A sharp edge surface three-dimensional reconstruction method based on line laser oblique cutting scanning

Abstract

To address the technical challenge of achieving low-cost, high-precision, and high-efficiency measurement and 3D reconstruction of compressor blade inlet edges using existing measurement methods, this invention proposes a 3D reconstruction method for sharp edge surfaces based on line laser oblique scanning. This invention provides theoretical support for increasing the number of measurement points by establishing a quantification model for the number of measurement points in line laser oblique scanning; it provides accurate quantification of the high reflectivity and complex characteristics of the sharp edge surface by establishing a quantification model for the reflectivity of the surface; it achieves synergistic optimization of the oblique angle by establishing an optimization model for the oblique angle, thus balancing the increase in the number of measurement points with the impact of high reflectivity; and finally, it uses bidirectional oblique scanning and point cloud processing to achieve 3D reconstruction of the sharp edge surface. This invention provides a feasible solution for achieving low-cost, high-precision, and high-efficiency measurement of sharp edge surfaces.

Description

Technical Field

[0001] This invention belongs to the field of image data processing, and specifically relates to a method for three-dimensional reconstruction of sharp-edged surfaces. Background Technology

[0002] Compressor blades are core components of aero-engines. To meet the demands for lightweight design and high aerodynamic performance, blades are typically designed as thin-walled curved surfaces with high geometric precision and complex profiles, manufactured using difficult-to-machine materials such as titanium alloys or high-temperature alloys. Therefore, the machining process presents significant challenges. The geometric features of a blade mainly include the inlet edge, exhaust edge, blade head, and blade back. Among these, the inlet edge, located on the blade inlet side, has the most significant impact on aerodynamic performance; even minor machining errors can cause significant fluctuations in the aero-engine's aerodynamic performance. However, this region is thinner, has more drastic curvature changes, and is more difficult to machine, making it highly susceptible to deviations from tolerances.

[0003] Besides relying on strict manufacturing processes, measurement technology is also an indispensable means of controlling blade machining errors. Currently, blade inspection mainly relies on contact probes to sample point-by-point on specific two-dimensional target sections, with commonly used equipment including coordinate measuring machines (CMMs) and in-machine measurement systems. While this method offers high accuracy, it is inefficient. With the development of next-generation aero-engines, the inlet edge size of compressor blades has been further reduced, resulting in sharp edges, typical of pointed surfaces. Due to the rapid curvature changes at such extremely small dimensions, contact probes suffer from difficulties in precise positioning and large cosine errors, making accurate measurement challenging. To address the small-sized features such as tool edges, the Austrian company Alicona developed an automatic zoom scanning device called InfiniteFocus. This device utilizes focus-changing technology to achieve high-precision three-dimensional reconstruction of pointed surfaces, but it has limitations in terms of efficiency, flexibility, and cost.

[0004] In recent years, structured light measurement technology has gained widespread attention in the inspection of aero-engine blades due to its flexibility and efficiency. Based on the different light morphologies, it can be divided into three types: point light, line light, and surface light. Point light measurement offers high accuracy and low cost, but its point-by-point sampling results in low efficiency. Surface light measurement can obtain a large-scale point cloud of the blade's three-dimensional profile in a single acquisition, but its accuracy is lower, its cost is higher, and it is prone to point cloud gaps in sharp edge areas, making it primarily used for the inspection of large-sized fan blades. In contrast, line light measurement achieves a good balance between accuracy, efficiency, and cost, possessing comprehensive advantages.

[0005] In blade line laser measurement, the measurement plane is typically parallel to the target cross-section to directly acquire the cross-sectional point cloud for comparative evaluation. However, this measurement method presents two key problems when applied to the inlet edge of compressor blades:

[0006] 1. Due to the limited thickness of the air intake edge and the resolution of the line laser sensor, the surface measurement points are sparse, making it difficult to accurately represent the true profile.

[0007] 2. Although the measurement and evaluation of specific two-dimensional target sections of the blades conforms to industry standards, only limited contour information can be obtained, which poses a potential risk of missed detection, especially in the critical area of ​​the air intake edge.

[0008] Therefore, there is an urgent need to study a low-cost, high-precision, and high-efficiency method for measuring and reconstructing sharp edge surfaces to effectively support the control and evaluation of machining errors at the inlet edge of compressor blades. Summary of the Invention

[0009] To address the technical challenge of achieving low-cost, high-precision, and high-efficiency measurement and three-dimensional reconstruction of the inlet edge of compressor blades using existing measurement methods, this invention proposes a three-dimensional reconstruction method for the sharp edge surface based on line laser oblique scanning.

[0010] The technical solution of this invention is:

[0011] A method for 3D reconstruction of sharp-edge surfaces based on line laser oblique scanning, characterized by the following steps:

[0012] The surface to be measured is projected onto a two-dimensional plane to obtain the projection area. A quantitative model for the number of measurement points in the line laser oblique scanning is established to describe the relationship between the number of measurement points in the projection area and the oblique angle. A quantitative model for the reflected light intensity of the surface to be measured is established to describe the relationship between the reflected light intensity of the surface to be measured and the oblique angle. The two-dimensional plane is a plane perpendicular to the beam axis of the line laser sensor in the parallel scanning pose. The intersection of the light plane of the line laser sensor and the surface to be measured in the parallel scanning pose is defined as the cross section. In the oblique scanning pose, the intersection of the light plane of the line laser sensor and the surface to be measured is defined as the oblique section. The oblique angle refers to the angle between the oblique section and the cross section.

