Accurate fusion method of line-structured light ring stitching based on affine space discrete transformation

Abstract

The present invention discloses a precise fusion method for annular splicing of line structured light based on affine space discrete transformation, which is applied to a line structured light sensor in an annular layout to collect point cloud data; the method includes: collecting the cross-section point cloud of the calibration block by the line structured light sensor; obtaining the theoretical point cloud corresponding to each frame of the cross-section point cloud using a three-dimensional coordinate measuring machine; converting both the three-dimensional theoretical point cloud and the scanned point cloud into two-dimensional point clouds; fusing and registering the scanned point cloud and the theoretical point cloud through the intersection of the contour lines and the angle bisector of the inferior angle for each frame of the point cloud; selecting any sensor as the main line structured light sensor, and then transforming it to the coordinate system of the first cross-section of the main line structured light sensor through a three-dimensional registration matrix; and performing linear interpolation on the fusion matrices before and after the cross-section when the cross-section position is not in the cross-section with the three-dimensional registration matrix during measurement. The present invention solves the affine error problem that cannot be eliminated by traditional six-degree-of-freedom registration, and has high point cloud fusion accuracy and accurate size and shape of the measured workpiece.

Description

Technical Field

[0001] The present invention relates to the technical field of point cloud fusion in different coordinate systems, and in particular to a precise fusion method for line-structured light ring splicing based on affine space discrete transformation. Background Art

[0002] As the level of automation in the manufacturing industry continues to increase, the requirements for product quality inspection speed and accuracy are also gradually increasing. Product quality inspection equipment is also transitioning from traditional contact measurement equipment such as contact measurement fixtures and three-dimensional coordinate measuring machines to non-contact measurement equipment such as linear structured light sensors. As a result, the measurement accuracy of linear structured light sensor measurement systems is becoming increasingly important to more and more companies. One of the key factors affecting measurement accuracy is the calibration of the relative position of each linear structured light sensor coordinate system.

[0003] Non-contact measurement is a technology that obtains information about an object's properties or geometry without requiring direct contact between the object and the measuring device. Linear structured light sensors calculate distance or depth by measuring the time it takes for a laser beam to travel from the device to reflect off the object. This technology enables fast, accurate, and non-destructive measurement and inspection, and has important applications in many fields, including industry, healthcare, and scientific research, helping to improve production efficiency, quality control, and the feasibility of scientific research.

[0004] The point cloud data acquired by the linear structured light sensor is located in different coordinate systems. Before fusion, these point cloud data need to be converted into a common global coordinate system so that they can be properly aligned. This involves point cloud registration technologies such as the ICP algorithm. Due to installation errors and other reasons, the normal direction of the light plane of the linear structured light sensor is generally not parallel to the direction of motion of the linear slide. This causes affine deformation of the point cloud in three-dimensional space and affects the accuracy of point cloud fusion. Traditional six-degree-of-freedom fusion methods ignore the distortion problem and directly perform point cloud fusion, which inevitably leads to reduced point cloud fusion accuracy and, in turn, reduced detection equipment accuracy. This brings many limitations to the linear structured light measurement system in application scenarios requiring high-precision detection.

[0005] Therefore, it is necessary to improve the existing technology to solve the shortcomings of the existing technology. Summary of the Invention

[0006] 1. Technical problems to be solved:

[0007] The purpose of the present invention is to address the problems of low point cloud fusion accuracy, high calibration cost, complex structure and long time consumption under the existing structured light scanning system. A low-cost fusion method using affine distortion in multiple two-dimensional plane discrete spaces is invented to quickly and accurately calculate the relative position between the coordinate systems of each sensor, and then fuse the point clouds under the coordinate system to effectively solve problems such as stratification and misalignment of the scanning point cloud.

[0008] 2. Technical solution:

[0009] A precise fusion method for annular splicing of line-structured light based on affine spatial discrete transformation is characterized by: a line-structured light sensor system in an annular layout realizes distortion-free acquisition of spliced and fused point cloud data; the line-structured light sensor in the annular layout includes multiple line-structured light sensors, each of which can scan at least two intersecting straight lines on the cross-sectional profile of a calibration block within its field of view; during calibration, the calibration block moves at a constant speed in a direction approximately perpendicular to the light plane, so that the position to be calibrated passes through the light plane;

[0010] The calibration and fusion process of the calibration block specifically includes the following steps:

[0011] Step 1: Each line structured light sensor simultaneously collects the cross-sectional contour of the calibration block at a preset distance and generates a corresponding cross-sectional point cloud of the calibration block. Each collected point cloud is called a frame of cross-sectional point cloud until the entire travel path to be calibrated is collected.

