CN101246595A - Multi-view point cloud data combination method in optical 3D scanning system - Google Patents

Abstract

A method for combining multi-view point cloud data in an optical 3D scanning system. The method is realized with the support of an optical scanning imaging system, and includes the following steps: ① Use a patch composed of a white solid circle in the middle and a black periphery as a reference unit; ② Paste at least three reference units on the scanned object; ③ Direct information extraction, obtain the data of the center of the reference unit, ④ convert the obtained reference unit center data into the centroid position information of the three-dimensional reference unit, ⑤ calculate the distance matrix with the same number as the number of digital cameras by means of the analytical method, ⑥ determined as two For the corresponding public reference points in the point cloud set, ⑦Use at least three pairs of corresponding coordinate parameters corresponding to the public reference points to find out the rotation matrix R and translation matrix T about the conversion of the two coordinate systems, and then according to the conversion formula: P=R*P′ +T directly completes the transformation of the point cloud coordinate data, and completes the combination of two point cloud data. The advantage of the present invention is that the data combination is simple and accurate.

Description

Multi-view point cloud data combination method in optical 3D scanning system

technical field

The method belongs to the computer digital image processing technology, specifically a multi-viewpoint cloud data combination method in an optical three-dimensional scanning system.

Background technique

The 3D scanning system has been continuously improved and developed in practical applications, and a complete location cloud data set that can reproduce the image of the object can be obtained by means of direct shooting of the scanned object. However, it is difficult to obtain the complete point cloud data of the object through one measurement when surveying and mapping large-scale objects. A camera often cannot scan the panorama, and it is often necessary to convert different angles for measurement; or use at least two or more digital cameras. The point cloud data set that can reflect the complete appearance of the object can only be obtained by scanning and measuring at the same time. Only the relevant data of objects in a certain viewing angle can be obtained each or every time, and only the complete point cloud data can be obtained, and the point cloud data of different viewing angles need to be stitched together.

At present, the existing stitching methods mainly include: 1) stitching based on the premise of the topological feature of the surface, that is, the curvature of the surface, that is, searching for points with higher curvature in the point cloud data, and using these feature information as a reference point to stitch the local point cloud. This makes those surfaces smooth, surface feature information is not obvious, and point cloud data without excessive curvature is difficult to get a better stitching effect. And it is a time-consuming task to calculate the curvature of each point in the point cloud data. The selection of the number of neighborhood points directly affects the calculation result of the curvature of the point, and there are many uncertain factors; ②The method of human-computer interaction. The typical program steps are to manually select several pairs of corresponding points, first to complete the overall merging, that is, the rough merging work, and then use the ICP algorithm to perform iterative operations on each element in the public point cloud data set to achieve the best merging result. . The disadvantage of this method is that it has higher requirements on the operator, and the selected corresponding points are required to be accurate, and the ICP algorithm is more sensitive to the initial merging results. If the deviation of the rough merging is large, it will bring a large amount of calculation for the precise merging .

Contents of the invention

The purpose of the present invention is to find a method for quickly and accurately realizing data stitching for the separately independent point cloud data appearing in the multi-viewpoint scanning imaging system, realize the standardization of three-dimensional point cloud data acquired under different viewing angles, and seek a precision High, fast, and widely applicable methods are used as guidance to realize the stitching of point cloud data from different perspectives. The design basis of the present invention is based on the coordinate data established by the coordinate information of the online reference system obtained during the point cloud data extraction process of the image three-dimensional scanning system. The process of merging point cloud data is actually a process of coordinate transformation between two reference systems. Therefore, the method of the present invention is the operation process of finding out the undetermined coefficients of the two coordinate system transformation formulas.

This method is realized with the support of a computer, a supporting memory with optical scanning imaging system management software, and at least two digital cameras. The key point is that this method includes the following steps:

①A patch consisting of a white solid circle in the middle and a black closed area outside the circle is used as a reference unit,

②Paste at least three reference units on the scanned object within the range that the relevant digital cameras at different viewpoints can scan at the same time,

③ With the help of the online operating system, the reference unit pasted on the scanned object is directly extracted to obtain the data information of the center of the reference unit, and then the black and white transition boundary of the reference point is extracted by using the area identification method, and the least squares is used to extract the reference unit. The method of combining multiplication and multiplication to obtain the center position data of the reference unit,

