Stereoscopic image pair point cloud registration method and computer equipment

Abstract

The invention discloses a three-dimensional image pair point cloud registration method and computer equipment, and the method comprises the steps: S1, obtaining a point cloud through a three-dimensional laser scanner with a high-precision camera, and shooting a plurality of photos with overlapped regions; s2, utilizing a plurality of photos with overlapped areas to construct a stereo image pair model for the feature points of the common area, and converting the stereo image pair model under an image space auxiliary coordinate system to a point cloud space coordinate system where a three-dimensional laser scanner is located; s3, selecting a first optimal matching feature point set and a second optimal matching feature point set; and S4, selecting a final optimal matching feature point set, and carrying out point cloud data matching alignment by using the final optimal matching feature point set. According to the method, matching can be performed more efficiently and quickly according to the feature points, and the point cloud matching precision and reliability are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of three-dimensional laser scanning, and specifically to a method for point cloud registration of a stereo pair and a computer device. Background Art

[0002] Three-dimensional point cloud data has richer object, scene information and scene applications than two-dimensional images, and point clouds can also efficiently store information data with less storage space. At present, the point cloud data collected by a single sensor can no longer meet the development needs, and point cloud matching of the point cloud data obtained by multiple sensors can more accurately describe the real three-dimensional world. However, the cross-source data collected by multiple sensors has situations such as missing point clouds, inconsistent point cloud density distributions, and different local data patterns, which are not directly applicable to point cloud matching. Therefore, preprocessing the point cloud is a necessary step.

[0003] In the prior art, several common point cloud matching algorithms are roughly the following four: ICP algorithm, feature-based method, deep learning-based method, and local feature descriptor matching method. The ICP algorithm is prone to falling into a local optimal solution, is sensitive to the initial estimate value, is sensitive to noise and outliers, and requires a good initial estimate; the feature-based method has a high computational complexity, requires high computing resources, and may require a large amount of memory and computing time; the deep learning-based method requires a large amount of labeled data for training, and has high requirements for computing resources and training time; the local feature descriptor matching method may have a low matching efficiency for large-scale data sets. Summary of the Invention

[0004] The present invention provides a method for point cloud registration of a stereo pair and a computer device that can effectively improve the matching efficiency of large-scale data sets, and can solve at least one of the above technical problems.

[0005] In order to solve the above technical problems, the present invention adopts the following technical solutions:

[0006] A method for point cloud registration of a stereo pair includes the following steps:

[0007] S1. While acquiring point clouds using a three-dimensional laser scanner with a high-precision camera, take multiple photos with overlapping regions.

[0008] S2. Use multiple photos with overlapping regions to construct a stereo pair model for the feature points in the common region. Assume an image space auxiliary coordinate system through the positional relationship of each photo. The stereo pair model is located in this image space auxiliary coordinate system, and then, through the grating positional relationship between the high-precision camera and the three-dimensional laser scanner, convert the stereo pair model in the image space auxiliary coordinate system to the point cloud space coordinate system where the three-dimensional laser scanner is located.

[0009] S3. Respectively perform two-way constraint matching on the feature point sets in the point cloud space coordinate system obtained by scanning adjacent two grating emitters in the three-dimensional laser scanner and the feature point sets in the point cloud space coordinate system solved by the stereo pair model, and select the first optimal matching feature point set and the second optimal matching feature point set;

[0010] S4. Perform two-way constraint matching on the first optimal matching feature point set and the second optimal matching feature point set again, select the final optimal matching feature point set, and perform point cloud data matching alignment with the final optimal matching feature point set.

[0011] Further, in S2, during the process of obtaining point clouds by the three-dimensional laser scanner, use a high-precision camera to take photos with a specific overlap rate, take the same feature point set P through multiple images with overlapping areas, and project the feature point set P onto the first photo or the second photo to obtain the projected feature point set p 1 or the projected feature point set p 2 , the first image space coordinate system of the first photo is different from the second image space coordinate system of the second photo. Assume the second image space coordinate system as the image space auxiliary coordinate system, and use the position relationship of the camera centers of the two photos and the length B of the photographic baseline to uniformly convert the first image space coordinate system and the second image space coordinate system to the image space auxiliary coordinate system.