[0012] Step 2: Use a three-dimensional coordinate measuring machine to obtain the three-dimensional coordinate measuring machine measurement point cloud corresponding to each frame of the cross-section point cloud as the theoretical point cloud of the cross-section point cloud of the frame. The theoretical point cloud is a three-dimensional point cloud located in the world coordinate system;

[0013] Step 3: Since the theoretical point cloud of each frame and the scanned point cloud used for cross-section registration are in the same cross-section, i.e., the Y-axis coordinates of the cross-section are the same, the coordinates corresponding to the Y-axis are discarded to convert the three-dimensional theoretical point cloud and the scanned point cloud of each frame into corresponding two-dimensional point clouds; the two-dimensional point clouds include X-axis coordinates and Z-axis coordinates;

[0014] Step 4: Perform fusion registration of the scanned point cloud and the theoretical point cloud for each frame of point cloud; including: obtaining the intersection of the two intersecting contour lines and the angle bisector of the inferior angle of the corresponding two-dimensional point cloud of the section, and at the same time obtaining the corresponding straight line intersection and the angle bisector of the inferior angle of the theoretical point cloud; through two-dimensional posture transformation, first rotate the angle Δα between the two angle bisectors so that the inferior angle bisector of the scanned point cloud and the inferior angle bisector of the theoretical point cloud are parallel to each other, and then translate the point cloud to make the intersection of the scanned point cloud coincide with the intersection of the theoretical point cloud, record the two-dimensional registration matrix of the process, and thus complete the registration of a section;

[0015] Step 5: Repeat steps 3 and 4 until all scanned point clouds from the line structured light sensors are fused and registered to the coordinate system of the theoretical point cloud.

[0016] Step 6: Select any sensor as the mainline structured light sensor, and then transform the fused point cloud into the coordinate system of the first section of the mainline structured light sensor using the 3D registration matrix to ensure that the subsequent measurement point cloud is distortion-free.

[0017] Step 7: After obtaining the fusion matrix of the point cloud at each section position, if the section position is not in the section with a three-dimensional registration matrix during actual measurement, linear interpolation is performed on the fusion matrices before and after the section to generate a point cloud fusion matrix suitable for the section.

[0018] Furthermore, the theoretical point cloud in step 4 is registered with the scanned point cloud. The specific process is as follows:

[0019] The angle between the two angle bisectors is expressed as Δα, and the intersection points of the scanned point cloud and the theoretical point cloud are The new intersection point of the scanned point cloud lines when the two angle bisectors are parallel after rotation but::

[0020]

[0021] Let the two-dimensional registration matrix be M 2dTrans ,but:

[0022]

[0023] Scan point cloud origin} Multiply M on the left 2dTrans The scanned point cloud can be registered to the coordinate system of the theoretical point cloud, that is, the registration of the scanned point cloud and the theoretical point cloud. The point cloud after registration is {p scan},but:

[0024] {p scan}=M 2dTrans ×{p origin} (3).

[0025] Furthermore, in step six, the fused point cloud is transformed into the coordinate system of the first section of the main line structured light sensor through the three-dimensional registration matrix and a discrete transformation precise fusion method is used, specifically:

[0026] S51: Convert the two-dimensional registration matrix into a three-dimensional registration matrix, that is, the three-dimensional registration matrix M of the point cloud in each sensor coordinate system needs to be converted into a three-dimensional registration matrix M. 2dTrans Add a dimension based on , namely:

[0027]

[0028] S52: Select any line structured light sensor coordinate system as the target coordinate system of the fused point cloud. This sensor is recorded as the main line structured light sensor, and its three-dimensional registration matrix is recorded as M C1 , the 3D registration matrix of the remaining sensors is recorded as M Ci ,i∈2,3…n c , n c is the maximum number of sensors; based on formula (4), we can get M C1 and M Ci ;