[0013] Based on the quantitative model of the number of measurement points in the line laser oblique scanning and the quantitative model of the reflected light intensity of the sharp edge surface, an optimization model for the oblique angle is constructed and solved to obtain the optimal oblique angle.

[0014] Based on the optimized oblique cutting angle and employing a bidirectional scanning strategy, continuous line laser oblique cutting scanning is performed on the surface to be measured to obtain the oblique cutting measurement point cloud. and Continuous parallel line laser scanning is performed on the surface of the sharp edge to be measured to obtain a reference point cloud. Oblique measurement point cloud and and benchmark cloud All are located in the line laser sensor base coordinate system; the line laser sensor base coordinate system refers to the coordinate system of the mechanical device used to control the movement of the line laser sensor;

[0015] Based on benchmark point cloud Using the nearest neighbor concept, we eliminate obliquely tangent measurement point clouds. and From the anomaly point cloud, the oblique measurement point cloud is obtained. and ;

[0016] Based on benchmark point cloud Oblique measurement point cloud and After precise registration, duplicate point clouds in overlapping areas are removed to obtain the point cloud of the sharp edge surface to be measured. This enables three-dimensional reconstruction of the sharp edge surface to be measured.

[0017] Furthermore, the method for constructing the intensity quantization model of reflected light from sharp-edge surfaces is as follows:

[0018] Images of line laser stripes on the surface of the sharp edge under different bevel angles are acquired, and the global normalized grayscale sum of all pixels in each line laser stripe image is calculated.

[0019] A double exponential function is used to fit the global normalized gray sum of all pixels in the image of the chamfered angle and the corresponding line laser stripe to obtain a quantitative model of the reflected light intensity of the sharp edge surface.

[0020] Furthermore, when using a double exponential function to fit and establish a quantitative model of reflected light intensity on a sharp edge surface, it is also necessary to center and scale the chamfer angle.

[0021] Furthermore, the method for constructing the oblique angle optimization model is as follows:

[0022] The logarithmic form of the quantization model for the number of measurement points in a line laser oblique scanning is normalized to obtain the function. The quantization model of reflected light intensity from the sharp-edge surface is normalized to obtain the function. ;

[0023] For functions and Weighted summation yields the objective function used to solve for the optimal oblique angle; It is a beveled angle;

[0024] With minimizing the objective function as the solution objective and the range of values ​​for the chamfer angle as the constraint, a chamfer angle optimization model is constructed.

[0025] Furthermore, for symmetrical sharp edges, the range of values ​​for the oblique angle is... The quantization model for the number of measurement points in the line laser oblique scanning, the quantization model for the reflected light intensity of the sharp edge surface, and the optimization model for the oblique angle are all in... Constructed within the range of the chamfer angle; for asymmetric sharp edges, the chamfer angle ranges from [value missing]. The quantization model for the number of measurement points in the line laser oblique scanning, the quantization model for the reflected light intensity of the sharp edge surface, and the optimization model for the oblique angle are all respectively in... and Constructed within the range of the oblique angle.

[0026] Furthermore, the bidirectional scanning strategy refers to:

[0027] For symmetrical sharp edges, the preferred chamfer angle is negatively determined, and chamfer scanning is performed based on the preferred chamfer angle and the chamfer angle after taking the negative value, respectively.

[0028] For asymmetric sharp edges, based on respectively The first preferred chamfer angle of the range and The second preferred bevel angle within the range is used to perform a bevel scan.

[0029] Furthermore, based on the benchmark point cloud Using the nearest neighbor concept, we eliminate obliquely tangent measurement point clouds. and The method for handling anomalous point clouds is:

[0030] Calculate the distances from each measuring point in the oblique measurement point cloud to each measuring point in the reference point cloud, and obtain the shortest distance and the shortest point vector;

[0031] Abnormal point cloud identification and handling: When the nearest distance to a certain measuring point in the oblique measurement point cloud is greater than the set distance threshold, and the angle between the nearest point vector and the specified direction is less than 90°, the measuring point is identified as an abnormal point cloud and is removed; otherwise, the measuring point is a non-abnormal point cloud and is retained.

[0032] The specified direction refers to the direction of the beam axis of the line laser sensor in the base coordinate system of the line laser sensor under parallel scanning pose.

[0033] Furthermore, based on the benchmark point cloud Oblique measurement point cloud and The method for fine registration is:

[0034] Construct point clouds separately , , The axis is aligned with the bounding box;

[0035] Point cloud and Axial alignment bounding box, point cloud and Intersection region detection is performed using axis-aligned bounding boxes to obtain overlapping region point clouds. , , and ,in, Represents the benchmark point cloud The point cloud located in the intersecting region, Represents oblique measurement point cloud The point cloud located in the intersecting region, Represents the benchmark point cloud The point cloud located in the intersecting region, Represents oblique measurement point cloud Point clouds located in the intersecting region;

[0036] Point clouds of overlapping regions , , and Voxel downsampling was performed separately to obtain point clouds. , , , ;

[0037] Using the coherent point drift algorithm, the point cloud Register to point cloud Obtain the transformation matrix , will dot clouds Register to point cloud Obtain the transformation matrix ;

[0038] Transformation matrix , Acting on oblique measurement point clouds respectively and Obtain the point cloud after fine registration and .