[0012] Further, both the first photo and the second photo are photos taken by a high-precision camera, and both the first photo and the second photo are located between the first grating emitter and the second grating emitter. The first grating emitter is located on the left, and the second grating emitter is located on the right. When constructing the stereo pair model, the first photo taken by the high-precision camera on the right side of the first grating emitter and the second photo taken by the high-precision camera on the left side of the second grating emitter.

[0013] Further, in S2, determine the conversion parameters between the image space auxiliary coordinate system and the point cloud space coordinate system through the actual relative position relationship between the first grating emitter and the high-precision camera on the right side of the first grating emitter, and the actual relative position relationship between the second grating emitter and the high-precision camera on the left side of the second grating emitter, and solve the point cloud space coordinates of the feature point set P under the three-dimensional laser scanner.

[0014] Further, the coordinates of the projected feature point set p 1 and the projected feature point set p 2 in the image space auxiliary coordinate system are respectively {(x 1 , y 1 , z 1)|x 1 ∈ the first image space coordinate system, y 1 ∈ the first image space coordinate system, z 1 ∈ the first image space coordinate system}, {(x 2 , y 2 , z 2 )|x 2 ∈ the second image space coordinate system, y 2 ∈ the second image space coordinate system, z 2 ∈ the second image space coordinate system}, the first image space coordinate system and the second image space coordinate system respectively take the first camera center S 1 and the second camera center S 2 as the coordinate origin, the x-axis and y-axis of the first image space coordinate system and the second image space coordinate system are respectively parallel to the x-axis and y-axis of the image plane coordinate system, the image plane coordinate system is the positional relationship in the image plane where the image points are located, that is, the two-dimensional plane coordinate system where the image is located, and the z-axes of the first image space coordinate system and the second image space coordinate system coincide with the principal optical axis of the high-precision camera;

[0015] Based on the relative positional relationship between the first camera center S 1 and the second camera center S 2 and the first grating emitter and the second grating emitter, the conversion parameters between the image space auxiliary coordinate system and the point cloud space coordinate system under the three-dimensional laser scanner are calculated, and the coordinate sets of the feature points in the image space auxiliary coordinate system {(x 1 , y 1 , z 1 )|x 1 ∈ the image space auxiliary coordinate system, y 1 ∈ the image space auxiliary coordinate system, z 1 ∈ the image space auxiliary coordinate system} and {(x 2 , y 2 , z 2 )|x 2 ∈ the image space auxiliary coordinate system, y 2 ∈ the image space auxiliary coordinate system, z 2 ∈ the image space auxiliary coordinate system} are respectively converted into the point cloud space coordinates {(X K1 , Y K1 , Z K1 )|X K1 ∈ the point cloud space coordinate system where the first grating emitter is located, Y K1 ∈ the point cloud space coordinate system where the first grating emitter is located, Z K1 ∈ the point cloud space coordinate system where the first grating emitter is located} and {(X K2 , Y K2 , Z K2 )|XK2 ∈ the point cloud space coordinate system where the second grating emitter is located, Y K2 ∈ the point cloud space coordinate system where the second grating emitter is located, Z K2 ∈ the point cloud space coordinate system where the second grating emitter is located}.

[0016] Further, in the step S3, the 3D laser scanner with a high-precision camera has multiple grating emitters. Each grating emitter can obtain a set of feature points in the point cloud space coordinate system under the grating emitter through scanning, and this set of feature points is the point cloud data obtained by the grating emitter;

[0017] The set of feature points in the point cloud space coordinate system scanned by the first grating emitter among adjacent grating emitters is subjected to two-way constraint matching with the set of feature points in the point cloud space coordinate system solved by the stereo pair model, and the first optimal matching set of feature points is selected. The set of feature points in the point cloud space coordinate system scanned by the second grating emitter among adjacent grating emitters is subjected to two-way constraint matching with the set of feature points in the point cloud space coordinate system solved by the stereo pair model, and the second optimal matching set of feature points is selected;

[0018] The point cloud space coordinates obtained by scanning the first grating emitter and the second grating emitter are converted from the point cloud space coordinates obtained by scanning the first grating emitter to the point cloud space coordinates obtained by scanning the second grating emitter through the positional relationship between adjacent two grating emitters, that is, the vertical angle α, the rotation angle ω, and the distance R. The coordinate conversion formula is as follows:

[0019] X K2 = R × cosα × sinω

[0020] Y K2 = R × cosα × sinω

[0021] Z K2 = R × sinω.