[0029] Fuse all sensors of the jth section into the main sensor coordinate system of the current section, j∈1,2…n sec , where the registration matrix of the single-section line structured light sensor point cloud transformed into the main line structured light sensor coordinate system is n sec is the maximum number of cross sections; All can be obtained based on formula (4);

[0030] At this point, all line structured light sensor point clouds are fused into the main sensor coordinate system of the current section. Since the affine distortion between the sections of the main sensor still exists, the point clouds of each section cannot be directly spliced and fused. The point clouds of each line structured light sensor in each section before and after fusion are recorded as The registration process of fusing any Y-coordinate position section into the first section coordinate system of the main line structured light sensor is:

[0031]

[0032] Furthermore, in step 7, linear interpolation is performed on the fusion matrix before and after the section. Specifically, the linear interpolation between j and j+1 is taken as an example. The specific process is:

[0033] First, obtain the ratio k of the distance between the section to be interpolated and the j section and the distance between the j section and the j+1 section, and determine the increase or decrease ratio of the rotation and translation of the registration matrix of the section to be interpolated, that is:

[0034]

[0035] in and They are the 2D rotation and translation corresponding to each sensor point cloud of the interpolation section, and the 2D rotation between section j and section j+1 and translation Obtained when the scanned point cloud is registered with the theoretical point cloud, thereby obtaining the registration matrix of the section to be interpolated

[0036]

[0037] The point cloud of each cross-section and line structured light sensor before and after fusion is recorded as Then the registration process of fusing any Y coordinate position section into the first section coordinate system of the main line structured light sensor is expressed as:

[0038]

[0039] Furthermore, the annular layout line structured light sensor system includes a gantry base, an annular layout line structured light sensor, and a linear slide; the annular layout line structured light sensor is fixed on the side panel of the gantry base; the linear slide is fixedly installed under the gantry base; during calibration, the calibration block is fixed to the surface of the linear slide to realize the scanning area formed by the annular layout line structured light sensor; the calibration block is a polygonal calibration block; the length of the calibration block and the stroke of the linear slide are the same as the actual length of the workpiece to be measured.

[0040] 3.Beneficial effects:

[0041] (1) In the present invention, a 3D cross-section point cloud is first converted into a 2D point cloud, and then multiple 2D cross-sections are registered, thereby avoiding the affine errors that are difficult to avoid when directly fusing 3D point clouds. This fusion method eliminates the slight errors, noise, and distortion that may exist between each sensor, ensuring that the errors in the overlapping areas of the fused point clouds are minimized. This also reduces the difficulty of device installation while improving the measurement accuracy of the device. The overall calibration cost is only about one-tenth of that of 6-DOF fusion.

[0042] (2) The calibration block structure of the present invention is simple. As long as the field of view of all linear structured light sensors contains at least two intersecting straight lines of the calibration block cross section, the processing cost is low. It is compatible with the calibration of most measurement systems composed of linear slides and linear structured light sensors on the market and has a wide range of applications. It is particularly suitable for calibration during the large motion stroke scanning detection process of slender workpieces.

[0043] (3) The theoretical point cloud for point cloud registration in this invention is obtained by measuring multiple cross sections of a calibration block with a three-dimensional coordinate measuring machine and then reverse modeling. Its registration accuracy is higher than that of a discrete point cloud of a CAD model. Furthermore, the registration data only uses one point (the intersection of two straight lines of the contour) and one line (the angle bisector). The registration speed is faster than the general ICP method. Experimental data shows that compared with six-degree-of-freedom point cloud fusion, it reduces the operator's difficulty in use, improves the efficiency of point cloud fusion by 50%, and improves the accuracy by more than 70%. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 Schematic diagram of the circular layout of the line structured light sensor system and its fusion calibration in the present invention;

[0045] Figure 2Schematic diagram of the shape of a calibration block to which the present invention can be applied for calibration;

[0046] Figure 3 Schematic diagram of the single-section point cloud registration principle in the present invention;

[0047] Figure 4 Schematic diagram of the coordinate transformation principle of affine space discrete transformation in the present invention;

[0048] Figure 5 Specific flow chart of the precise fusion method of affine space discrete transformation in the present invention.

[0049] Reference numerals: gantry base 1 , linear structured light sensor 2 , calibration fixture 3 , linear slide 4 , calibration block 5 , sensor light plane 6 . DETAILED DESCRIPTION

[0050] The present invention will be described in detail below with reference to the accompanying drawings.

