A mirror-assisted multi-view 3D laser scanning system and a method for panoramic measurement of complex surfaces

Abstract

The present invention discloses a mirror-assisted multi-view 3D laser scanning system and a method for panoramic measurement of complex surfaces. By introducing the assistance of a plane mirror, a laser scanning system (TLS) can simultaneously image an object from three different viewpoints. By accurately calibrating the two plane mirrors in advance, the 3D data reconstructed by the two virtual systems can be converted into the coordinate system of the real system, thereby achieving 360-degree full-surface 3D contour reconstruction of the object. This solves the problem of complex setup and high computational cost of traditional multi-view measurement systems, and can perform 3D measurement of objects within a larger depth of field. This allows us to perform 3D panoramic reconstruction of larger objects with more complex materials and surface textures. The technical solution of the present invention can be applied to a wider range of measurement scenarios.

Description

Technical Field

[0001] The present invention relates to the field of optical measurement, in particular to a mirror-assisted multi-view three-dimensional laser scanning system and a complex surface panoramic measurement method. Background Art

[0002] Optical 3D shape measurement technology is widely used in mechanical manufacturing, industrial quality control, object recognition, virtual reality, cultural relic preservation, medical plastic surgery, and other fields. As a non-contact, high-precision 3D profile measurement technology, laser scanning has proven its ability to reconstruct complex surfaces with complex textures and surface materials. Furthermore, panoramic measurement is crucial for accurately characterizing the complete characteristics of the test sample.

[0003] In recent years, some scholars have achieved 360-degree full-surface three-dimensional shape measurement of objects with complex surfaces by taking multiple measurements from different perspectives. One method is to use three or more synchronized cameras to observe the surface of the object from different angles at the same time, and then use a complex and time-consuming point cloud matching algorithm to merge the data points of each camera into the same coordinate system. Another good solution is to use a mirror-assisted multi-view measurement system to achieve panoramic three-dimensional profile measurement. This method is equivalent to three sets of measuring devices (i.e., one set of actual measuring devices and two virtual measuring devices imaged by plane mirrors) arranged in a "surround" configuration to capture the surface data of the object from three angles at the same time, but in fact only one set of measuring devices and two plane mirrors are required. Epstein et al. introduced plane mirrors into the traditional tassel projection profilometer (FPP) system to achieve high-speed panoramic three-dimensional shape measurement [1]. (Chen and Pan, 2019a) et al. constructed a plane mirror-assisted multi-view digital image correlation (DIC) system to achieve panoramic human skin shape and deformation measurement [2].

[0004] However, due to its limited and obstructed field of view, the traditional laser scanning system is not sufficient to perform three-dimensional reconstruction of the 360-degree full surface of the entire region of interest. At the same time, the method based on spatiotemporal multiplexing requires the use of multiple cameras, which leads to high costs, complex system settings, and high computational costs for matching camera data. The above-mentioned method based on the plane mirror assisted by the FPP system and the DIC system is easily affected by the surface texture and surface material of the target object, and is not effective for measuring surfaces with rich textures or high reflectivity. In addition, the three regions of interest (ROIs), including the real object and the virtual objects in the two plane mirrors, are positioned at different depths. When the object to be measured is large, the three ROIs may not be able to be clearly imaged at the same time due to the depth of field of the lens. The blurring of the ROI may cause errors in the reconstructed surface. Summary of the Invention

[0005] To address the deficiencies of the prior art, the present invention aims to provide a mirror-assisted multi-view 3D laser scanner and a method for panoramic measurement of complex surfaces. By introducing the assistance of a plane mirror, a laser scanning system (TLS) can simultaneously image an object from three different viewpoints. By precisely calibrating the two plane mirrors in advance, the 3D data reconstructed by the two virtual systems can be converted into the coordinate system of the real system, thereby achieving 360-degree 3D reconstruction of the object's full surface. This solves the problem of complex setup and high computational cost of traditional multi-view measurement systems. In addition, since the laser pattern is determined only by the surface contour of the object during laser scanning, it is not sensitive to the texture and material of the object's surface. At the same time, the broadening of the laser line due to defocusing has little effect on the extraction of the laser line centerline. This means that objects can be measured in three dimensions within a larger depth of field, allowing us to perform 3D panoramic reconstruction of larger objects with more complex materials and surface textures. The technical solution of the present invention can be applied to more types of measurement scenarios.