[0022] Further, in the step S4, two adjacent point cloud data obtained by scanning adjacent two grating emitters have common feature points at the adjacent edges. The two point cloud data with common feature points are the point cloud data that need to be stitched and matched. The two point cloud data that need to be stitched and matched are respectively compared with the stereo pair model for two-way constraint coordinate differences, and the first optimal matching set of feature points and the second optimal matching set of feature points are matched through restricting the coordinate differences;

[0023] The first optimal matching feature set and the second optimal matching feature set are respectively located in the point cloud space coordinate system where the first grating emitter and the second grating emitter are located. After rotating, translating, and scaling the point cloud space coordinate system where the first grating emitter is located in sequence according to the positional relationship between two adjacent grating emitters, it is transformed into the point cloud space coordinate system where the second grating emitter is located. After unifying the coordinate systems, the first optimal matching feature point set and the second optimal matching feature point set are compared for the coordinate difference with two-way constraints again, and finally the final optimal matching feature point set for point cloud matching is obtained.

[0024] Further, the feature point set P includes at least high-curvature points, edge points, and corner points.

[0025] Further, the first grating emitter and the second grating emitter project structured light stripes containing encoded information onto the feature point set P.

[0026] A computer device includes a memory and a processor. When a computer program stored in the memory is executed by the processor, the processor executes the steps of the point cloud registration method for the stereo image pair.

[0027] The beneficial effects of the present invention are as follows:

[0028] 1. In the present invention, the optimal matching feature point set is obtained by two-way constraints between the feature points in the point cloud space coordinates solved by the stereo image pair model and the feature points of the point cloud data. With this feature point set, point cloud matching can be more efficient and fast based on the feature points.

[0029] 2. In the present invention, through two-way constraints twice, the matching of feature points is higher and more accurate. Point cloud matching based on these feature points will reduce the problem of point cloud data offset, making the accuracy and reliability of point cloud matching higher. Description of the Drawings

[0030] The drawings described herein are used to provide a further understanding of the present application, and constitute a part of the present application. The schematic embodiments of the present application and their descriptions are used to explain the present application, and do not constitute an improper limitation to the present application.

[0031] Figure 1 It is a schematic diagram of the overall process of the point cloud registration method according to an embodiment of the present invention.

[0032] Figure 2 It is a schematic diagram of the structure of the stereo image pair model construction according to an embodiment of the present invention.

[0033] Figure 3 It is a schematic diagram of the relative position structure between the grating emitter and the photo according to an embodiment of the present invention.

[0034] Figure 4 It is a schematic diagram of the conversion of the positional relationship between adjacent grating emitters in an embodiment of the present invention.

[0035] Figure 5 It is a structural block diagram of a computer device in an embodiment of the present invention.

[0036] The markings of each component in the drawings are: 1. Feature point set P; 2. Feature point set projection p 1 ; 3. Feature point set projection p 2 ; 4. First image space coordinate system; 5. Second image space coordinate system; 6. First photography center S 1 ; 7. Second photography center S 2 ; 8. Photography baseline length B; 9. First grating emitter; 10. Second grating emitter; 11. First photo; 12. Second photo; 13. Vertical angle α; 14. Rotation angle ω; 15. Distance R. Specific embodiments

[0037] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Without conflict, the embodiments in the present application and the features in the embodiments can be combined with each other. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0038] It should be noted that if there are descriptions involving "first", "second", etc. in the embodiments of the present invention, such descriptions of "first", "second", etc. are only for descriptive purposes and should not be understood as indicating or implying their relative importance or implicitly indicating the quantity of the indicated technical features. Thus, the features defined with "first" and "second" may explicitly or implicitly include at least one such feature. In addition, "a plurality" means more than two.

[0039] See Figure 1 - Figure 2 , an embodiment of the present invention provides a point cloud registration method for a stereo image pair, including the following steps:

[0040] S1. While acquiring point clouds using a three-dimensional laser scanner with a high-precision camera, take multiple photos with overlapping areas;

[0041] S2. Use multiple photos with overlapping areas to construct a stereo pair model for the feature points in the common area. Assume an auxiliary image space coordinate system based on the position relationships of the photos. The stereo pair model is located in this auxiliary image space coordinate system. Then, based on the grating position relationship between the high-precision camera and the 3D laser scanner, convert the stereo pair model in the auxiliary image space coordinate system to the point cloud space coordinate system where the 3D laser scanner is located.