[0051] A precise fusion method for annular splicing of line-structured light based on affine space discrete transformation is characterized by: a line-structured light sensor system applied to an annular layout realizes distortion-free acquisition of spliced and fused point cloud data; the line-structured light sensor in the annular layout includes multiple line-structured sensors, and each line-structured sensor can scan at least two intersecting straight lines on the cross-sectional contour of a calibration block within its field of view; during calibration, the calibration block moves at a uniform speed in a direction approximately perpendicular to the light plane, so that the position to be calibrated passes through the light plane.

[0052] This solution is suitable for line structured light sensors with a ring layout, such as the attached Figure 1 It is a structured light sensor with three wire rings, as shown in the attached figure. Figure 3 There are five line structured light sensors arranged in a ring. The workpiece to be collected is located in the middle of the ring and moves. The shape of the workpiece is as shown in the attached figure. Figure 2 As shown, attached Figure 2 The workpieces with pentagonal cross-sections, octagonal cross-sections and quadrilateral cross-sections are provided; it can be seen that the cross-sections collected by the pentagon and octagon include two intersecting contour edges, so this method can be used for sampling. However, if the field of view collected by the sensor only includes one contour edge as shown in the figure, this method is not applicable to the quadrilateral. Since the calibration block is actually driven by a linear slide, it is difficult to ensure that the movement direction of the linear slide is parallel to the normal of the light plane. The generation direction of the point cloud section is not strictly increased along the Y axis according to the distance of the scanning interval, but also includes a small translation in the X and Z directions, so the point cloud distortion is generated. This precise fusion method can solve the problem of the angle between the normal of the light plane and the direction of movement.

[0053] As attached Figure 5As shown in the flowchart, the calibration and fusion process of the calibration block specifically includes the following steps:

[0054] Step 1: Each line structured light sensor simultaneously collects the cross-sectional contour of the calibration block at a preset distance and generates the corresponding cross-sectional point cloud of the calibration block. Each collected point cloud is called a frame of cross-sectional point cloud until the entire travel path to be calibrated is collected.

[0055] In practice, the cross-sectional point cloud is usually collected at a preset interval of about 1 mm.

[0056] Step 2: Use a three-dimensional coordinate measuring machine to obtain the three-dimensional coordinate measuring machine measurement point cloud corresponding to each frame of the cross-section point cloud as the theoretical point cloud of the cross-section point cloud of the frame. The theoretical point cloud is a three-dimensional point cloud located in the world coordinate system.

[0057] In subsequent steps, the point cloud scanned by each line structured light sensor is used as the theoretical point cloud and is registered with the point cloud obtained by the coordinate measuring machine. The theoretical point cloud is obtained through reverse modeling of the coordinate measuring machine, and its registration accuracy is higher than that of the discrete point cloud of the CAD model.

[0058] Step 3: Since the theoretical point cloud of each frame and the scanned point cloud used for section registration are in the same section, that is, the Y-axis coordinates of the section are the same, the coordinates corresponding to the Y-axis are discarded, and the three-dimensional theoretical point cloud and the scanned point cloud of each frame are converted into corresponding two-dimensional point clouds; the two-dimensional point cloud includes X-axis coordinates and Z-axis coordinates.

[0059] Step 4: Perform fusion registration of the scanned point cloud and the theoretical point cloud for each frame of point cloud; including: obtaining the intersection of the two intersecting contour lines and the angle bisector of the inferior angle of the corresponding two-dimensional point cloud of the section, and at the same time obtaining the corresponding straight line intersection and the angle bisector of the inferior angle of the theoretical point cloud; through two-dimensional posture transformation, first rotate the angle Δα between the two angle bisectors so that the inferior angle bisector of the scanned point cloud and the inferior angle bisector of the theoretical point cloud are parallel to each other, and then make the intersection of the scanned point cloud coincide with the intersection of the theoretical point cloud by translating the point cloud, record the two-dimensional registration matrix of the process, and thus complete the registration of a section.

[0060] As attached Figure 2 As shown, the data in this registration process only uses one point (intersection point) and one line (angle bisector), and its registration speed is faster than the general ICP method.

