Three-dimensional point cloud registration method, system, storage medium and terminal for ship curved plate

Abstract

The present invention discloses a three-dimensional point cloud registration method, system, storage medium and terminal for a ship curved plate. The method comprises: determining a central module from a plurality of modules that can enclose an area to be registered for the ship curved plate; arranging a plurality of marker points in each overlapping field of view; performing point cloud data acquisition on the central module and each remaining module respectively to obtain a set of target point cloud data and a plurality of groups of original point cloud data; performing point cloud coarse registration on the target point cloud data and the original point cloud data to achieve coarse registration of the overlapping area of ​​the ship outer plate point cloud; preprocessing the target point cloud data and the plurality of groups of original point cloud data respectively; extracting a plurality of overlapping area point cloud data corresponding to each overlapping field of view from the target point cloud data and the plurality of groups of original point cloud data respectively; processing two overlapping area point cloud data in the same overlapping field of view to replace the overlapping area point cloud data; and performing fine registration, wherein the central module and each remaining module have at least one overlapping field of view.

Description

Technical Field

[0001] The present invention relates to the technical field of point cloud registration, and in particular to a three-dimensional point cloud registration method, system, storage medium and terminal for a curved plate of a ship. Background Art

[0002] Point cloud registration technology is widely used in reverse engineering, construction engineering, machine vision and other fields. The registration of three-dimensional point clouds is essentially to minimize the average distance between corresponding points in the overlapping area of ​​the two data sets by solving the optimal rotation and translation matrix in an appropriate metric space. Traditional registration algorithms mainly include algorithms based on local feature descriptors (such as the fast point feature histogram algorithm (FPFH), 3D shape context feature algorithm (3Dsc), etc.), algorithms based on probability distribution (such as the normal distribution transform algorithm (NDT)), and point-to-point distance-based registration algorithms such as the iterative closest point search algorithm (ICP). Among them, algorithms based on local feature descriptors and algorithms based on probability distribution are usually used for coarse registration; registration algorithms based on point-to-point distance are usually used for fine registration.

[0003] Because fine registration algorithms have high algorithmic overhead and the initial registration values ​​of two completely unknown coordinate systems significantly affect the algorithm's computational efficiency and accuracy, current mainstream registration solutions typically use a cascade of coarse and fine registration, and perform appropriate data preprocessing before registration. This includes filtering and sampling the dataset to remove noise to improve algorithm accuracy, and downsampling to reduce the data volume to improve computational efficiency.

[0004] Unlike traditional applications such as complex workpieces and autonomous driving, ship plating is mostly smooth, curved plates with minimal curvature variation. To fully utilize the field of view of the 3D laser scanning hardware and cover a larger measurement volume, measurement systems typically minimize the overlapping field of view between imaging devices. Furthermore, within the overlapping area, the surface of the ship plating is mostly smooth, with minimal curvature variation, resulting in fewer texture features. Traditional coarse registration algorithms alone produce poor registration results. Furthermore, due to the uneven distribution of point cloud data in the overlapping area, directly using the data from the overlapping area for a fine registration algorithm based on point-to-point distances cannot meet accuracy requirements. Summary of the Invention

[0005] The purpose of the present invention is to provide a three-dimensional point cloud registration method, system, storage medium and terminal for a ship curved plate, so as to solve the problem in the prior art that when the coverage measurement space is large, there are few texture features in the data overlap area, which leads to low accuracy of the fine registration algorithm.

[0006] The above-mentioned purpose of the present invention can be achieved by adopting the following technical solutions:

[0007] The present invention provides a three-dimensional point cloud registration method for a ship curved plate, comprising:

[0008] Determine a central module from a plurality of modules that can enclose the area to be registered of the ship curved plate;

[0009] Arrange multiple markers in each overlapping field of view according to preset rules;

[0010] Performing point cloud data collection on the central module and each of the remaining modules to obtain a set of target point cloud data and multiple sets of original point cloud data;

[0011] Based on the attributes of the corresponding multiple landmark points, the target point cloud data and the original point cloud data are coarsely registered to achieve coarse registration of the overlapping area of ​​the ship outer plate point cloud;

[0012] Preprocessing the target point cloud data and the multiple groups of original point cloud data after rough registration respectively;

[0013] Based on the attributes of each of the marker points, extracting a plurality of overlapping area point cloud data corresponding to each of the overlapping fields of view from the target point cloud data and the plurality of groups of original point cloud data respectively;

[0014] Performing sparse fitting processing on the two overlapping area point cloud data in the same overlapping field of view to replace the overlapping area point cloud data;

[0015] Precise registration is performed based on the target point cloud data, the multiple groups of original point cloud data, and the multiple overlapping area point cloud data.