[0039] Furthermore, after fine registration, the method for deduplicating point clouds in overlapping regions is as follows:

[0040] Point cloud based on axis-aligned bounding box and Perform intersection region detection to obtain the overlapping area between the two.

[0041] Remove point clouds or The point cloud located within the overlapping region.

[0042] The beneficial effects of this invention are as follows:

[0043] This invention employs oblique scanning to increase the number of measurement points in sharp-edge regions. However, this approach presents a technical challenge: while increasing the number of measurement points, it exacerbates high reflectivity, leading to abnormal point cloud interference in the obtained oblique measurement point cloud and affecting measurement accuracy. Although existing point cloud processing methods exist for highly reflective objects, they cannot be applied to line laser oblique measurement because they do not consider the balance between increasing the number of measurement points and intensifying high reflectivity. Therefore, this invention overcomes this technical challenge through the following technological innovations.

[0044] 1. A quantitative model for the number of measurement points in line laser oblique scanning was established.

[0045] Unlike scanning methods where the optical plane is parallel to the target cross-section, this invention uses directional control of the line laser sensor's pose to transform the cross-section formed by its optical plane and the surface to be measured (the sharp edge) into an oblique section. This expands the lateral scale of the contour within the optical plane, thereby increasing the number of measurement points on the sharp edge surface. Based on this, a quantitative model for the number of measurement points in line laser oblique scanning is established through theoretical derivation, enabling a quantitative analysis of the impact of changes in the oblique angle on the increase in the number of measurement points on the sharp edge surface.

[0046] 2. A quantitative model for the reflected light intensity of a sharp-edge surface was established.

[0047] To address the problem that the increased oblique angle in line laser scanning exacerbates the high reflectivity of sharp-edge surfaces and causes severe interference from anomalous point clouds in the measurement data, this invention establishes a quantitative model of the reflected light intensity of sharp-edge surfaces through line laser light stripe image processing technology and double exponential fitting. This enables a quantitative analysis of the complex influence of oblique angle changes on the reflected light intensity of sharp-edge surfaces.

[0048] 3. An optimization model for the oblique cutting angle was established.

[0049] By combining the logarithmic form of the normalized linear laser oblique scanning measurement point quantification model with the normalized sharp-edge surface reflective light intensity quantification model using logarithmic transformation and weighting, an oblique angle optimization model is formed. Solving this optimization model yields an optimal oblique angle. Performing linear laser oblique scanning on sharp-edge surfaces based on this optimal angle achieves a balance between increasing the number of measurement points and exacerbating high reflectivity.

[0050] 4. A method for laser oblique scanning and point cloud data processing of sharp-edge surface lines was established.

[0051] To reduce interference from anomalous point clouds in oblique cutting measurement point clouds, oblique cutting measurement point clouds of sharp-edge surfaces are obtained through bidirectional oblique cutting scanning. A method is proposed to remove outliers and perform fine registration in oblique cutting measurement point clouds by using the point cloud obtained from parallel scanning as the reference point cloud, thereby improving the accuracy of line laser oblique cutting measurement point clouds of sharp-edge surfaces.

[0052] Among the aforementioned technological innovations: the linear laser oblique scanning measurement point quantification model provides theoretical support for increasing the number of measurement points; the sharp edge surface reflection intensity quantification model provides accurate quantification for the complex characteristics of high reflectivity; the oblique angle optimization model achieves synergistic optimization of increasing the number of measurement points and the influence of high reflectivity; and the bidirectional oblique scanning and point cloud processing method is used to realize the three-dimensional reconstruction of the sharp edge surface, ultimately providing a feasible solution for achieving low-cost, high-precision, and high-efficiency measurement of sharp edge surfaces.

[0053] Furthermore, the three-dimensional reconstruction results of the sharp edge surface obtained by this invention can avoid the risk of missed detection of sharp edge surfaces under current testing standards. Attached Figure Description

[0054] Figure 1 This is a flowchart of the method of the present invention.

[0055] Figure 2 This is a schematic diagram of the optical plane for parallel scanning of a line laser in an embodiment of the present invention.

[0056] Figure 3 This is a schematic diagram illustrating the conversion of the linear laser sensor from parallel scanning to oblique scanning in an embodiment of the present invention.

[0057] Figure 4 This is a schematic diagram of the optical plane during line laser oblique scanning in an embodiment of the present invention.

[0058] Figure 5 This is a schematic diagram (two-dimensional plane) of the measuring points on the inlet side surface of the compressor blade during linear laser parallel scanning in an embodiment of the present invention.

[0059] Figure 6 This is a schematic diagram (two-dimensional plane) of the measuring points on the inlet side surface of the compressor blade during linear laser oblique scanning in an embodiment of the present invention.

[0060] Figure 7 This is a schematic diagram (two-dimensional plane) of the scanning area during linear laser oblique scanning in an embodiment of the present invention.

[0061] Figure 8 This is the three-dimensional reconstruction result of the compressor blade inlet edge region in an embodiment of the present invention.