[0042] S3. Perform two-way constraint matching on the feature point sets in the point cloud space coordinate system obtained by scanning with two adjacent grating emitters in the 3D laser scanner respectively and the feature point set in the point cloud space coordinate system solved by the stereo pair model, and select the first optimal matching feature point set and the second optimal matching feature point set.

[0043] S4. Perform two-way constraint matching on the first optimal matching feature point set and the second optimal matching feature point set again, select the final optimal matching feature point set, and perform point cloud data matching and alignment based on the final optimal matching feature point set.

[0044] See Figure 2 - Figure 3 , in this embodiment, in S2, during the process of obtaining point cloud by the 3D laser scanner, use the high-precision camera to take photos with a specific overlap rate. Take the same feature point set P1 by taking multiple images with overlapping areas, and project the feature point set P1 onto the first photo 11 or the second photo 12 to obtain the feature point set projection p 1 2 or the feature point set projection p 2 3. The first image space coordinate system 4 of the first photo 11 is different from the second image space coordinate system 5 of the second photo 12. Assume the second image space coordinate system 5 as the auxiliary image space coordinate system, and based on the position relationship of the camera centers of the two photos and the length of the photographic baseline B8, uniformly convert the first image space coordinate system 4 and the second image space coordinate system 5 to the auxiliary image space coordinate system.

[0045] See Figure 2 - Figure 3 , in this embodiment, both the first photo 11 and the second photo 12 are photos taken by the high-precision camera, and both the first photo 11 and the second photo 12 are located between the first grating emitter 9 and the second grating emitter 10. The first grating emitter 9 is arranged on the left side, and the second grating emitter 10 is arranged on the right side. When constructing the stereo pair model, the first photo 11 taken by the high-precision camera on the right side of the first grating emitter 9 and the second photo 12 taken by the high-precision camera on the left side of the second grating emitter 10.

[0046] See Figure 2 - Figure 3, in this embodiment, in S2, based on the actual relative position relationship between the first grating emitter 9 and the high-precision camera on the right side of the first grating emitter 9, and the actual relative position relationship between the second grating emitter 10 and the high-precision camera on the left side of the second grating emitter 10, the conversion parameters between the image space auxiliary coordinate system and the point cloud space coordinate system are determined, and the point cloud space coordinates of the feature point set P1 under the three-dimensional laser scanner are calculated. The above actual relative position relationship includes the geometric relationship of optical imaging and the spatial relative relationship among the high-precision camera, the three-dimensional laser scanner, and the grating emitter.

[0047] See Figure 2 - Figure 3 , in this embodiment, the projection p 1 2 of the feature point set and the projection p 2 3 of the feature point set in the image space auxiliary coordinate system are respectively {(x 1 , y 1 , z 1 ) | x 1 ∈ the first image space coordinate system 4, y 1 ∈ the first image space coordinate system 4, z 1 ∈ the first image space coordinate system 4}, {(x 2 , y 2 , z 2 ) | x 2 ∈ the second image space coordinate system 5, y 2 ∈ the second image space coordinate system 5, z 2 ∈ the second image space coordinate system 5}. The first image space coordinate system 4 and the second image space coordinate system 5 take the first photography center S 1 6 and the second photography center S 2 7 as the coordinate origins respectively. The x-axis and y-axis of the first image space coordinate system 4 and the second image space coordinate system 5 are respectively parallel to the x-axis and y-axis of the image plane coordinate system. The image plane coordinate system is the position relationship in the photographed image plane where the image points are located, that is, the two-dimensional plane coordinate system where the image is located. The z-axes of the first image space coordinate system 4 and the second image space coordinate system 5 coincide with the principal optical axis of the high-precision camera;