[0006] The system of the present invention comprises a 3D laser scanner and two plane mirrors placed behind the sample. With the aid of these two mirrors, a 360-degree 3D contour reconstruction of the object's entire surface is achieved. Because the laser pattern during laser scanning is determined solely by the object's surface contour, it is insensitive to the object's texture and material. Furthermore, the broadening of the laser line due to defocusing has little effect on the extraction of the laser line's centerline. This allows for 3D panoramic reconstruction of larger objects and is applicable to a variety of measurement scenarios.

[0007] In order to achieve the above objectives, the technical methods adopted by the present invention are specifically as follows:

[0008] The present invention discloses a mirror-assisted multi-view 3D laser scanning system, comprising a camera, a line laser, a linear translation stage, a controller, a computer connected to the camera and the controller for processing data, used to achieve synchronous control of the camera and the linear translation stage, and two plane mirrors for capturing panoramic images. The camera and the line laser are fixed on a sliding bracket, which slides up and down relative to the linear translation stage. There is a height difference between the laser and the camera to generate the parallax required for 3D reconstruction. The two plane mirrors are placed behind the object to be measured and form an angle. The linear translation stage is connected to the controller.

[0009] As a further improvement, the laser plane of the present invention is perpendicular to the two plane mirrors. To ensure continuous laser stripes projected onto the object, the laser plane and the two plane mirrors must be kept as perpendicular as possible. This configuration ensures that the stereoscopic scanning system can simultaneously capture views of the object's front surface and both rear surfaces. In this configuration, all three surfaces of the object can be captured by the laser stereoscopic scanning system in a single shot.

[0010] As a further improvement, the angle between the two plane mirrors described in the present invention is between 90 degrees and 180 degrees.

[0011] As a further improvement, the relative position of the two mirrors of the present invention is at an angle of 120 degrees. Considering the balance between the camera field of view and the camera depth of field, 120 degrees is selected as the optimal angle between the two plane mirrors.

[0012] The present invention also discloses a complex surface panoramic measurement method based on a mirror-assisted multi-view three-dimensional laser scanning system, comprising:

[0013] Place the calibration plate in at least 15 different postures, and obtain N pictures of the calibration plate with the laser on and N pictures of the calibration plate with the laser off in the corresponding postures;

[0014] Extract corner points from N calibration plate images taken with the laser turned off, and then use the classic Zhang calibration method to calculate the camera's intrinsic and extrinsic parameters;

[0015] Take N calibration plate images with the laser on, and use the Steger algorithm to extract the pixel coordinates of the laser center line. Use the calculated camera intrinsic and extrinsic parameters to solve for the 3D coordinates corresponding to the laser line coordinates. Finally, use the 3D coordinates of the laser line under the N poses to calculate the plane equation of the laser plane.

[0016] While keeping the linear translation stage in constant motion, use the camera fixed on the linear translation stage to continuously shoot the calibration plate in a fixed pose to obtain at least 10 pictures;

[0017] The obtained camera intrinsic parameters are used to calculate the 3D coordinates of the checkerboard corners of the calibration plate captured by the camera at different positions. Finally, the direction vector of the translation stage is obtained by calculating the average value of the 3D coordinate differences of each checkerboard corner at different camera positions.

[0018] Use the camera to simultaneously photograph the calibration plate at its initial position, obtaining at least 10 high-precision virtual images of the calibration plate and its image in the mirror.