[0062] Explanation of reference numerals in the attached figures:

[0063] 1-Linear laser sensor; 2-Compressor blade inlet edge. Detailed Implementation

[0064] The following, in conjunction with the accompanying drawings, uses the measurement of the inlet edge of the compressor blade cascade in an aero-engine as an example to provide a detailed explanation of the three-dimensional reconstruction method for sharp edge surfaces based on line laser oblique scanning proposed in this invention.

[0065] Reference Figure 1A method for three-dimensional reconstruction of sharp-edge surfaces based on line laser oblique scanning, specifically including the following steps:

[0066] Step 1: Establish a quantitative model for the number of measurement points in line laser oblique scanning.

[0067] See Figure 2 Define the light plane of line laser sensor 1 in the initial measurement pose of parallel scanning as... , For the beam axis of the line laser sensor, at this time, the light plane The cross section intersecting with the compressor blade inlet edge 2 is the cross-section. See also... Figure 3 When the line laser sensor 1 rotates around z s0 Shaft rotation hour, The light plane under the corresponding measurement pose is converted into See also Figure 4 Light plane The section intersecting with the compressor blade inlet edge 2 is an oblique section. Defined as the oblique angle, it is also the angle between the oblique section and the cross section.

[0068] Define along y s0 The axially moving line laser sensor 1 passes through the optical plane. The method for scanning the compressor blade inlet edge 2 is line laser parallel scanning; the y-axis is defined as... s0 The axially moving line laser sensor 1 passes through the optical plane. The compressor blade inlet edge 2 is scanned using a line laser oblique cut scanning method. Where, y s0 Axis based on light plane And determined by the right-hand rule.

[0069] In line laser measurement, two adjacent measurement points are defined as follows: The theoretical distance along the axis is d Re Linear laser sensor 1 along y s0 The sampling interval during axial movement is d Sa (d) Sa Values ​​related to d Re (equal), the thickness of the compressor blade inlet edge 2 is a, and the height is b (a and b can be obtained based on the compressor blade design model).

[0070] The compressor blade inlet edge 2 is projected onto a two-dimensional plane to obtain a rectangular projection area. The horizontal dimension of the rectangular projection area corresponds to the inlet edge thickness a, and the vertical dimension of the rectangular projection area corresponds to the moving direction of the line laser sensor 1 and the inlet edge height b, respectively.

[0071] The two-dimensional plane is Figure 2The o-yz plane in the o-xyz base coordinate system of the line laser sensor; the coordinate system in the o-xyz base coordinate system of the line laser sensor under the parallel scanning pose of the line laser sensor. In Plane parallelism, the coordinate system of the o-xy plane and the line laser sensor under parallel scanning pose. In The plane (i.e., the light plane when the line laser sensor performs parallel scanning) is parallel. In other words, the two-dimensional plane is perpendicular to the beam axis of the line laser sensor in its parallel scanning pose. The plane.

[0072] The line laser sensor base coordinate system o-xyz is the coordinate system of the mechanical device used to control the movement of the line laser sensor 1. The mechanical device can be a CNC machine tool or an industrial robot.

[0073] After projection, the light plane of the line laser sensor 1 is obliquely scanned. The cross section intersecting with the compressor blade inlet edge 2 is then transformed into a smooth plane. A straight line that intersects the rectangular projection area.

[0074] Figure 5 , Figure 6 , Figure 7 The figures show schematic diagrams of measuring points on the compressor blade inlet edge surface during parallel linear laser scanning, oblique linear laser scanning, and the scanned area during oblique linear laser scanning within a two-dimensional plane (o-yz). As can be seen from the figures, the quantitative relationship between the number of measuring points and the oblique angle can be described using a mathematical model constructed based on trigonometric functions. Therefore, a quantitative model for the number of measuring points in oblique linear laser scanning is established within the two-dimensional plane (o-yz):

[0075] When using a beveled angle When scanning a rectangular projection area of ​​width a and height b, and requiring the scanned area S to completely cover the projected rectangular area, we can obtain:

[0076]

[0077] Based on the above formula, the following is introduced for adjacent measuring points: Theoretical distance d in the axial direction Re With line laser sensor 1 along y s0 Sampling interval d during axial movement Sa By further calculating the number of measuring points within the scanning area S, the measuring point density E can be derived. Den With the oblique angle The changing relationship:

[0078]

[0079] The area of ​​the rectangular projection region and the density of measuring points E are used to... Den The product of these two numbers yields the number of measuring points within the rectangular projection area under oblique measurement. With the oblique angle The changing relationship, i.e., the quantitative model of the number of measurement points in line laser oblique scanning:

[0080]

[0081] Since the compressor blade inlet edge 2 in this embodiment has a symmetrical structure, the same value but different directions of the oblique angle have the same effect on the number of measuring points in the projected area. Therefore, to simplify data processing, the range of oblique angle values ​​can be narrowed down to... Only need to A quantitative model of the number of measurement points for line laser oblique scanning is established within the oblique angle range.

[0082] In other embodiments, if the edge to be measured is asymmetrical, the range of the bevel angle is not narrowed; instead, it needs to be... and A quantitative model of the number of linear laser oblique scanning measurement points was established within the oblique angle range.