[0048] Based on the relative position relationship between the first photography center S 1 6 and the second photography center S 2 7 and the first grating emitter 9 and the second grating emitter 10, the conversion parameters between the image space auxiliary coordinate system and the point cloud space coordinate system under the three-dimensional laser scanner are calculated. Through the conversion parameters, the coordinate of the feature point set in the image space auxiliary coordinate system {(x 1 , y 1 , z 1 ) | x 1∈Image space auxiliary coordinate system, y 1 ∈Image space auxiliary coordinate system, z 1 ∈Image space auxiliary coordinate system} and {(x 2 , y 2 , z 2 )|x 2 ∈Image space auxiliary coordinate system, y 2 ∈Image space auxiliary coordinate system, z 2 ∈Image space auxiliary coordinate system} are respectively converted into point cloud space coordinates {(X K1 , Y K1 , Z K1 )|X K1 ∈The point cloud space coordinate system where the first grating emitter 9 is located, Y K1 ∈The point cloud space coordinate system where the first grating emitter 9 is located, Z K1 ∈The point cloud space coordinate system where the first grating emitter 9 is located} and {(X K2 , Y K2 , Z K2 )|X K2 ∈The point cloud space coordinate system where the second grating emitter 10 is located, Y K2 ∈The point cloud space coordinate system where the second grating emitter 10 is located, Z K2 ∈The point cloud space coordinate system where the second grating emitter 10 is located}.

[0049] Under normal circumstances, the X-axis, Y-axis, and Z-axis of the image space auxiliary coordinate system are respectively parallel to the corresponding axis systems of the image space coordinate system of the first photo in the flight line, and the coordinate origin is the first projection center S 1 6 and the second projection center S 2 7. This method uses the image space coordinate system of the first photo or the left photo in the flight line to replace the image space auxiliary coordinate system.

[0050] See Figure 2 - Figure 4 , in this embodiment, in S3, the three-dimensional laser scanner with a high-precision camera has multiple grating emitters. Each grating emitter can obtain a set of feature points in the point cloud space coordinate system under the grating emitter through scanning, and this set of feature points is the point cloud data obtained by the grating emitter;

[0051] The set of feature points in the point cloud space coordinate system scanned by the first grating emitter 9 among adjacent grating emitters is subjected to two-way constraint matching with the set of feature points in the point cloud space coordinate system solved by the stereo pair model to select the first optimal matching set of feature points. The set of feature points in the point cloud space coordinate system scanned by the second grating emitter 10 among adjacent grating emitters is subjected to two-way constraint matching with the set of feature points in the point cloud space coordinate system solved by the stereo pair model to select the second optimal matching set of feature points;

[0052] The spatial coordinates of the point cloud obtained by scanning the first grating emitter 9 and the second grating emitter 10 are converted from the spatial coordinates of the point cloud obtained by scanning the first grating emitter 9 to the spatial coordinates of the point cloud obtained by scanning the second grating emitter 10 through the positional relationship between two adjacent grating emitters, namely the vertical angle α13, the rotation angle ω14, and the distance R15. The coordinate conversion formula is as follows:

[0053] X K2 = R × cosα × sinω

[0054] Y K2 = R × cosα × sinω

[0055] Z K2 = R × sinω.

[0056] See Figure 1 - Figure 3 , in this embodiment, in S4, two point cloud data obtained by scanning two adjacent grating emitters have common feature points at adjacent edges. The two point cloud data with common feature points are the point cloud data that need to be stitched and matched. The two point cloud data that need to be stitched and matched are respectively compared with the stereo pair model for the two-way constrained coordinate difference. By restricting the coordinate difference, the first optimal matching feature point set and the second optimal matching feature point set are matched bidirectionally;

[0057] The first optimal matching feature set and the second optimal matching feature set are respectively located in the point cloud space coordinate system where the first grating emitter 9 and the second grating emitter 10 are located. After the point cloud space coordinate system where the first grating emitter 9 is located is rotated, translated, and scaled in sequence through the positional relationship between two adjacent grating emitters, it is converted to the point cloud space coordinate system where the second grating emitter 10 is located. After unifying the coordinate system, the first optimal matching feature point set and the second optimal matching feature point set are compared again for the two-way constrained coordinate difference, and finally the final optimal matching feature point set for point cloud matching is obtained.