[0019] The high-precision calibration plate and its virtual image in the mirror are used to obtain multiple three-dimensional feature point pairs (respectively and ), then the Levenberg-Marquardt algorithm with bundle adjustment strategy is used to obtain the reflection matrix of the plane mirror at the initial position of the camera;

[0020] Projecting a line laser on the surface of an object to obtain a laser stripe image of the deformation of the object's surface;

[0021] Using the obtained plane equation of the participating laser plane in the camera in combination with the obtained laser stripe image, three-dimensional data of the object line profile in one real perspective and two virtual perspectives are obtained;

[0022] Calculate the reflection matrix of the plane mirror at the current position using the reflection matrix of the plane mirror at the initial position of the camera and the direction vector of the translation stage;

[0023] Then, the reflection matrix is ​​used to transform the three-dimensional data of the object's line profile at the virtual perspective to its real position, and merged with the three-dimensional data of the real perspective to obtain the panoramic line profile data of the object at that position;

[0024] As the linear translation stage slides, the camera collects the line profile information of the object at different positions, merges and reconstructs the panoramic line profile data of the object at all positions, and realizes the panoramic three-dimensional shape measurement of the object.

[0025] As a further improvement, the reflection matrix calibration of the plane mirror of the camera at the initial position of the present invention includes:

[0026] After obtaining N sets of feature point pairs (including real points and virtual points ), a Levenberg-Marquardt nonlinear optimization algorithm with a bundle adjustment strategy is used to accurately calculate the reflection matrix of the plane mirror.

[0027] Let G = {a m ,b m ,c m ,d}, we get:

[0028]

[0029] g1(G)=[1-2(a m ) 2 ]x v -2a m b m y v -2a m c m z v +2a m dx r

[0030] g2(G)=-2a m b m x v +[1-2(b m ) 2 ]y v -2b m c m z v +2bm dy r

[0031] g3(G)=-2a m c m x v -2b m c m y v +[1-2(c m ) 2 ]z v +2c m dz r

[0032] Where g1(G), g2(G), g3(G) are residual errors, and N is the number of three-dimensional point pairs; however, due to the limited manufacturing quality of the plane mirror, the virtual point The accuracy of cannot be guaranteed, which will introduce systematic errors to the final calibration results. Therefore, in order to overcome this shortcoming, a bundle adjustment strategy is introduced to optimize the coordinates of the virtual points as variables, that is:

[0033]

[0034] Minimizing the above formula is a nonlinear minimization problem that can be solved by the Levenberg-Marquardt algorithm.

[0035] As a further improvement, the calculation of the reflection matrix of the plane mirror at the current position of the present invention includes: Since the camera position in the proposed system changes with time, we take the camera coordinate system at the initial scanning position as the global coordinate system (X0, Y0, Z0) of the system, and the camera coordinate system at the i-th scanning position as the i-th local coordinate system (X i ,Y i ,Z i ). Because the motion of the linear displacement slide is rigid motion and only includes translation motion but not rotation motion, the transformation relationship between the global coordinate system and the i-th local coordinate system can be expressed as:

[0036]

[0037] Where step is the distance the stage moves each time.

[0038] Because the relative position of the camera and the plane mirror is constantly changing, the reflection matrix of the plane mirror in each local coordinate system changes accordingly. In the global coordinate system, the plane equation of the plane mirror can be described as:

[0039] in

[0040] Then the equation of the plane mirror in the i-th local coordinate system can be expressed as

[0041]

[0042] Right now

[0043] It can be found that in different local coordinate systems, the normal vector of the plane mirror is the same, and the only thing that changes is the distance between the plane mirror and the origin of the coordinate system, that is,

[0044]

[0045] In the following, we will use To describe the unit normal vector of the plane mirror in the i-th local coordinate system, the distance from the plane mirror to the origin in the i-th local coordinate system can be described as:

[0046]

[0047] Then the reflection matrix of the plane mirror in the local coordinate system can be expressed as:

[0048]

[0049] As a further improvement, the present invention converts the three-dimensional data of the object line profile in the virtual perspective to its real position, including:

[0050] In order to reconstruct the 360-degree surface of an object, each part of the object surface should be reconstructed at its real position. The image captured by the plane mirror is a virtual image. The positional relationship between the plane mirror and the camera is required to transform the reconstructed virtual surface to its real position through reflection transformation, and then calculate the world coordinates of the point in the real space.