[0083] In other embodiments, if the projected area obtained by projecting the surface of the sharp edge to be measured onto the two-dimensional plane o-yz is not rectangular, those skilled in the art can still establish the measurement point density E by referring to the principles of the above method and using existing mathematical methods. Den With the oblique angle The changing relationship.

[0084] Step 2: Establish a quantitative model of the reflected light intensity on the inlet side surface of the compressor blade cascade.

[0085] In online laser measurement, the grayscale of the light stripe image is a digital representation of the light intensity on the surface of the object being measured. Therefore, the globally normalized grayscale of all pixels in the line laser light stripe image can be used to characterize the reflected light intensity.

[0086] Control line laser sensor 1 from initial measurement pose (light plane is Starting from z, circle around z s0 Shaft rotation bevel angle Collect different bevel angles Linear laser stripe image of the inlet edge 2 of the lower compressor blade cascade Thus, line lasers can be obtained at different bevel angles. The image of the line laser stripe on the surface of the compressor blade inlet edge 2 during the measurement.

[0087] Define the collected image set as , ,in, Indicates the bevel angle Corresponding line laser stripe image The grayscale value of the pixel in the m-th row and n-th column is given, where M and N represent the number of pixels in the height and width directions of the image, respectively. The linear laser stripe image is calculated using the following formula. Globally normalized grayscale sum of all pixels in:

[0088]

[0089] Where 255 represents the maximum pixel grayscale value in the line laser light stripe image.

[0090] Using a double exponential function to measure the chamfer angle With global normalized grayscale By fitting the data, a quantitative model of the reflected light intensity from the inlet side surface of the compressor blades is formed:

[0091]

[0092] The above-mentioned quantitative model for the reflected light intensity from the inlet side surface of the compressor blade cascade is used to describe the change in reflected light intensity from the inlet side surface of the compressor blade cascade with the oblique angle. The changing relationship.

[0093] Collected different oblique angles and the corresponding line laser stripe image Globally normalized grayscale values ​​of all pixels in the range Substituting these values ​​into the above equation, we can obtain the coefficients A, B, C, and D in the quantitative model of reflected light intensity on the inlet side surface of the compressor blade.

[0094] To improve the stability of numerical calculations, a double exponential function can be used to adjust the angle of the oblique section during fitting. After centering and scaling, the quantification model of reflected light intensity from the inlet edge surface of the compressor blades can be further expressed as:

[0095]

[0096] In the formula, mean represents the mean, std represents the standard deviation, and A1, B1, C1, and D1 are the angles of tangency. The coefficients in the quantization model of reflected light intensity from the inlet side surface of the compressor blades, after centering and scaling.

[0097] Collected different oblique angles and the corresponding line laser stripe image Globally normalized grayscale values ​​of all pixels in the range Substituting these values ​​into the above formula, we can obtain the coefficients A1, B1, C1, and D1 in the quantitative model of reflected light intensity on the inlet side surface of the compressor blade.

[0098] Steps 1 and 2 above can be interchanged.

[0099] In this embodiment, the compressor blade inlet edge 2 is a symmetrical pointed edge; therefore, it is only necessary to... A quantitative model of the reflected light intensity on the inlet side surface of the compressor blade cascade is established within the range of the oblique cutting angle.

[0100] In other embodiments, if the surface of the sharp edge to be tested has an asymmetric structure, it is necessary to follow the method in step 2 of this embodiment to respectively... and A quantitative model of the reflected light intensity of the sharp edge surface is established within the range of the oblique angle.

[0101] Step 3: Establish and solve the oblique angle optimization model to obtain the optimal oblique angle.

[0102] Based on the quantification model of the number of measurement points in line laser oblique scanning and the quantification model of the reflected light intensity on the inlet side surface of the compressor blade, it can be seen that as the oblique angle increases... As the magnitude of light increases, both the number of measuring points and the intensity of reflected light within the projection area show an upward trend, especially the number of measuring points. The range of values ​​is too large, when the bevel angle is too large. When the angle approaches 90 degrees, the number of measuring points approaches infinity. Therefore, a logarithmic transformation is used to numerically compress the quantification model of the number of measuring points in linear laser oblique scanning, allowing for different oblique angles to be measured. The change in the number of measurement points is more stable, that is, the logarithmic form of the quantitative model of the number of measurement points in line laser oblique scanning is obtained through logarithmic transformation. Then, the logarithmic form of the quantification model for the number of measurement points in the linear laser oblique scanning and the quantification model for the reflected light intensity on the inlet side surface of the compressor blade cascade were normalized respectively. and Mapping to the range [0, 1], we get:

[0103]

[0104]

[0105] For laser oblique scanning of the inlet edge surface of compressor blades, this invention aims to obtain a higher number of measurement points and a lower reflected light intensity. Therefore, The higher the value, the better. The smaller the value, the better.

[0106] Based on the concept of weighting, an objective function is established to solve for the optimal oblique angle:

[0107]

[0108] In the formula, This is a weighting coefficient, with a value range of (0, 1). In this embodiment... The value is 0.5.

[0109] The chamfer angle optimization model is represented as follows:

[0110]

[0111] in, This indicates a constraint condition.

[0112] The optimal cut angle model can be solved by traversing the search path, thus obtaining the preferred cut angle for the compressor blade inlet edge in line laser cut angle scanning measurement. .