[0061] Step 5: Repeat steps 3 and 4 until all scanned point clouds from the line structured light sensors are fused and registered to the coordinate system of the theoretical point cloud.

[0062] Step 6: Select any sensor as the mainline structured light sensor, and then transform the fused point cloud into the coordinate system of the first section of the mainline structured light sensor through the three-dimensional registration matrix to achieve distortion-free point cloud measurement.

[0063] Step 7: After obtaining the fusion matrix of the point cloud at each section position, if the section position is not in the section with a three-dimensional registration matrix during actual measurement, linear interpolation is performed on the fusion matrices before and after the section to generate a point cloud fusion matrix suitable for the section.

[0064] Furthermore, as attached Figure 3 As shown in Figure 2, the theoretical point cloud in step 4 is registered with the scanned point cloud. The specific process is as follows:

[0065] The angle between the two angle bisectors is expressed as Δα, and the intersection points of the scanned point cloud and the theoretical point cloud are The new intersection point of the scanned point cloud lines when the two angle bisectors are parallel after rotation but::

[0066]

[0067] Let the two-dimensional registration matrix be M 2dTrans ,but:

[0068]

[0069] Scan point cloud origin} Multiply M on the left 2dTrans The scanned point cloud can be registered to the coordinate system of the theoretical point cloud, that is, the registration of the scanned point cloud and the theoretical point cloud. The point cloud after registration is {p scan},but:

[0070] {p scan}=M 2dTrans ×{p origin} (3).

[0071] Attachment Figure 3 In the figure, the point cloud captured by the sensor at the lower right corner of the pentagon is used as a specific example. In the figure, 10 represents the two-dimensional scanned point cloud obtained by scanning with a line structured light sensor, 11 is the corresponding two-dimensional theoretical point cloud measured using a three-dimensional coordinate measuring machine for the section containing the scanned point cloud; 12 represents the angle bisector of the two straight lines fitted to the scanned point cloud; the theoretical point cloud is obtained by discrete measurements of the polygonal section of the calibration block, and 13 represents the angle bisector of the polygonal interior angle fitted by the theoretical point cloud. The figure illustrates the registration process of the scanned point cloud and the theoretical point cloud for a cross section: first, the point cloud is translated so that the intersection of the straight lines of the scanned point cloud 10 coincides with the vertex of the theoretical point cloud 11, and then the scanned point cloud 10 is rotated around the vertex by an angle equal to the angle Δα between the two angle bisectors.

[0072] Furthermore, in step six, the fused point cloud is transformed into the coordinate system of the first section of the main line structured light sensor through the three-dimensional registration matrix and a discrete transformation precise fusion method is used, specifically:

[0073] S51: Convert the two-dimensional registration matrix into a three-dimensional registration matrix, that is, the three-dimensional registration matrix M needs to be converted into a three-dimensional registration matrix M. 2dTrans Add a dimension based on:

[0074]

[0075] S52: Select any line structured light sensor coordinate system as the target coordinate system of the fused point cloud. This sensor is recorded as the main line structured light sensor, and its three-dimensional registration matrix is recorded as M C1 , the 3D registration matrix of the remaining sensors is recorded as M Ci ,i∈2,3…n c , n c is the maximum number of sensors; based on formula (4), we can get M C1 and M Ci ;

[0076] Fuse all sensors of the jth section into the main sensor coordinate system of the current section, j∈1,2…n sec , where the registration matrix of the single-section line structured light sensor point cloud transformed into the main line structured light sensor coordinate system is n sec is the maximum number of cross sections; All can be obtained based on formula (4);

[0077] At this point, all line structured light sensor point clouds are fused into the main sensor coordinate system of the current section. Since the affine distortion between the sections of the main sensor still exists, the point clouds of each section cannot be directly spliced and fused. The point clouds of each line structured light sensor in each section before and after fusion are recorded as The registration process of fusing any Y-coordinate position section into the first section coordinate system of the main line structured light sensor is:

[0078]

[0079] The specific process of step six is as follows Figure 4 As shown, attached Figure 4 The schematic diagram of the collected full cross section is arranged in order of the size of the Y coordinate of the scanned point cloud cross section from front to back, and the specific processing process is shown in S52.