[0058] When comparing the two-way constrained coordinate difference between the spatial coordinate feature points of the point cloud calculated from the point cloud data and the spatial coordinate feature points of the point cloud calculated from the stereo pair model, based on the feature points on the spatial coordinates of the point cloud calculated from the stereo pair model, search one by one in the feature set of the point cloud data obtained by the first grating emitter (9) on the left for the feature points whose coordinate differences are less than the tolerance requirements for matching, and obtain the feature point matching set of the left point cloud data relative to the stereo pair. Similarly, obtain the feature point matching set of the stereo pair relative to the left point cloud data. If some feature points can satisfy both of these two matching sets, they are the optimal matching feature point set P 左 = {feature point 1, feature point 2, feature point 3...}.

[0059] Similarly, based on the feature points in the spatial coordinates of the point cloud solved by the stereo pair model, search one by one for the feature points with a coordinate difference less than the tolerance requirement in the feature set of the point cloud data obtained by the second grating emitter (10) on the right side, to obtain the feature point matching set of the right-side point cloud data relative to the stereo pair, and similarly obtain the feature point matching set of the stereo pair relative to the right-side point cloud data. If some feature points satisfy both of these two matching sets at the same time, they are the optimal matching feature point set P 右 ={feature point 1, feature point 2, feature point 3...}.

[0060] Again, perform a two-way constrained coordinate difference comparison on these two sets of P 左 and P 右 Finally, the obtained optimal matching feature point set P = {feature point 1, feature point 2, feature point 3...}, and use this set P for point cloud matching alignment.

[0061] In this embodiment, the feature point set P1 at least includes high-curvature points, edge points, and corner points. These above-mentioned points have important feature information, can be used to describe the local shape and structure of the point cloud, and are saved and transmitted as the representative features of the point cloud data.

[0062] In this embodiment, the first grating emitter 9 and the second grating emitter 10 project structured light stripes containing encoded information onto the feature point set P1. When this emitted light beam irradiates the surface of the object to be scanned, it will be modulated into deformed light stripes, and these modulated gratings contain the three-dimensional information of the surface of the object to be scanned.

[0063] The embodiment of the present invention also provides a computer-readable storage medium, storing a computer program, when the computer program is executed by a processor, enabling the processor to execute the steps of the point cloud registration method of the stereo pair as described above.

[0064] See Figure 5 , the embodiment of the present invention also provides a computer device, including a memory and a processor, the memory stores a computer program, when the computer program is executed by the processor, enabling the processor to execute the steps of the point cloud registration method of the stereo pair as described above.

[0065] The embodiment of the present invention also provides a computer program product containing instructions, when it runs on a computer, enabling the computer to execute the steps of the point cloud registration method of the stereo pair as described above.

[0066] It is understandable that the systems, devices, and storage media provided in the embodiments of the present invention correspond to the methods provided in the embodiments of the present invention. For the explanations, examples, and beneficial effects of related content, reference can be made to the corresponding parts in the above-mentioned point cloud registration method for stereo image pairs.

[0067] It should be noted that those of ordinary skill in the art can understand that all or part of the steps implemented in the embodiments of the present invention can be implemented in whole or in part through software, hardware, firmware, or any combination thereof. When implemented using hardware, it can be implemented in whole or in part in the form of purchasing standard components or modified components. When implemented using software, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of the present application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (such as coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (such as infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that the computer can access or a data storage device such as a server or data center that includes one or more integrated available media. The available medium can be a magnetic medium (such as a floppy disk, hard disk, magnetic tape), an optical medium (such as a DVD), or a semiconductor medium (such as a solid-state drive (SSD)).

[0068] In summary, the present application provides a new method for point cloud matching of stereo image pairs that can integrate different point cloud data into a globally consistent point cloud model in both homologous point cloud matching and cross-source point cloud matching. It mainly constructs a stereo image pair model through multiple images with overlapping regions, determines the coordinates of the feature points of the stereo image pair model in the point cloud space coordinate system, and performs corresponding searches and transformation estimations with the actual scanned point cloud coordinates to select the best edge feature points, and then performs least squares iterative matching again to determine the optimal edge feature points that finally meet the coordinate difference requirements, and performs high-precision point cloud matching based on these feature points.