[0113] In this embodiment, the compressor blade inlet edge 2 is a symmetrical pointed edge; therefore, it is only necessary to... Establish and solve the cut angle optimization model within the range of the cut angle.

[0114] In other embodiments, if the surface of the sharp edge to be tested has an asymmetric structure, it is necessary to follow the method in step 3 of this embodiment to respectively... and Establish and solve the oblique angle optimization model within the oblique angle range, and obtain the results for each oblique angle range. and The two preferred chamfer angles within the range are denoted as the first preferred chamfer angle. Second preferred bevel angle .

[0115] Step 4: Based on the optimized oblique cutting angle and using a bidirectional scanning strategy, perform continuous line laser oblique cutting scans on the inlet edge surface of the compressor blades to obtain the oblique cutting measurement point cloud. and Linear laser parallel scanning was performed on the inlet edge surface of the compressor blades to obtain a reference point cloud. .

[0116] Oblique scan:

[0117] Due to the high reflectivity during the oblique scanning process, severe anomalous point clouds appear on one side of the compressor blade inlet edge surface. To reduce interference from these anomalous point clouds, a bidirectional scanning strategy is adopted.

[0118] For symmetrical sharp edges (as in this embodiment), a preferred bevel angle is preferred. Take negative values, based on the preferred chamfer angle. The angle of the cut after taking the negative value Perform a slant scan.

[0119] Define the preferred bevel angle The corresponding optical plane of the line laser sensor is , along y s0 The axially moving line laser sensor 1 passes through the optical plane. Oblique scanning is performed on the inlet edge of the compressor blade cascade to obtain oblique measurement point cloud. .

[0120] Define the chamfer angle after taking the negative value. The corresponding optical plane of the line laser sensor is , along y s0 The axially moving line laser sensor 1 passes through the optical plane. Perform a second oblique scan on the inlet edge of the compressor blades to obtain the oblique measurement point cloud. .

[0121] For asymmetrical sharp edges (in other embodiments), based on the first preferred chamfer angle respectively With the second preferred oblique angle Perform oblique scanning to obtain the corresponding oblique measurement point cloud. and .

[0122] Parallel scan:

[0123] Along y s0 The axially moving line laser sensor 1 passes through the optical plane. Parallel scanning is performed on the inlet edge of the compressor blade cascade under test to obtain point cloud data. Due to point clouds The point cloud with the fewest anomalies is therefore the point cloud... As a reference point cloud, it is used to assist in oblique measurement point cloud in subsequent steps. and The processing.

[0124] It should be noted that the measurement point clouds acquired under the different scanning methods described above... , , All points are located in the line laser sensor base coordinate system. The hand-eye transformation matrix required to convert the measurement points in the optical plane of line laser sensor 1 into a measurement point cloud in the line laser sensor base coordinate system can be solved by the hand-eye calibration method, which is a conventional method in this field.

[0125] Step 5: Based on the benchmark point cloud Using the nearest neighbor concept, we eliminate obliquely tangent measurement point clouds. and From the anomaly point cloud, the oblique measurement point cloud is obtained. and .

[0126] In online laser beveling measurement, as the beveling angle... The improvement in reflectivity exacerbates the high reflectivity phenomenon and the problem of abnormal point clouds becomes serious.

[0127] Baseline point cloud Expressed as , obliquely cut the measurement point cloud and Unified expression ;in, Represents the benchmark point cloud The k-th measuring point, Represents the benchmark point cloud The number of measurement points in the data; i takes values ​​of 1 and 2 respectively. Represents oblique measurement point cloud The k-th measuring point, Represents oblique measurement point cloud The number of measuring points in the system.

[0128] Based on the nearest neighbor concept, calculate the oblique measurement point cloud respectively. ( and Clouds from each measuring point within the area to the benchmark point The distance between each measuring point within the area is used to obtain the shortest distance. and the nearest point vector ; where the nearest point vector The direction is defined as the measuring point. Pointing to the measuring point The nearest point vector The defined plane (the optical plane of line laser sensor 1 in the base coordinate system of the line laser sensor) is defined as... Two-dimensional vectors within a parallel plane; obliquely tangent measurement point cloud The shortest distances between all measuring points constitute the oblique measurement point cloud. nearest distance set The nearest point vectors of all measuring points constitute the oblique measurement point cloud. The nearest point vector set , , i takes values ​​of 1 and 2 respectively.

[0129] Based on the set distance threshold , (The values ​​range from 0.040 to 1.000 mm. In this embodiment,) , The values ​​are taken as 0.062 mm and 0.063 mm respectively, along with the corresponding vector constraints, for the obliquely measured point cloud. ( and Outlier detection is performed at each measuring point within the range:

[0130] When obliquely measuring point cloud The closest distance to a certain measuring point When, and the nearest point vector In the prescribed direction (z of line laser sensor 1) s0 The angle between the axis and the direction in the base coordinate system of the online laser sensor. When the angle is less than 90°, the corresponding measuring point These are abnormal point clouds and are removed; among them, .

[0131] Conversely, the corresponding measuring points Point clouds that are not anomalous are preserved.

[0132] After removing outlier point clouds, oblique measurement point clouds are obtained. and .