[0080] Furthermore, in step 7, linear interpolation is performed on the fusion matrix before and after the section. Specifically, the linear interpolation between j and j+1 is taken as an example. The specific process is:

[0081] First, obtain the ratio k of the distance between the section to be interpolated and the j section and the distance between the j section and the j+1 section, and determine the increase or decrease ratio of the rotation and translation of the registration matrix of the section to be interpolated, that is:

[0082]

[0083] in and They are the 2D rotation and translation corresponding to each sensor point cloud of the interpolation section, and the 2D rotation between section j and section j+1 and translation Obtained when the scanned point cloud is registered with the theoretical point cloud, thereby obtaining the registration matrix of the section to be interpolated

[0084]

[0085] The point cloud of each cross-section and line structured light sensor before and after fusion is recorded as The registration process of fusing the full cross section into the first cross section coordinate system of the main line structured light sensor is expressed as:

[0086]

[0087] Furthermore, as attached Figure 1 As shown, the annular linear structured light sensor system includes a gantry base 1, an annular linear structured light sensor 2, and a linear slide 4. The annular linear structured light sensor is fixed to the side panel of the gantry base; the linear slide is fixedly mounted below the gantry base. During calibration, a calibration block is fixed to the surface of the linear slide to achieve the scanning area formed by the annular linear structured light sensor. The calibration block is a polygonal calibration block. The length of the calibration block and the stroke of the linear slide are the same as the actual length of the workpiece to be measured. Reference 3 in the figure represents the calibration fixture used to fix the calibration block.

[0088] Although the present invention has been disclosed above in terms of preferred embodiments, they are not intended to limit the present invention. Anyone skilled in the art can make various changes or modifications without departing from the spirit and scope of the present invention. Therefore, the scope of protection of the present invention should be based on the scope of protection defined by the claims of this application.

Claims

1. A precise fusion method for line-structured light ring splicing based on affine space discrete transformation, characterized by: A line structured light sensor system used in a ring layout achieves distortion-free acquisition of spliced and fused point cloud data. The ring-shaped line structured light sensor includes multiple line structure sensors, each of which can scan at least two intersecting lines on the cross-sectional profile of a calibration block within its field of view. During calibration, the calibration block moves at a constant speed in a direction approximately perpendicular to the light plane, so that the position to be calibrated passes through the light plane. The calibration and fusion process of the calibration block specifically includes the following steps: Step 1: Each line structured light sensor simultaneously collects the cross-sectional contour of the calibration block at a preset distance and generates a corresponding cross-sectional point cloud of the calibration block. Each collected point cloud is called a frame of cross-sectional point cloud until the entire travel path to be calibrated is collected. Step 2: Use a three-dimensional coordinate measuring machine to obtain the three-dimensional coordinate measuring machine measurement point cloud corresponding to each frame of the cross-section point cloud as the theoretical point cloud of the cross-section point cloud of the frame. The theoretical point cloud is a three-dimensional point cloud located in the world coordinate system; Step 3: Since the theoretical point cloud of each frame and the scanned point cloud used for cross-section registration are in the same cross-section, i.e., the Y-axis coordinates of the cross-section are the same, the coordinates corresponding to the Y-axis are discarded to convert the three-dimensional theoretical point cloud and the scanned point cloud of each frame into corresponding two-dimensional point clouds; the two-dimensional point clouds include X-axis coordinates and Z-axis coordinates; Step 4: Perform fusion registration of the scanned point cloud and the theoretical point cloud for each frame of point cloud; including: obtaining the intersection of the two intersecting contour lines and the angle bisector of the inferior angle of the corresponding two-dimensional point cloud of the section, and at the same time obtaining the corresponding straight line intersection and the angle bisector of the inferior angle of the theoretical point cloud; through two-dimensional posture transformation, first rotate the angle △α between the two angle bisectors so that the inferior angle bisector of the scanned point cloud and the inferior angle bisector of the theoretical point cloud are parallel to each other, and then translate the point cloud to make the intersection of the scanned point cloud coincide with the intersection of the theoretical point cloud, record the two-dimensional registration matrix of the process, and thus complete the registration of a section; Step 5: Repeat steps 3 and 4 until all scanned point clouds from the line structured light sensors are fused and registered to the coordinate system of the theoretical point cloud. Step 6: Select any sensor as the mainline structured light sensor, and then transform the fused point cloud into the coordinate system of the first section of the mainline structured light sensor using the 3D registration matrix to ensure that the subsequent measurement point cloud is distortion-free. Step 7: After obtaining the fusion matrix of the point cloud at each section position, if the section position is not in the section with a three-dimensional registration matrix during actual measurement, linear interpolation is performed on the fusion matrices before and after the section to generate a point cloud fusion matrix suitable for the section.