[0069] It should be understood that the examples and embodiments described herein are only for illustration and are not used to limit the present invention. Those skilled in the art can make various modifications or changes according to it. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. A point cloud registration method for a stereo image pair, characterized in that: The following steps are involved: S1. Using a 3D laser scanner with a high-precision camera to obtain a point cloud, take multiple photos with overlapping areas; S2. Use multiple photos with overlapping areas to construct a stereo image pair model for feature points in a common area, assume an image space auxiliary coordinate system based on the positional relationship of each photo, and the stereo image pair model is located in this image space auxiliary coordinate system. Then, through the grating position relationship between the high-precision camera and the 3D laser scanner, the stereo image pair model in the image space auxiliary coordinate system is converted to the point cloud space coordinate system where the 3D laser scanner is located; S3, performing bidirectional constraint matching on the feature point set in the point cloud space coordinate system obtained by scanning two adjacent grating transmitters in the three-dimensional laser scanner and the feature point set in the point cloud space coordinate system solved by the stereo image pair model, and selecting a first optimal matching feature point set and a second optimal matching feature point set; S4. Perform bidirectional constraint matching again on the first optimal matching feature point set and the second optimal matching feature point set to select a final optimal matching feature point set, and use the final optimal matching feature point set to perform point cloud data matching and alignment.

2. The point cloud registration method of a stereo image pair as claimed in claim 1, characterized in that: In S2, during the point cloud acquisition process, the three-dimensional laser scanner uses a high-precision camera to take photos with a specific overlap rate, and uses multiple images with overlapping areas to take the same feature point set P (1), and projects the feature point set P (1) onto the first photo (11) or the second photo (12) to obtain the feature point set projection p1 (2) or the feature point set projection p2 (3). The first image space coordinate system (4) of the first photo (11) is different from the second image space coordinate system (5) of the second photo (12). The second image space coordinate system (5) is assumed to be an auxiliary image space coordinate system. The first image space coordinate system (4) and the second image space coordinate system (5) are uniformly converted to the auxiliary image space coordinate system using the photographic center image pair position relationship and the photographic baseline length B (8) of the two photos.

3. The point cloud registration method for a stereoscopic image pair as claimed in claim 2, characterized in that: The first photo (11) and the second photo (12) are both photos taken by a high-precision camera, and the first photo (11) and the second photo (12) are both located between a first grating transmitter (9) and a second grating transmitter (10), the first grating transmitter (9) being located on the left side, and the second grating transmitter (10) being located on the right side. When constructing a stereo pair model, the first photo (11) is taken by the high-precision camera located on the right side of the first grating transmitter (9), and the second photo (12) is taken by the high-precision camera located on the left side of the second grating transmitter (10).

4. The point cloud registration method for a stereoscopic image pair as claimed in claim 3, characterized in that: In S2, the conversion parameters between the image space auxiliary coordinate system and the point cloud space coordinate system are determined based on the actual relative position relationship between the first grating transmitter (9) and the high-precision camera located on the right side of the first grating transmitter (9), and the actual relative position relationship between the second grating transmitter (10) and the high-precision camera located on the left side of the second grating transmitter (10), and the point cloud space coordinates of the feature point set P (1) under the three-dimensional laser scanner are calculated.

5. The point cloud registration method of a stereoscopic image pair as claimed in claim 4, characterized in that: The coordinates of the feature point set projection p1(2) and the feature point set projection p2(3) in the auxiliary image space coordinate system are respectively {(x1, y1, z1)|x1∈first image space coordinate system (4), y1∈first image space coordinate system (4), z1∈first image space coordinate system (4)}, {(x2, y2, z2)|x2∈second image space coordinate system (5), y2∈second image space coordinate system (5), z2∈second image space coordinate system (5)}, wherein the first image space coordinate system (4) and the second image space coordinate system (5) are respectively The image space coordinate system (5) takes the first photographic center S1 (6) and the second photographic center S2 (7) as the coordinate origins respectively, the x-axis and y-axis of the first image space coordinate system (4) and the second image space coordinate system (5) are respectively parallel to the x-axis and y-axis of the image plane coordinate system, the image plane coordinate system is the positional relationship of the image points in the plane of the photographed film, that is, the z-axis of the first image space coordinate system (4) and the second image space coordinate system (5) are both coincident with the main optical axis of the high-precision camera; Based on the relative positional relationship between the first photographic center S1 (6) and the second photographic center S2 (7) and the first grating transmitter (9) and the second grating transmitter (10), the conversion parameters between the image space auxiliary coordinate system and the point cloud space coordinate system under the three-dimensional laser scanner are calculated, and the feature point set coordinates {(x1, y1, z1)|x1∈image space auxiliary coordinate system, y1∈image space auxiliary coordinate system, z1∈image space auxiliary coordinate system} and {(x2, y2, z2)|x2∈image space auxiliary coordinate system, y2∈image space auxiliary coordinate system, z2∈image space auxiliary coordinate system} in the image space auxiliary coordinate system are converted into point cloud space coordinates {(x1, y1, z1)|x1∈image space auxiliary coordinate system, y1∈image space auxiliary coordinate system, z1∈image space auxiliary coordinate system} in the image space auxiliary coordinate system respectively through the conversion parameters. K1 , Y K1 , Z K1 )|X K1 ∈The point cloud space coordinate system where the first grating transmitter (9) is located, Y K1 ∈The point cloud space coordinate system where the first grating transmitter (9) is located, Z K1 ∈ the point cloud space coordinate system where the first grating transmitter (9) is located} and {(X K2 , Y K2 , Z K2 )|X K2 ∈The point cloud space coordinate system where the second grating transmitter (10) is located, Y K2 ∈The point cloud space coordinate system where the second grating transmitter (10) is located, Z K2 ∈The point cloud space coordinate system where the second grating transmitter (10) is located}.