[0051] Compared with the prior art, the present invention has the following beneficial effects:

[0052] 1. Simple setup and low cost: Only one laser scanning system and two mirrors are needed to achieve 360-degree panoramic measurement. Compared with traditional multi-view panoramic surface measurement systems that require multiple sets of synchronized laser stereo scanning systems, the required cost and space are greatly reduced.

[0053] 2. Ease of implementation: Traditional panoramic measurement systems require either complex calibration or complex point cloud registration algorithms (e.g., iterative closest point (ICP)). The proposed method only requires the calibration of a laser scanning system and two plane mirrors to merge 3D scan data obtained from multiple angles, thereby achieving direct full-surface 360-degree 3D surface measurement.

[0054] 3. More application scenarios: When using a plane mirror to assist in multi-view imaging, the plane mirror should maintain a certain distance from the sample to be measured so that the back of the sample is clearly visible. This will cause a significant depth difference between the real surface and the virtual surface, making it impossible for all areas of interest to be clearly imaged at the same time. This phenomenon is more obvious when the sample to be measured is large. Thanks to the characteristics of the TLS reconstruction algorithm and the maturity of the laser line centerline extraction algorithm, accurate extraction of the centerline can be achieved even when the laser line is widened due to defocusing. This allows the proposed system to collect three-dimensional data of the sample surface within a larger depth range, that is, to achieve panoramic three-dimensional reconstruction of larger samples. At the same time, since the laser has the characteristics of concentrated and stable energy, the proposed system is not easily affected by external lighting conditions, and can also avoid measurement errors caused by surface reflectivity differences caused by rich textures on the sample surface or local pixel saturation caused by high-reflective surfaces. Therefore, the technical solution of the present invention has a wider range of applications, and samples with larger volumes, richer surface textures and more types of materials can also use a plane mirror-assisted method to achieve panoramic measurement. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] Figure 1 It is a structural schematic diagram of the device of the present invention;

[0056] Figure 2 It is a data flow diagram of the method of the present invention;

[0057] Figure 3 It is a diagram of the plane mirror calibration device of the present invention;

[0058] In the figure, 1 is a plane mirror, 2 is a linear translation stage, 3 is a camera, 4 is a line laser, 5 is a computer, 6 is a controller, 7 is an object to be measured, 8 is the image of the object to be measured in the plane mirror, and 10 is a calibration plate. DETAILED DESCRIPTION

[0059] The technical solution of the present invention will be further described below with reference to the accompanying drawings.

[0060] Figure 1Figure 2 is a schematic diagram of the structure of the apparatus of the present invention; the system includes a camera 3, a line laser 4, a linear stage 2, a controller 6, two plane mirrors 1 for capturing panoramic images, and a computer 5 for processing the data. The camera 3 and line laser 4 have a fixed relative position and are fixed on a linear slide. A certain height difference is required between the laser and the camera 3 to produce the parallax required for 3D reconstruction. The two mirrors are placed behind the object to be measured 7, forming an angle of approximately 120 degrees. The linear stage 2 is connected to the controller 6 via a data cable, and the camera 3 and controller 6 are connected to the computer 5, enabling synchronous control of the camera 3 and the linear stage 2. Furthermore, to ensure a continuous laser stripe projected onto the object, the laser plane must be kept as perpendicular as possible to the two plane mirrors 1. This configuration ensures that the stereoscopic scanning system can simultaneously capture views of the front surface and both rear surfaces of the object. 8 is the image of the object to be measured in the plane mirrors. In this configuration, all three surface areas of the object can be captured by the laser stereo scanning system in a single shot.