2. The method for precise fusion of line-structured light ring splicing based on affine space discrete transformation according to claim 1 is characterized by: The theoretical point cloud and the scanned point cloud are registered in step 4. The specific process is as follows: The angle between the two angle bisectors is expressed as △α, and the intersection points of the scanned point cloud and the theoretical point cloud are The new intersection point of the scanned point cloud lines when the two angle bisectors are parallel after rotation but: Let the two-dimensional registration matrix be M 2dTrans ,but: Scan point cloud origin } Multiply M on the left 2dTrans The scanned point cloud can be registered to the coordinate system of the theoretical point cloud, that is, the registration of the scanned point cloud and the theoretical point cloud. The point cloud after registration is {p scan },but: {p scan }=M 2dTrans ×{p origin } (3)。 3. The method for precise fusion of line-structured light ring splicing based on affine space discrete transformation according to claim 2 is characterized by: In step 6, the fused point cloud is transformed into the coordinate system of the first section of the main line structured light sensor through the three-dimensional registration matrix and the discrete transformation precise fusion method is used. Specifically: S51: Convert the two-dimensional registration matrix into a three-dimensional registration matrix, that is, the three-dimensional registration matrix M of the point cloud in each sensor coordinate system needs to be converted into a three-dimensional registration matrix M. 2dTrans Add a dimension based on , namely: S52: Select any line structured light sensor coordinate system as the target coordinate system of the fused point cloud. This sensor is recorded as the main line structured light sensor, and its three-dimensional registration matrix is recorded as M C1 , the 3D registration matrix of the remaining sensors is recorded as M Ci ,i∈2,3…n c , n c is the maximum number of sensors; Based on formula (4), we can get M C1 and M Ci ; Fuse all sensors of the jth section into the main sensor coordinate system of the current section, j∈1,2…n sec , where the registration matrix of the single-section line structured light sensor point cloud transformed into the main line structured light sensor coordinate system is n sec is the maximum number of cross sections; All can be obtained based on formula (4); At this point, all line structured light sensor point clouds are fused into the main sensor coordinate system of the current section. Since the affine distortion between the sections of the main sensor still exists, the point clouds of each section cannot be directly spliced and fused. The point clouds of each line structured light sensor in each section before and after fusion are recorded as The registration process of fusing any Y-coordinate position section into the first section coordinate system of the main line structured light sensor is:

4. The method for precise fusion of line-structured light ring splicing based on affine space discrete transformation according to claim 3 is characterized by: In step 7, linear interpolation is performed on the fusion matrix before and after the section. Taking the linear interpolation between j and j+1 as an example, the specific process is as follows: First, obtain the ratio k of the distance between the section to be interpolated and the j section and the distance between the j section and the j+1 section, and determine the increase or decrease ratio of the rotation and translation of the registration matrix of the section to be interpolated, that is: in and They are the 2D rotation and translation corresponding to each sensor point cloud of the interpolation section, and the 2D rotation between section j and section j+1 and translation Obtained when the scanned point cloud is registered with the theoretical point cloud, thereby obtaining the registration matrix of the section to be interpolated The point cloud of each cross-section and line structured light sensor before and after fusion is recorded as Then the registration process of fusing any Y coordinate position section into the first section coordinate system of the main line structured light sensor is expressed as:

5. The method for precise fusion of line-structured light ring splicing based on affine space discrete transformation according to claim 1 is characterized by: The annular layout line structured light sensor system includes a gantry base, an annular layout line structured light sensor, and a linear slide; the annular layout line structured light sensor is fixed on the side panel of the gantry base; the linear slide is fixedly installed under the gantry base; during calibration, the calibration block is fixed to the surface of the linear slide to realize the scanning area formed by the annular layout line structured light sensor; the calibration block is a polygonal calibration block; the length of the calibration block and the stroke of the linear slide are the same as the actual length of the workpiece to be measured.

Images