[0016] The central module and each of the remaining modules have at least one overlapping field of view.

[0017] Preferably, the step of performing point cloud coarse registration on the target point cloud data and the original point cloud data based on the corresponding attributes of the plurality of landmark points to achieve coarse registration of overlapping areas of the ship outer plate point cloud comprises:

[0018] Obtaining the central three-dimensional coordinates of each of the marker points using a marker point recognition algorithm;

[0019] Based on the fact that the attribute relationship between the multiple marker points in the central module and the multiple marker points in one of the remaining modules meets the preset requirements, the target point cloud data and the corresponding original point cloud data are roughly aligned to achieve rough alignment of the overlapping area of ​​the ship outer plate point cloud.

[0020] Preferably, the preprocessing includes denoising, filtering and completing the data.

[0021] Preferably, the filtering function used in the filtering in the preprocessing is:

[0022] p(i+1)=k×p(i)+(1-k)×p(i-1);

[0023] Wherein, k is the filter coefficient, and i is an integer not less than 1.

[0024] Preferably, the filter coefficient is not less than 0.1 and not greater than 0.4.

[0025] Preferably, when fitting the point cloud data of the two overlapping areas in the same overlapping field of view, a quadratic surface fitting method is adopted; and / or

[0026] When performing sparse processing on the point cloud data of the two overlapping areas in the same overlapping field of view, an equal interval method is adopted.

[0027] Preferably, the precise registration is performed using an iterative closest point search algorithm.

[0028] The present invention also provides a system, which includes: a module for executing any of the aforementioned ship curved plate three-dimensional point cloud registration methods.

[0029] The present invention also provides a storage medium having a computer program stored thereon, wherein the storage medium is a computer-readable storage medium, and when the program is executed, any of the aforementioned ship curved plate three-dimensional point cloud registration methods is implemented.

[0030] The present invention also provides a terminal comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, any one of the aforementioned ship curved plate three-dimensional point cloud registration methods is implemented.

[0031] The present invention has at least the following characteristics and advantages:

[0032] 1. The present invention solves the problem of splicing cloud data collected in blocks of ship shell plates by coupling a method based on coarse registration based on landmark points and fast fine registration based on overlapping areas, and enables 3D equipment to obtain complete shell plate 3D point cloud data through one data collection, thereby improving data collection efficiency and accuracy.

[0033] 2. The data obtained by the present invention through sparse fitting is more regular and the data volume is smaller, so that the correspondence between the original point cloud and the target point cloud data is better, thereby improving the calculation efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0035] Figure 1 This is a flowchart of the three-dimensional point cloud registration method for a curved plate of a ship according to the present invention;

[0036] Figure 2 This is a flowchart of the three-dimensional point cloud registration method for a curved plate of a ship according to the present invention;

[0037] Figure 3 is a flow chart of the coarse registration method of the present invention;

[0038] Figure 4 is a flow chart of the precise registration method of the present invention;

[0039] Figure 5 It is a structural block diagram of the terminal of the present invention.

[0040] Reference numerals and descriptions:

[0041] 10000, terminal; 11000, memory; 11100, computer program; 12000, processor. DETAILED DESCRIPTION

[0042] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0043] Implementation Method 1

[0044] The present invention provides a method for 3D point cloud registration of ship curved plates. Figures 1 to 4 , including the following steps:

[0045] S1. Determine a central module from multiple modules that can enclose the area to be registered for the ship's curved plate;

[0046] S2. Arrange multiple markers in each overlapping field of view according to preset rules;

[0047] S3, collecting point cloud data from the central module and each of the remaining modules to obtain a set of target point cloud data and multiple sets of original point cloud data;

[0048] S4. Based on the attributes of the corresponding multiple landmark points, the target point cloud data and the original point cloud data are coarsely registered to achieve coarse registration of the overlapping area of ​​the ship shell point cloud;

[0049] S5, pre-processing the target point cloud data after coarse registration and the multiple groups of original point cloud data respectively;

[0050] S6. Based on the attributes of each marker point, extracting a plurality of overlapping area point cloud data corresponding to each overlapping field of view from the target point cloud data and the plurality of sets of original point cloud data;

[0051] S7, performing sparse fitting processing on the point cloud data of two overlapping areas in the same overlapping field of view to replace the point cloud data of the overlapping area;

[0052] S8. Perform precise registration based on the target point cloud data, multiple sets of original point cloud data, and multiple overlapping area point cloud data.