[0133] Step 6: Based on the benchmark point cloud Oblique measurement point cloud after removing outlier point clouds and After fine registration, duplicate points in the overlapping areas are removed to obtain the point cloud of the compressor blade inlet edge surface. .

[0134] Oblique measurement point cloud and Essentially a multi-view measurement point cloud, fine registration processing is required to further enhance the stitching accuracy of the multi-view measurement point cloud. This is because the reference point cloud... The point cloud is obtained from a single scan, and there is no stitching error as in multi-view measurement point clouds. Therefore, this invention is based on a reference point cloud. Achieving oblique measurement point cloud and Precise registration:

[0135] Building point clouds , , The axis is aligned with the bounding box, i.e. , and .

[0136] To each and , and Perform intersection region detection to obtain point clouds of overlapping regions: and In the diagram, the overlapping point clouds are denoted as follows: , , Represents the benchmark point cloud The point cloud located in the intersecting region, Represents oblique measurement point cloud Point clouds located in the intersecting region; and In the diagram, the overlapping point clouds are denoted as follows: , , Represents the benchmark point cloud The point cloud located in the intersecting region, Represents oblique measurement point cloud Point clouds located in the intersecting region.

[0137] Because the number of points in the scanned point cloud is large, it affects the registration efficiency. Therefore, for the overlapping area point cloud... , , and Voxel downsampling was performed separately, and the overlapping point clouds after downsampling were denoted as follows: , , and In this embodiment, the number of points in the overlapping region point cloud after downsampling is 0.2 times the number of points in the overlapping region point cloud before downsampling.

[0138] The point clouds of the overlapping regions after downsampling were registered using the coherent point drift algorithm: the point clouds were then... Register to point cloud Obtain the transformation matrix ; to point clouds Register to point cloud Obtain the transformation matrix .

[0139] The obtained transformation matrix , Acting on oblique measurement point clouds respectively and Obtain the point cloud after fine registration and .

[0140] Point cloud based on axis-aligned bounding box and Intersection region detection is performed to obtain the overlapping area between the two components. Removing one of them completes the deduplication process, and the point cloud of the compressor blade inlet edge is obtained after deduplication. dot clouds This is the final three-dimensional reconstruction result of the compressor blade inlet edge obtained in this embodiment, such as... Figure 8 As shown.

[0141] Technical effectiveness verification:

[0142] The inlet edge of a compressor blade cascade of a certain model was used as the inspection object, with a radius of curvature of 0.14 mm. Three-dimensional point clouds of the inlet edge of this compressor blade cascade were obtained using a high-precision automatic zoom three-dimensional topography measuring instrument, a line laser parallel scanning method, and the method of this invention, respectively. The high-precision automatic zoom three-dimensional topography measuring instrument took approximately 5 hours, the line laser parallel scanning method took approximately 1 minute, and the method of this invention took approximately 3.5 minutes.

[0143] Using measurements from a high-precision automatic zoom 3D topography measuring instrument as true values, the line laser parallel scanning method and the method of this invention were evaluated. The results show that, within the same measurement range, compared to the line laser parallel scanning method, the method of this invention increases the number of measurement points by more than 60% and reduces the measurement deviation by more than 25%, achieving low-cost, high-precision, and high-efficiency measurement and 3D reconstruction of the compressor blade inlet side surface.

[0144] The above explanation only uses the compressor blade inlet edge as an example. The three-dimensional reconstruction method for the sharp edge surfaces of other components is the same in principle and steps as the above technical solution, and will not be repeated here.

Claims

1. A method for three-dimensional reconstruction of sharp-edge surfaces based on line laser oblique scanning, characterized in that, Including the following steps: The surface to be measured is projected onto a two-dimensional plane to obtain the projection area. A quantitative model for the number of measurement points in the line laser oblique scanning is established to describe the relationship between the number of measurement points in the projection area and the oblique angle. A quantitative model for the reflected light intensity of the surface to be measured is established to describe the relationship between the reflected light intensity of the surface to be measured and the oblique angle. The two-dimensional plane is a plane perpendicular to the beam axis of the line laser sensor in the parallel scanning pose. The intersection of the light plane of the line laser sensor and the surface to be measured in the parallel scanning pose is defined as the cross section. In the oblique scanning pose, the intersection of the light plane of the line laser sensor and the surface to be measured is defined as the oblique section. The oblique angle refers to the angle between the oblique section and the cross section. Based on the quantitative model of the number of measurement points in the line laser oblique scanning and the quantitative model of the reflected light intensity of the sharp edge surface, an optimization model for the oblique angle is constructed and solved to obtain the optimal oblique angle. Based on the optimized oblique cutting angle and employing a bidirectional scanning strategy, continuous line laser oblique cutting scanning is performed on the surface to be measured to obtain the oblique cutting measurement point cloud. and ; A continuous line laser parallel scan is performed on the surface to be measured to obtain a reference point cloud. Oblique measurement point cloud and and benchmark cloud All are located in the line laser sensor base coordinate system; the line laser sensor base coordinate system refers to the coordinate system of the mechanical device used to control the movement of the line laser sensor; Based on benchmark point cloud Using the nearest neighbor concept, we eliminate obliquely tangent measurement point clouds. and From the anomaly point cloud, the oblique measurement point cloud is obtained. and ; Based on benchmark point cloud Oblique measurement point cloud and After precise registration, duplicate point clouds in overlapping areas are removed to obtain the point cloud of the sharp edge surface to be measured. This enables three-dimensional reconstruction of the sharp edge surface to be measured.