6. The point cloud registration method of a stereoscopic image pair as claimed in claim 5, characterized in that: In S3, the three-dimensional laser scanner with a high-precision camera has a plurality of grating transmitters, and each grating transmitter can obtain a feature point set in a point cloud space coordinate system under the grating transmitter by scanning, and the feature point set is the point cloud data obtained by the grating transmitter; A feature point set in a point cloud space coordinate system scanned by the first grating transmitter (9) among the adjacent grating transmitters is matched with a feature point set in a point cloud space coordinate system solved by a stereo image pair model by two-way constraint matching, and the first optimal matching feature point set is selected; a feature point set in a point cloud space coordinate system scanned by the second grating transmitter (10) among the adjacent grating transmitters is matched with a feature point set in a point cloud space coordinate system solved by a stereo image pair model by two-way constraint matching, and the second optimal matching feature point set is selected; The point cloud space coordinates obtained by scanning the first grating transmitter (9) and the second grating transmitter (10) are converted into the point cloud space coordinates obtained by scanning the second grating transmitter (10) through the positional relationship between two adjacent grating transmitters, that is, the vertical angle α (13), the rotation angle ω (14) and the distance R (15). The coordinate conversion formula is as follows: X K2 =R×cosα×sinω AND K2 =R×cosα×sinω WITH K2 =R×sinω。 7. The point cloud registration method of a stereoscopic image pair as claimed in claim 6, characterized in that: In S4, the two point cloud data scanned by two adjacent grating transmitters have common feature points at adjacent edges, and the two point cloud data having the common feature points are the point cloud data that need to be spliced ​​and matched. The two point cloud data that need to be spliced ​​and matched are compared with the stereo image pair model by performing bidirectional constrained coordinate difference, and the first optimal matching feature point set and the second optimal matching feature point set are obtained by performing bidirectional constrained matching by limiting the coordinate difference; The first optimal matching feature set and the second optimal matching feature set are respectively located in the point cloud space coordinate system where the first grating transmitter (9) and the second grating transmitter (10) are located. The point cloud space coordinate system where the first grating transmitter (9) is located is rotated, translated and scaled in sequence according to the positional relationship between the two adjacent grating transmitters, and then converted to the point cloud space coordinate system where the second grating transmitter (10) is located. After unifying the coordinate system, the first optimal matching feature point set and the second optimal matching feature point set are compared again in terms of bidirectional constrained coordinate differences, and finally the final optimal matching feature point set for point cloud matching is obtained.

8. The point cloud registration method for a stereoscopic image pair as claimed in claim 2, characterized in that: The feature point set P(1) includes at least high curvature points, edge points and corner points.

9. The point cloud registration method of a stereo image pair as claimed in claim 3, characterized in that: The first grating transmitter (9) and the second grating transmitter (10) project structured light stripes containing coded information toward the feature point set P (1).

10. A computer device, characterized in that: The method comprises a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor executes the steps of the point cloud registration method for a stereoscopic image pair as claimed in any one of claims 1 to 9.

Images