[0053] The central module and each of the remaining modules have at least one overlapping field of view.

[0054] In some embodiments, see Figure 2 S4, based on the attributes of the corresponding multiple landmark points, performing point cloud coarse registration of the target point cloud data with the original point cloud data to achieve coarse registration of the overlapping area of ​​the ship outer plate point cloud, including:

[0055] S41, using a landmark recognition algorithm to obtain the central three-dimensional coordinates of each landmark;

[0056] S42. Based on the fact that the attribute relationship between the multiple landmark points in the central module and the multiple landmark points in one of the remaining modules meets the preset requirements, the target point cloud data and the corresponding original point cloud data are coarsely aligned to achieve coarse alignment of the overlapping area of ​​the ship outer plate point cloud.

[0057] In some embodiments, satisfying the preset requirements based on the attribute relationship between multiple marker points in the central module and multiple marker points in a remaining module means that the multiple marker points in the central module and the multiple marker points in a remaining module are points with the same name, that is, the numbers of the multiple marker points in the central module are the same as the numbers of the multiple marker points in a remaining module.

[0058] In some embodiments, preprocessing includes denoising, filtering, and completing the data.

[0059] Furthermore, the filtering function used in the pre-processing is:

[0060] p(i+1)=k×p(i)+(1-k)×p(i-1);

[0061] Wherein, k is a filter coefficient, and i is an integer not less than 1. Furthermore, the filter coefficient is not less than 0.1 and not greater than 0.4.

[0062] In some embodiments, a quadratic surface fitting method is used when fitting the point cloud data of two overlapping areas in the same overlapping field of view; in other embodiments, an equal interval method is used when sparsely processing the point cloud data of two overlapping areas in the same overlapping field of view.

[0063] In some embodiments, fine registration is performed using an iterative closest point search algorithm.

[0064] The present invention has at least the following characteristics and advantages:

[0065] 1. The present invention solves the problem of splicing cloud data collected in blocks of ship shell plates by coupling a method based on coarse registration based on landmark points and fast fine registration based on overlapping areas, and enables 3D equipment to obtain complete shell plate 3D point cloud data through one data collection, thereby improving data collection efficiency and accuracy.

[0066] 2. The data obtained by the present invention through sparse fitting is more regular and the data volume is smaller, so that the correspondence between the original point cloud and the target point cloud data is better, thereby improving the calculation efficiency.

[0067] The present invention is further studied below through a specific embodiment. Figures 1 to 4 :

[0068] like Figure 3 As shown, the coarse registration of the present invention includes the following steps:

[0069] Step 1: Place markers in the overlapping field of view between the central module and each remaining module. The placement mainly follows the following principles: the markers cover as much overlapping area as possible; ensure that the distances between each pair of markers are different and the shapes formed are irregular;

[0070] Step 2: Collect point cloud data. Since there are overlapping fields of view between modules, the collection of point cloud data is carried out in different time periods to obtain multiple sets of point cloud data.

[0071] Step 3: Use the landmark recognition algorithm to obtain the three-dimensional coordinates of the center point of the landmark;

[0072] Step 4: Rough registration of point clouds based on homonymous points.

[0073] Assuming that the center coordinates of the landmarks in the n overlapping areas are obtained, n sets of homonymous points can be constructed. Then, the n sets of spatial homonymous points are used to calculate the coarse registration rotation matrix R1 and translation matrix T1 of the original point cloud to the target point cloud;

[0074] like Figure 4As shown, the precise registration of the present invention includes the following steps:

[0075] Step 1: Preprocess the data by denoising and filtering the coarsely registered original point cloud and target point cloud data.

[0076] Specifically, the denoising process is as follows: calculate the mean of the non-zero depth values ​​in the eight neighborhoods of each point in the depth map, then calculate the error between the depth value of the point and the mean of the eight neighborhoods. If the error is greater than 5mm, the point is considered to be noise and is removed; traverse each row of data and filter the point cloud data row by row. If the point cloud is m rows × n columns of (x, y, z) data, for a row of data there is a data set [p(1), p(2), …p(i), …, p(n)], where p(i) = {xi, yi, zi}. Use the filtering function to process each row of data to obtain new point cloud data. The filtering function is as follows: p(i+1) = k×p(i) + (1-k)×p(i-1), where k is the filtering coefficient, and the value range is [0.1, 0.4].