④ Use the computer vision theory to perform stereo matching on the centroid position data of the reference unit obtained above, that is, convert the center data of the plane reference unit into the centroid position information of the three-dimensional reference unit,

⑤According to the centroid position information of the reference unit, refer to the data information of each point in the point cloud set obtained by scanning and mapping of the digital camera, and use the analytical method to obtain the distance parameter between the selected reference point in the point cloud set and the centroid position of the reference unit form the same number of distance matrices as the number of digital cameras employed,

⑥ By comparing the element values in the two matrices to find their largest isomorphic subset, so as to find the reference point with the same topological characteristics, and determine it as the corresponding common reference point in the two point cloud sets,

⑦ Use at least three pairs of corresponding coordinate parameters corresponding to the common reference points to find out the rotation matrix R and translation matrix T about the conversion of the two coordinate systems, and then directly complete the point cloud coordinate data according to the conversion formula: P=R*P′+T The transformation is completed on the two point clouds

In fact, this method uses the designed reference patches to artificially set boundaries and labels for large objects to be scanned and mapped. By processing the patch information in point cloud datasets collected from two different viewing angles, a sufficient The public corresponding points with reference significance. All corresponding points point to the same position but have completely different coordinate information, thus providing the necessary conditions for finding out the undetermined coefficients in the conversion formula between the corresponding position points of the two coordinate systems. Converting each point in the point cloud of different coordinate systems according to the formula can be easily completed, that is, the task of data stitching is easily and accurately completed. With the assistance of the existing system and management software, the complex and difficult work in the prior art becomes so simple and accurate under the guidance of the method and the support of the computer system, which is the positive effect of the present invention.

Description of drawings

Figure 1 is a patch with a circle as the reference unit.

Figure 2 is a patch with a rectangle as the reference unit.

Fig. 3 is a flow chart of merging equivalence marks during processing.

Detailed ways

A method for combining multi-view point cloud data in an optical three-dimensional scanning system. The method is implemented with the support of a computer, a supporting memory with optical scanning imaging system management software, and at least two digital cameras. The method includes the following steps:

①A patch consisting of a white solid circle in the middle and a black closed area outside the circle is used as the reference unit;

②Paste at least three reference units on the scanned object within the range that relevant digital cameras at different viewpoints can scan simultaneously;

③ With the help of the online operating system, the reference unit pasted on the scanned object is directly extracted to obtain the data information of the center of the reference unit, and then the black and white transition boundary of the reference point is extracted by using the area identification method, and the least squares is used to extract the reference unit. The method of combining multiplication to obtain the center position data of the reference unit;

④ Use the computer vision theory to perform stereo matching on the centroid position data of the reference unit obtained above, that is, convert the center data of the plane reference unit into the centroid position information of the three-dimensional reference unit;

⑤According to the centroid position information of the reference unit, refer to the data information of each point in the point cloud set obtained by scanning and mapping of the digital camera, and use the analytical method to obtain the distance parameter between the selected reference point in the point cloud set and the centroid position of the reference unit Form the same distance matrix as the number of digital cameras used;

⑥ Find the largest isomorphic subset by comparing the element values in the two matrices, so as to find the reference point with the same topological characteristics, and determine it as the corresponding common reference point in the two point cloud sets;

⑦ Use at least three pairs of corresponding coordinate parameters corresponding to the common reference points to find out the rotation matrix R and translation matrix T about the conversion of the two coordinate systems, and then directly complete the point cloud coordinate data according to the conversion formula: P=R*P′+T Transformation to complete the combination of two point cloud data.

During specific implementation, the reference unit used in step ① is a black patch with a white solid circle in the middle of a circle or a rectangle.

The reference unit mentioned in step ② mentioned above is evenly attached to the common area that can be scanned by the two digital cameras.

The above step ③ uses the online operating system to directly extract the reference unit pasted on the scanned object. It uses the method of combining area identification and least squares to obtain the centroid position of the reference point. First, the sequential algorithm is used to scan Image, the image of the object to be tested is identified and divided into equivalent pairs, the marked points belonging to the same class are merged, and the same equivalent mark is given; then the boundary of the reference point is obtained by using the boundary point recognition method, that is, for the boundary inside point, the points in the eight directions around this point are all black; otherwise, the points on the boundary do not have such properties; The centroid position of the reference point and related data information are stored in the intermediate register for use.