2. The method for three-dimensional reconstruction of sharp-edge surfaces based on line laser oblique scanning according to claim 1, characterized in that, The method for constructing the intensity quantization model of reflected light from sharp-edged surfaces is as follows: Images of line laser stripes on the surface of the sharp edge under different bevel angles are acquired, and the global normalized grayscale sum of all pixels in each line laser stripe image is calculated. A double exponential function is used to fit the global normalized gray sum of all pixels in the image of the chamfered angle and the corresponding line laser stripe to obtain a quantitative model of the reflected light intensity of the sharp edge surface.

3. The method for three-dimensional reconstruction of sharp-edge surfaces based on line laser oblique scanning according to claim 2, characterized in that, When using a double exponential function to fit and establish a quantitative model of reflected light intensity on a sharp edge surface, it is also necessary to center and scale the oblique angle.

4. The method for three-dimensional reconstruction of sharp-edge surfaces based on line laser oblique scanning according to claim 1, 2, or 3, characterized in that, The method for constructing the bevel angle optimization model is as follows: The logarithmic form of the quantization model for the number of measurement points in a line laser oblique scanning is normalized to obtain the function. The quantization model of reflected light intensity from the sharp-edge surface is normalized to obtain the function. ; For functions and Weighted summation yields the objective function used to solve for the optimal oblique angle; It is a beveled angle; With minimizing the objective function as the solution objective and the range of values ​​for the chamfer angle as the constraint, a chamfer angle optimization model is constructed.

5. The method for three-dimensional reconstruction of sharp-edge surfaces based on line laser oblique scanning according to claim 4, characterized in that: For a symmetrical pointed edge, the range of the chamfer angle is: The quantization model for the number of measurement points in the line laser oblique scanning, the quantization model for the reflected light intensity of the sharp edge surface, and the optimization model for the oblique angle are all in... Constructed within the range of the oblique angle; For asymmetric sharp edges, the range of the chamfer angle is: The quantization model for the number of measurement points in the line laser oblique scanning, the quantization model for the reflected light intensity of the sharp edge surface, and the optimization model for the oblique angle are all respectively in... and Constructed within the range of the oblique angle.

6. The method for three-dimensional reconstruction of sharp-edge surfaces based on line laser oblique scanning according to claim 5, characterized in that, The bidirectional scanning strategy refers to: For symmetrical sharp edges, the preferred chamfer angle is negatively determined, and chamfer scanning is performed based on the preferred chamfer angle and the chamfer angle after taking the negative value, respectively. For asymmetric sharp edges, based on respectively The first preferred chamfer angle of the range and The second preferred bevel angle within the range is used to perform a bevel scan.

7. The method for three-dimensional reconstruction of sharp-edge surfaces based on line laser oblique scanning according to claim 6, characterized in that, Based on benchmark point cloud Using the nearest neighbor concept, we eliminate obliquely tangent measurement point clouds. and The method for handling anomalous point clouds is: Calculate the distances from each measuring point in the oblique measurement point cloud to each measuring point in the reference point cloud, and obtain the shortest distance and the shortest point vector; Abnormal point cloud identification and handling: When the nearest distance to a certain measuring point in the oblique measurement point cloud is greater than the set distance threshold, and the angle between the nearest point vector and the specified direction is less than 90°, the measuring point is identified as an abnormal point cloud and is removed. Conversely, if the measurement point is a non-abnormal point cloud, it should be retained. The specified direction refers to the direction of the beam axis of the line laser sensor in the base coordinate system of the line laser sensor under parallel scanning pose.

8. The method for three-dimensional reconstruction of sharp-edge surfaces based on line laser oblique scanning according to claim 7, characterized in that, Based on benchmark point cloud Oblique measurement point cloud and The method for fine registration is: Construct point clouds separately , , The axis is aligned with the bounding box; Point cloud and axis-aligned bounding box, point cloud and Intersection region detection is performed using axis-aligned bounding boxes to obtain overlapping region point clouds. , , and ,in, Represents the benchmark point cloud The point cloud located in the intersecting region, Represents oblique measurement point cloud The point cloud located in the intersecting region, Represents the benchmark point cloud The point cloud located in the intersecting region, Represents oblique measurement point cloud Point clouds located in the intersecting region; Point clouds of overlapping regions , , and Voxel downsampling was performed separately to obtain point clouds. , , , ; Using the coherent point drift algorithm, the point cloud Register to point cloud Obtain the transformation matrix , will dot clouds Register to point cloud Obtain the transformation matrix ; Transformation matrix , Acting on oblique measurement point clouds respectively and Obtain the point cloud after fine registration and .

9. The method for three-dimensional reconstruction of sharp-edge surfaces based on line laser oblique scanning according to claim 8, characterized in that, After fine registration, the method for deduplicating point clouds in overlapping regions is as follows: Point cloud based on axis-aligned bounding box and Perform intersection region detection to obtain the overlapping area between the two. Remove point clouds or The point cloud located within the overlapping region.

Images