[0077] Step 2: In the depth maps of the original point cloud and the target point cloud, the point cloud data is cut out according to the three-dimensional coordinates of the center points of the acquired landmark points to obtain the original point cloud and target point cloud in the overlapping area;

[0078] Step 3: Perform quadratic surface fitting on the overlapping area of ​​the original point cloud and the target point cloud to obtain the equation of the fitted surface. Then, based on the obtained surface equation, recalculate the point cloud data in the marker area at equal intervals to obtain sparse point cloud data.

[0079] Step 4: Call the iterative closest point search algorithm to perform precise registration of the sparse fitting point cloud data to obtain the precise registration rotation matrix R2 and translation matrix T2.

[0080] The present invention is a method that couples coarse registration based on marker points with fast fine registration based on overlapping areas. Coarse registration of overlapping areas of a ship's outer plate point cloud is achieved through marker point deployment, point cloud data acquisition, marker point recognition, and point cloud coarse registration based on points of the same name. The coarse registration point cloud data is pre-processed by denoising and filtering, the overlapping area point cloud is intercepted according to the marker point center coordinates, and the intercepted point cloud data is parameterized and fitted, and fine registration is performed using sparsely fitted data. The present invention can solve the problem of splicing cloud data collected in blocks of ship outer plate, enabling a three-dimensional device to obtain complete three-dimensional point cloud data of the outer plate through a single data acquisition, thereby improving registration efficiency and accuracy.

[0081] Implementation Method 2

[0082] The embodiment of the present invention further provides a system, which includes a module for executing the steps of the method in any one of the embodiments in Implementation Method 1. Those skilled in the art will appreciate that the system provided by the present invention has the same beneficial effects as the embodiments in Implementation Method 1, and will not be described in detail here.

[0083] Implementation Method 3

[0084] The embodiment of the present invention further provides a storage medium having a computer program 11100 stored thereon. The storage medium is a computer-readable storage medium, and when the program is executed by the processor 12000, the steps of the method of any embodiment in Implementation Method 1 are implemented. The computer-readable storage medium may include, but is not limited to, any type of disk, including a floppy disk, an optical disk, a DVD, a CD-ROM, a microdrive, a magneto-optical disk, a ROM, a RAM, an EPROM, an EEPROM, a DRAM, a VRAM, a flash memory device, a magnetic card or an optical card, a nanosystem (including a molecular memory IC), or any type of medium or device suitable for storing instructions and / or data. The specific execution process can be found in the detailed description of the method embodiment in Implementation Method 1, and will not be repeated here.

[0085] Those skilled in the art should understand that the storage medium provided by the present invention has the same beneficial effects as the embodiment in Implementation Method 1, and will not be described in detail here.

[0086] Implementation Method 4

[0087] The embodiment of the present invention further provides a terminal 10000, see Figure 5 , including a memory 11000, a processor 12000, and a computer program 11100 stored in the memory 11000 and executable on the processor 12000. When the processor 12000 executes the computer program 11100, the method of any one of the embodiments in Implementation Method 1 is implemented. The specific implementation process can be found in the detailed description of the above-mentioned method embodiments and will not be described in detail here.

[0088] In the embodiment of the present invention, the processor 12000 is the control center of the computer system and can be the processor of a physical machine or a virtual machine. In the embodiment of the present invention, the memory 11000 stores at least one instruction, which is loaded and executed by the processor 12000 to implement the methods in the above embodiments.

[0089] In one embodiment of the present invention, the processor 12000 may include one or more processing cores, such as a quad-core processor or an octal-core processor. The processor 12000 may be implemented using at least one of the following hardware forms: a DSP (Digital Signal Processing), an FPGA (Field-Programmable Gate Array), and a PLA (Programmable Logic Array). The processor 12000 may also include a main processor and a coprocessor. The main processor is a processor for processing data in the awake state, also known as a CPU (Central Processing Unit); the coprocessor is a low-power processor for processing data in the standby state.