A more specific operation process is as follows: firstly, median filter processing is performed on the collected image including the reference unit, so as to remove the noise in the image. Then, the dynamic threshold segmentation method is used to realize the binarization of the image, and the white is recorded as 1, and the black is recorded as 0. In the process of scanning the image from left to right and from top to bottom, if the gray value of the current pixel is 1, check its 4 adjacent pixels on the left, top left, top, and top right. If the gray values of the above four adjacent pixels are all 0, give the current pixel a new position marker value. If only one of the four adjacent pixels has a gray value of 1, assign the position marker value of this pixel to the current pixel. If the gray value of m (1<m<=4) pixels among the 4 adjacent pixels is 1, then determine the position mark value of the current pixel according to the priority order of left, upper left, upper, upper right. Then record the equivalence pairs for the tag values owned by these m pixels, and put them into an equivalence pair linked list. Fig. 3 shows the relevant neighbor points when calculating the position number of point O. It can be seen that the marked value of point O is related to the marked values of A, B, C, and D. If A, B, C, D have a value of 1 in the binary image, the mark value of O takes the minimum value of the mark value, otherwise, add 1 to the current maximum mark value and assign it to point O. If there are marked values in A, B, C, and D, then mark the marked value with high priority at point O, and treat the remaining different marked values as equivalent, thus establishing a serial number matrix corresponding to the binary image M and the equivalent pair linked list EList. Arrange the equivalent pairs in the EList, remove the duplicates, and mark all the equivalent pixels with the same mark. Although EList contains all equivalence pairs, there may be cases where the same tag value corresponds to different tag values. For example, after sequential connection, there may be cases where 4 and 5 are equivalent and 4 and 3 are equivalent. In this way It needs to be processed to mark 3, 4, and 5 as the same tag value. Then construct a double-layer linked list SList, first store the head node in the equivalent pair linked list EList in the SList, and do the following processing for each other node in the EList: For the two tag values a and b in the current node of the EList, Traversing the SList, if a is found, record the pointer of the internal linked list containing a in pa, otherwise pa is empty; if b is found, record the pointer of the internal linked list containing b in pb, otherwise pb is empty. According to the values of pa and pb, do the processing as shown in Figure 3. After the above processing, store all the equivalent tag linked lists in the SList, and replace each group of equivalent pixel tags with the minimum tag value in the equivalent tag chain, so that all connected regions are marked with new Same mark.

Boundary extraction is performed on the region of the same type of mark. For the point inside the boundary, the points in the 8 directions around this point are all black; otherwise, the points on the boundary do not have such properties. By using this method, the closed edge point data of the reference point can be obtained and stored in a linked list. Use the least squares fitting method to find the centroid of the area enclosed by the boundary pixels:

Let the given sample point be p _i ( _xi , y _i ), i=1...n.

can get:

${<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; -</span>{\stackrel{<span\; class="notranslate">\; ^</span>}{\mathrm{<span\; class="notranslate">\; u</span>}}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; )</span>}^{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; +</span>{<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; the\; y</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; -</span>{\stackrel{<span\; class="notranslate">\; ^</span>}{\mathrm{<span\; class="notranslate">\; u</span>}}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; )</span>}^{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; =</span>{<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; -</span>{\stackrel{<span\; class="notranslate">\; ^</span>}{\mathrm{<span\; class="notranslate">\; u</span>}}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; )</span>}^{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; +</span>{<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; the\; y</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; -</span>{\stackrel{<span\; class="notranslate">\; ^</span>}{\mathrm{<span\; class="notranslate">\; u</span>}}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; )</span>}^{\mathrm{<span\; class="notranslate">\; 2</span>}}$

The best unbiased estimate of the geometric centroid can be obtained by the above formula:

${\stackrel{<span\; class="notranslate">\; ^</span>}{\mathrm{<span\; class="notranslate">\; u</span>}}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; =</span>\frac{\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}{{\mathrm{<span\; class="notranslate">\; b</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}}^{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; \&CenterDot;</span>\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}{\mathrm{<span\; class="notranslate">\; a</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}{\mathrm{<span\; class="notranslate">\; c</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; -</span>\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}{\mathrm{<span\; class="notranslate">\; b</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}{\mathrm{<span\; class="notranslate">\; c</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; \&Center\; Dot;</span>\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}{\mathrm{<span\; class="notranslate">\; a</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}{\mathrm{<span\; class="notranslate">\; b</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}}{\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}{{\mathrm{<span\; class="notranslate">\; a</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}}^{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; \&CenterDot;</span>\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}{{\mathrm{<span\; class="notranslate">\; b</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}}^{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; -</span>{<span\; class="notranslate">\; (</span>\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}{\mathrm{<span\; class="notranslate">\; a</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}{\mathrm{<span\; class="notranslate">\; b</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; )</span>}^{\mathrm{<span\; class="notranslate">\; 2</span>}}}$

${\stackrel{<span\; class="notranslate">\; ^</span>}{\mathrm{<span\; class="notranslate">\; u</span>}}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; =</span>\frac{\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}{{\mathrm{<span\; class="notranslate">\; a</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}}^{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; \&Center\; Dot;</span>\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}{\mathrm{<span\; class="notranslate">\; b</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}{\mathrm{<span\; class="notranslate">\; c</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; -</span>\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}{\mathrm{<span\; class="notranslate">\; a</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}{\mathrm{<span\; class="notranslate">\; c</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; \&Center\; Dot;</span>\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}{\mathrm{<span\; class="notranslate">\; a</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}{\mathrm{<span\; class="notranslate">\; b</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}}{\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}{{\mathrm{<span\; class="notranslate">\; a</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}}^{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; \&CenterDot;</span>\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}{{\mathrm{<span\; class="notranslate">\; b</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}}^{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; -</span>{<span\; class="notranslate">\; (</span>\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}{\mathrm{<span\; class="notranslate">\; a</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}{\mathrm{<span\; class="notranslate">\; b</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; )</span>}^{\mathrm{<span\; class="notranslate">\; 2</span>}}}$

Among them: a _i =2(x _i+1 -x _i ), b _i =2(y _i+1 -y _i ),

${\mathrm{<span\; class="notranslate">\; c</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; =</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; 1</span>}}^{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; the\; y</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; 1</span>}}^{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}^{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; the\; y</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}^{\mathrm{<span\; class="notranslate">\; 2</span>}}$

Considering that there may be errors in the actual measurement, the comparison error value is set when the equidistant isomorphic subset is selected, that is, the specific steps for determining the common reference point corresponding to the two point cloud sets in the above-mentioned step ⑥ are: Determine the minimum error value δ according to the basic error of the measurement system, and find out the largest isomorphic subset from the element values in the two matrices on the condition that the comparison error value is less than δ. Further find the common corresponding point pairs.

According to the obtained corresponding point pairs (more than 3 pairs), use Singular Value Decomposition (SVD) to complete the solution of the rotation matrix and translation vector to obtain the relevant coordinate transformation formula and further complete the conversion and point cloud data combination.

Further explanation is as follows:

Suppose the set of reference points corresponding to point cloud A is P={p ₁ , p ₂ ...p _n }, and the set of reference points corresponding to point cloud B is Q={q ₁ , q ₂ ...q _n }, the two The corresponding relationship of points in the reference point set can be described as q _i =R×p _i +T i=1, 2,...,n, n>3

In the formula, R is a 3×3 rotation matrix, T is a 3-dimensional translation vector, and n is the number of reference point pairs. Create the following error function

$\mathrm{<span\; class="notranslate">\; f</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; R</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; T</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\frac{\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; no</span>}}\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; no</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}{<span\; class="notranslate">\; ||</span>{\mathrm{<span\; class="notranslate">\; q</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; R</span>}<span\; class="notranslate">\; \&times;</span>{\mathrm{<span\; class="notranslate">\; p</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; T</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; ||</span>}^{\mathrm{<span\; class="notranslate">\; 2</span>}}$

The SVD algorithm is used to solve the placement matrix and translation vector, so that