[0061] In order to ensure that the laser stripes projected onto the object are continuous, the laser plane needs to be kept as perpendicular as possible to the two plane mirrors 1. This configuration ensures that the stereo scanning system can simultaneously capture views of the front surface and two rear surfaces of the object. In this configuration, all three surface parts of the object can be captured by the laser stereo scanning system in one shot.

[0062] The present invention also discloses a complex surface panoramic measurement method based on a mirror-assisted multi-view three-dimensional laser scanning system. Figure 2 It is a data flow diagram of the method of the present invention, comprising:

[0063] The calibration plate 10 is placed in at least 15 different postures, and N pictures of the calibration plate 10 are obtained with the laser on and N pictures of the calibration plate 10 with the laser off in the corresponding postures;

[0064] Extract corner points from N calibration plate 10 images taken with the laser turned off, and then use the classic Zhang calibration method to calculate the intrinsic and extrinsic parameters of camera 3;

[0065] N calibration plate 10 images are taken under the condition of split laser, and the sub-pixel coordinates of the laser center line are extracted using the Steger algorithm. The three-dimensional coordinates corresponding to the laser center line coordinates are obtained by using the calculated intrinsic and extrinsic parameters of camera 3. Finally, the plane equation of the laser plane (a) is calculated using the three-dimensional coordinates of the laser center line under the N poses. l x+b l y+c l z+d l =0);

[0066] While the linear translation stage 2 is kept in constant motion, the camera 3 fixed on the linear translation stage 2 is used to continuously photograph the calibration plate 10 in a fixed position to obtain at least 10 pictures;

[0067] The obtained camera 3 internal parameters are used to calculate the three-dimensional coordinates of the checkerboard corners of the calibration plate 10 captured by the camera 3 at different positions. Finally, the direction vector of the translation stage is obtained by calculating the average value of the three-dimensional coordinate differences of each checkerboard corner at different camera 3 positions.

[0068] The calibration plate 10 is placed in at least 10 different postures, and the camera 3 is used to simultaneously capture the calibration plate 10 and its virtual image in the plane mirror 1 at the initial position, as shown in the attached figure. Figure 3 As shown, at least 10 high-precision calibration plates 10 and their virtual images in the mirror are obtained;

[0069] The calibration plate 10 and its virtual image in the mirror are used to obtain multiple three-dimensional feature point pairs (real points and virtual points ).

[0070] Let G = {a m ,b m ,c m ,d}, we get:

[0071]

[0072] g1(G)=[1-2(a m ) 2 ]x v -2a m b m y v -2a m c m z v +2a m dx r

[0073] g2(G)=-2a m b m x v +[1-2(b m ) 2 ]y v -2b m c m z v +2b m dy r

[0074] g3(G)=-2a m c m x v-2b m c m y v +[1-2(c m ) 2 ]z v +2c m dz r

[0075] Where g1(G), g2(G), g3(G) are residual errors, N is the number of three-dimensional point pairs; limited by the manufacturing quality of the plane mirror 1, the virtual point The accuracy cannot be guaranteed, and a systematic error is introduced into the final calibration result. A bundle adjustment strategy is introduced, and the coordinates of the virtual points are optimized as variables, which is:

[0076]

[0077] Minimizing the above formula is a nonlinear minimization problem, which is solved by the Levenberg-Marquardt algorithm. By solving the above equation, the plane parameters of plane mirror 1 can be obtained, and then the reflection matrix of plane mirror 1 at the initial position of camera 3 can be obtained. (in );

[0078] Projecting a line laser on the surface of an object, and simultaneously obtaining three laser stripe images of the object's surface deformation;

[0079] Using the obtained plane equation of the participating laser plane in camera 3 combined with the obtained laser stripe image, a laser scanning system 3D reconstruction algorithm is used, and the same calibration parameters are used to simultaneously obtain 3D data of the object line profile in one real perspective and two virtual perspectives;

[0080] The reflection matrix of the plane mirror 1 at the initial position of the camera 3 and the direction vector of the translation stage are used to calculate the reflection matrix of the plane mirror 1 at the current position: the coordinate system of the camera 3 at the initial position of the scan is used as the global coordinate system (X0, Y0, Z0) of the system, and the coordinate system of the camera 3 at the i-th scanning position is used as the i-th local coordinate system (X i ,Y i ,Z i ). The transformation relationship between the global coordinate system and the i-th local coordinate system is expressed as:

[0081]

[0082] Where step is the distance the stage moves each time.