[0090] Memory 11000 may include one or more computer-readable storage media, which may be non-transitory. Memory 11000 may also include high-speed random access memory and non-volatile memory, such as one or more disk storage devices and flash memory storage devices. In some embodiments of the present invention, the non-transitory computer-readable storage medium in memory 11000 is used to store at least one instruction, which is used to be executed by processor 12000 to implement the method in the embodiment of the present invention.

[0091] Those skilled in the art should understand that the terminal 10000 provided by the present invention has the same beneficial effects as the embodiment in Implementation Method 1, and will not be described in detail here.

[0092] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any form. Although the present invention has been disclosed as above in terms of preferred embodiments, they are not intended to limit the present invention. Any technician familiar with the present profession can make some changes or modifications to equivalent embodiments using the technical contents disclosed above without departing from the scope of the technical solution of the present invention. However, any simple modifications, equivalent changes and modifications made to the above embodiments based on the technical essence of the present invention without departing from the content of the technical solution of the present invention are still within the scope of the technical solution of the present invention.

Claims

1. A three-dimensional point cloud registration method for a ship curved plate, characterized in that: include: Determine a central module from a plurality of modules that can enclose the area to be registered of the ship curved plate; Arrange multiple markers in each overlapping field of view according to preset rules; Performing point cloud data collection on the central module and each of the remaining modules to obtain a set of target point cloud data and multiple sets of original point cloud data; Based on the attributes of the corresponding multiple landmark points, the target point cloud data and the original point cloud data are coarsely registered to achieve coarse registration of the overlapping area of ​​the ship outer plate point cloud; Preprocessing the target point cloud data and the multiple groups of original point cloud data after rough registration respectively; Based on the attributes of each of the marker points, extracting a plurality of overlapping area point cloud data corresponding to each of the overlapping fields of view from the pre-processed target point cloud data and the plurality of groups of original point cloud data; Performing sparse fitting processing on the two overlapping area point cloud data in the same overlapping field of view to replace the overlapping area point cloud data; Performing precise registration based on the target point cloud data, the multiple sets of original point cloud data, and the multiple overlapping area point cloud data; The central module and each of the remaining modules have at least one overlapping field of view.

2. The ship curved plate 3D point cloud registration method according to claim 1, characterized in that: The step of performing point cloud coarse registration on the target point cloud data and the original point cloud data based on the corresponding attributes of the plurality of landmark points to achieve coarse registration of the overlapping area of ​​the ship outer plate point cloud comprises: Obtaining the central three-dimensional coordinates of each of the marker points using a marker point recognition algorithm; Based on the fact that the attribute relationship between the multiple marker points in the central module and the multiple marker points in one of the remaining modules meets the preset requirements, the target point cloud data and the corresponding original point cloud data are roughly aligned to achieve rough alignment of the overlapping area of ​​the ship outer plate point cloud.

3. The ship curved plate 3D point cloud registration method according to claim 2, characterized in that: The preprocessing includes denoising, filtering and completing the data.

4. The ship curved plate 3D point cloud registration method according to claim 3, characterized in that: The filtering function used in the filtering in the preprocessing is: p(i+1)=k×p(i)+(1-k)×p(i-1); Wherein, k is the filter coefficient, and i is an integer not less than 1.

5. The ship curved plate 3D point cloud registration method according to claim 4, characterized in that: The filtering coefficient is not less than 0.1 and not greater than 0.

4.

6. The method for 3D point cloud registration of a curved ship plate according to any one of claims 1 to 5, characterized in that: When performing fitting processing on the point cloud data of the two overlapping areas in the same overlapping field of view, a quadratic surface fitting method is adopted; and / or When performing sparse processing on the point cloud data of the two overlapping areas in the same overlapping field of view, an equal interval method is adopted.

7. The ship curved plate 3D point cloud registration method according to claim 6, characterized in that: The precise registration is performed using an iterative closest point search algorithm.

8. A three-dimensional point cloud registration system for ship curved plates, characterized in that: include: A module for executing the three-dimensional point cloud registration method for a curved plate of a ship according to any one of claims 1 to 7.

9. A storage medium having a computer program stored thereon, characterized in that: The storage medium is a computer-readable storage medium, and when the program is executed, the three-dimensional point cloud registration method for a curved plate of a ship according to any one of claims 1 to 7 is implemented.

10. A terminal comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the three-dimensional point cloud registration method for a curved plate of a ship according to any one of claims 1 to 7 is implemented.

Images