${\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; 3</span>}<span\; class="notranslate">\; \&times;</span>\mathrm{<span\; class="notranslate">\; 3</span>}}<span\; class="notranslate">\; =</span>\frac{\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; no</span>}}\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; no</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; p</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; -</span>\stackrel{<span\; class="notranslate">\; \&OverBar;</span>}{\mathrm{<span\; class="notranslate">\; p</span>}}<span\; class="notranslate">\; )</span>{<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; q</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; -</span>\stackrel{<span\; class="notranslate">\; \&OverBar;</span>}{\mathrm{<span\; class="notranslate">\; q</span>}}<span\; class="notranslate">\; )</span>}^{\mathrm{<span\; class="notranslate">\; T</span>}}$

in: $\stackrel{\&OverBar;}{p}=\frac{1}{\mathrm{no}}\underset{i=1}{\overset{\mathrm{no}}{\mathrm{\&Sigma;}}}{p}_i,$ $\stackrel{\&OverBar;}{q}=\frac{1}{\mathrm{no}}\underset{i=1}{\overset{\mathrm{no}}{\mathrm{\&Sigma;}}}{q}_i$ are the centroids of P and Q, respectively. Singular value decomposition of matrix H is obtained: H=UDV ^T , (D=diag(d _i ), d ₁ ≥d ₂ ≥d ₃ ≥0),

Rotation matrix: R=UV ^T

Translation vector: T=p-R×q

Solve the rotation matrix R and the translation matrix T, that is, the combination of two point cloud data can be realized, and finally a complete point cloud data of an object can be obtained.

Claims (5)

1. A method for combining multi-viewpoint cloud data in an optical three-dimensional scanning system. The method is implemented with the support of a computer, a supporting memory with optical scanning imaging system management software, and at least two digital cameras. It is characterized in that The method includes the following steps: ①A patch consisting of a white solid circle in the middle and a black closed area outside the circle is used as a reference unit, ②Paste at least three reference units on the scanned object within the range that the relevant digital cameras at different viewpoints can scan at the same time, ③ With the help of the online operating system, the reference unit pasted on the scanned object is directly extracted to obtain the data information of the center of the reference unit, and then the black and white transition boundary of the reference point is extracted by using the area identification method, and the least squares is used to extract the reference unit. The method of combining multiplication and multiplication to obtain the center position data of the reference unit, ④ Use the computer vision theory to perform stereo matching on the centroid position data of the reference unit obtained above, that is, convert the center data of the plane reference unit into the centroid position information of the three-dimensional reference unit, ⑤According to the centroid position information of the reference unit, refer to the data information of each point in the point cloud set obtained by scanning and mapping of the digital camera, and use the analytical method to obtain the distance parameter between the selected reference point in the point cloud set and the centroid position of the reference unit form the same number of distance matrices as the number of digital cameras employed, ⑥ By comparing the element values in the two matrices to find their largest isomorphic subset, so as to find the reference point with the same topological characteristics, and determine it as the corresponding common reference point in the two point cloud sets, ⑦ Use at least three pairs of corresponding coordinate parameters corresponding to the common reference points to find out the rotation matrix R and translation matrix T about the conversion of the two coordinate systems, and then directly complete the point cloud coordinate data according to the conversion formula: P=R*P′+T Transformation to complete the combination of two point cloud data. 2. The method for combining multi-view point cloud data in an optical three-dimensional scanning system according to claim 1, wherein the reference unit used in step ① is a circle or a black patch with a white solid circle in the middle of a rectangle. 3. The method for combining multi-view point cloud data in an optical three-dimensional scanning system according to claim 1, characterized in that the reference unit in step ② is evenly attached to the common area depicted by the two digital cameras. 4. The multi-view point cloud data merging method in the optical three-dimensional scanning system according to claim 1, characterized in that the step ③ directly extracts information from the reference unit pasted on the scanned object by means of an online operating system, and uses the area The method of combining identification and least squares to obtain the centroid position of the reference point, first scans the image with a sequential algorithm, makes the area identification and equivalence pair division of the image of the object to be measured, merges the marking points belonging to the same class, and assigns The same equivalent mark; then use the boundary point recognition method to obtain the boundary of the reference point, that is, for the point inside the boundary, the points in the 8 directions around this point are all black; otherwise, the points on the boundary do not have such properties; or judge the boundary point according to the 4 neighbors of the point, use the least square method to fit the centroid position of the reference point and related data information, and store the above related information in the intermediate register for use. 5. The multi-view point cloud data merging method in the optical three-dimensional scanning system according to claim 1, characterized in that in step ⑥, the specific step of determining that the two point cloud sets correspond to the common reference points is: according to the basic error existing in the measurement system Determine the minimum error value δ, and find out the largest isomorphic subset from the element values in the two matrices on the condition that the comparison error value is less than δ.

Images