[0083] In the global coordinate system, the plane equation of plane mirror 1 is described as:

[0084] in

[0085] Then the equation of plane mirror 1 in the i-th local coordinate system can be expressed as

[0086] Right now

[0087] Then the plane equation parameters of the plane mirror 1 of the camera 3 at the i-th position can be expressed as:

[0088]

[0089] in is the unit normal vector of plane mirror 1 in the i-th local coordinate system. The distance between plane mirror 1 and the origin in the i-th local coordinate system can be described as:

[0090]

[0091] Then the reflection matrix of plane mirror 1 in the local coordinate system can be expressed as:

[0092]

[0093] Then, the reflection matrix of the plane mirror 1 at the current position is used to convert the three-dimensional data of the object line profile of the virtual perspective to its real position, and merge it with the three-dimensional data of the real perspective to obtain the panoramic line profile data of the object at this position, that is:

[0094]

[0095] As the linear translation stage 2 slides, the camera 3 collects line profile information of the object at different positions, merges and reconstructs the panoramic line profile data of the object at all positions, and realizes the panoramic three-dimensional shape measurement of the object.

[0096] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the core technical features of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

[0097] [1] E. Epstein, M. Granger-Piché, P. Potilin, Exploiting Mirrors in Interactive Reconstruction with Structured Light, VMV2004, pp.125-132.

[0098] [2]B.Chen,B.J.M.Pan,Mirror-assisted panoramic-digital imagecorrelation for full-surface360-deg deformation measurement,132(2019)350-358.

Claims

1. A measurement method based on a mirror-assisted multi-view 3D laser scanning system, characterized in that: The system includes a camera, a line laser, a linear translation stage, a controller, a computer connected to the camera and the controller for processing data, and two plane mirrors for capturing panoramic images. The camera and line laser are fixed on a sliding bracket, and the sliding bracket slides up and down relative to the linear translation stage. A height difference exists between the laser and the camera to generate the parallax required for three-dimensional reconstruction. The two plane mirrors are placed behind the object to be measured and form an angle. The linear translation stage is connected to the controller. The laser plane is perpendicular to the two plane mirrors, and the angle between the two plane mirrors is 120 degrees. The method specifically includes the following steps: placing the calibration plate in at least 15 different postures, obtaining N calibration plate images with the laser on and N calibration plate images with the laser off in the corresponding postures; Extract corner points from N calibration plate images taken with the laser turned off, and then use the classic Zhang calibration method to calculate the camera's intrinsic and extrinsic parameters; Take N calibration plate images with the laser on, and use the Steger algorithm to extract the pixel coordinates of the laser center line. Use the calculated camera intrinsic and extrinsic parameters to solve for the 3D coordinates corresponding to the laser line coordinates. Finally, use the 3D coordinates of the laser line under the N poses to calculate the plane equation of the laser plane. While keeping the linear translation stage in constant motion, use the camera fixed on the linear translation stage to continuously shoot the calibration plate in a fixed pose to obtain at least 10 pictures; The obtained camera intrinsic parameters are used to calculate the 3D coordinates of the checkerboard corners of the calibration plate captured by the camera at different positions. Finally, the direction vector of the translation stage is obtained by calculating the average value of the 3D coordinate differences of each checkerboard corner at different camera positions. Use the camera to simultaneously shoot the calibration plate at the initial position to obtain at least 10 high-precision calibration plates and their virtual images in the mirror; use the captured high-precision calibration plates and their virtual images in the mirror to obtain multiple three-dimensional feature point pairs, which are and Then, the Levenberg-Marquardt algorithm with bundle adjustment strategy is used to obtain the reflection matrix of the plane mirror at the initial position of the camera; Projecting a line laser on the surface of an object to obtain a laser stripe image of the deformation of the object's surface; Using the obtained plane equation of the participating laser plane in the camera in combination with the obtained laser stripe image, three-dimensional data of the object line profile in one real perspective and two virtual perspectives are obtained; Calculate the reflection matrix of the plane mirror at the current position using the reflection matrix of the plane mirror at the initial position of the camera and the direction vector of the translation stage; Then, the reflection matrix is ​​used to transform the three-dimensional data of the object's line profile at the virtual perspective to its real position, and merged with the three-dimensional data of the real perspective to obtain the panoramic line profile data of the object at that position; As the linear translation stage slides, the camera collects the line profile information of the object at different positions, merges and reconstructs the panoramic line profile data of the object at all positions, and realizes the panoramic three-dimensional shape measurement of the object.

