and the noise impact at various locations.

[0129] The distribution of the statistical error histogram is shown in Figures 12-14. Figure 12 shows that in this measurement, the error of the sparse point cloud of the fringe center method is concentrated within 0.15 mm, but the error reaches 0.45 mm under large incident angles.

[0130] Figure 13 shows that the measurement error of the phase shift method in this measurement is concentrated within 0.3 mm;

[0131] Figure 14 shows that the error after point cloud fusion in this measurement is concentrated within 0.15 mm.

[0132] Therefore, the method proposed in this paper can effectively solve the problem of inaccurate measurement in the edge region by the fringe center method in existing structured light 3D measurement, and reduce the noise of point cloud generated by the phase shift method.

[0133] The above is only used to illustrate the technical solution of the present invention and not to limit it. Any other modifications or equivalent substitutions made by those skilled in the art to the technical solution of the present invention, as long as they do not depart from the spirit and scope of the technical solution of the present invention, should be covered within the scope of the claims of the present invention.

Claims

1. An adaptive point cloud fusion method based on error evaluation, characterized in that, Includes the following steps: Step 1: Calibrate the camera and projector in the structured light system; Step 2: Project binary-coded and phase-shifted sinusoidal-coded stripes using a projector, and capture structured light images using a camera; Step 3: Extract the sub-pixel coordinates of the center line of the structured light stripes, calculate the phase features of the phase-shifted coded stripes, and reconstruct two sets of three-dimensional point clouds respectively. Step 4: Analyze the incident angle characteristics of the point cloud, the relationship between the pre-evaluated point cloud error and the incident angle, and calculate the fusion weight; Step 5: Merge the matching points in the two sets of 3D point clouds. Based on the pixel positions of the 3D points in the 2D image, interpolate the fusion vector of the matching points to merge the non-matching points.

2. The adaptive point cloud fusion method based on error evaluation according to claim 1, characterized in that, The calibration method for the structured light system in step one includes the following steps: S1. Project a circular array onto the circular calibration plate, and then simultaneously capture images of the calibration plate with features of both the actual and projected circles using a camera. S2. Identify the center pixel coordinates of the circular features on the calibration board. S3. Use the least squares method to obtain the parameter matrices of the camera and the projector.

3. The adaptive point cloud fusion method based on error evaluation according to claim 1, characterized in that: In step two, the width of the binary encoded bright stripes is The total stripe width is A black and white alternating binary stripe with a 50% duty cycle has a grayscale value that matches: ； The width of the phase-shifted sinusoidal encoded stripes in each cycle is [missing information]. The number of phase shift steps is Step, its grayscale conforms to: 。 4. The adaptive point cloud fusion method based on error evaluation according to claim 1, characterized in that, The method for extracting the sub-pixel coordinates of the center line in step three includes the following steps: S1. Calculate the eigenvalues ​​and eigenvectors of each pixel in the image using the Hessian matrix. The Hessian matrix for any point in the image is represented as: ； S2. Based on the calculated eigenvalues ​​and eigenvectors, the sub-pixel coordinates of the stripe center are obtained as follows: ,in These are the pixel coordinates at the extreme values ​​of the eigenvalues. For the corresponding feature vector, ; S3. The phase calculation method used is as follows: 。 5. The adaptive point cloud fusion method based on error evaluation according to claim 1, characterized in that, The pre-evaluation method for the relationship between point cloud error and incident angle in step four includes the following steps: S1. Simulate the incident angle α, reflection angle β, and fringe width in a simulation environment. Errors in fringe center extraction under individual influences, among which These are real values, derived from simulation presets. The measured value comes from center extraction of the stripes captured by the camera. The error is the difference between the true value and the measured value, i.e.: ; S2. The evaluation method for the relationship between phase-shift method error and noise is as follows: Evaluate the impact of noise on the phase-shift method measurement error under the measurement environment, whereby... Indicates signal strength. Indicates noise intensity. To represent the impact of error, in pixels, we have: ; ; ; S3, The weight is calculated as follows: 。 6. The adaptive point cloud fusion method based on error evaluation according to claim 1, characterized in that, The method for calculating the matching point in step five is as follows: Using the sub-pixel coordinates at the fringe center point, the phase value at the fringe center point is calculated, corresponding to its three-dimensional coordinates under the two measurement methods. and , and denote them as a set of matching points.

7. The adaptive point cloud fusion method based on error evaluation according to claim 1, characterized in that, The method for fusing matching points in step five is as follows: The matching points between the fringe center point and the phase-shift point are weighted and fused, i.e.: ， The weights are derived from the error prediction based on the pre-error assessment results, using the incident angle. Where the matching point fusion vector is ; Image pixel coordinates of matching points The index is the fusion vector of the matching points. The coordinates of the point to be interpolated are Interpolation is performed on the fusion vector to obtain the fusion vector of the non-matching points. ; The interpolation along a single direction is calculated by the following formula: ； ； Extending this to the three directions of three-dimensional coordinates, we can represent them as follows: ； The fusion method for non-matching points is as follows: .

Images