2. The measurement method of the mirror-assisted multi-view 3D laser scanning system according to claim 1, characterized in that: The reflection matrix calibration of the plane mirror of the camera at the initial position includes: After obtaining N sets of feature point pairs, including real points and virtual points A Levenberg-Marquardt nonlinear optimization algorithm with a bundle adjustment strategy is used to accurately calculate the reflection matrix of the plane mirror. Let G = {a m ,b m ,c m ,d}, we get: g1(G)=[1-2(a m ) 2 ]x v -2a m b m y v -2a m c m z v +2a m d-x r g2(G)=-2a m b m x v +[1-2(b m ) 2 ]y v -2b m c m z v +2b m d-y r g3(G)=-2a m c m x v -2b m c m y v +[1-2(c m ) 2 ]z v +2c m d-z r Where g1(G), g2(G), g3(G) are residual errors, N is the number of 3D point pairs; limited by the manufacturing quality of the plane mirror, the virtual point The accuracy cannot be guaranteed, and a systematic error is introduced into the final calibration result. A bundle adjustment strategy is introduced, and the coordinates of the virtual points are optimized as variables, which is: Minimizing the above formula is a nonlinear minimization problem, which is solved by the Levenberg-Marquardt algorithm.

3. The measurement method of the mirror-assisted multi-view 3D laser scanning system according to claim 2, characterized in that: The calculation of the reflection matrix of the plane mirror at the current position includes: The camera coordinate system at the initial scanning position is used as the global coordinate system (X0, Y0, Z0) of the system, and the camera coordinate system at the i-th scanning position is used as the i-th local coordinate system (X i ,Y i ,Z i ), the transformation relationship between the global coordinate system and the i-th local coordinate system can be expressed as: Where step is the distance the stage moves each time; In the global coordinate system, the plane equation of the plane mirror is described as: in Then the equation of the plane mirror in the i-th local coordinate system can be expressed as Right now In different local coordinate systems, the normal vector of the plane mirror is the same, and the only thing that changes is the distance between the plane mirror and the origin of the coordinate system, that is, Unified use To describe the unit normal vector of the plane mirror in the i-th local coordinate system, the distance from the plane mirror to the origin in the i-th local coordinate system is described as: Then the reflection matrix of the plane mirror in the local coordinate system is expressed as:

4. The measurement method of the mirror-assisted multi-view 3D laser scanning system according to claim 1, 2 or 3, characterized in that: Converting the 3D data of the object line outline in the virtual perspective to its real position includes: In order to reconstruct the 360-degree surface of the object, each part of the object surface is reconstructed at its real position. The image captured by the plane mirror is a virtual image. The positional relationship between the plane mirror and the camera is required. The reconstructed virtual surface is converted to its real position through reflection transformation, and then the world coordinates of the virtual point in the real space are calculated. The reflection matrix can be used to merge the 3D data of the virtual perspective and the real perspective: in is the 3D coordinate of the reconstructed virtual point, are the coordinates of the real position of the virtual point, is the reflection matrix of the plane mirror.

Images