A three-dimensional point cloud splicing reconstruction method and device based on a rotating table

Abstract

The application discloses a three-dimensional point cloud splicing reconstruction method and device based on a rotating table, and relates to the technical field of point cloud splicing reconstruction.The method comprises the following steps: placing a measured object at the center of a rotating table, fixing a 3D camera at a position capable of covering the object on the rotating table, controlling the rotating table to rotate at fixed angle intervals, triggering the 3D camera to collect a local three-dimensional point cloud of the measured object at each rotating angle, obtaining a point cloud sequence and an angle sequence under a plurality of visual angles, solving the rotating shaft parameter of the rotating table under the 3D camera coordinate system through a joint iterative calibration algorithm, calculating the rigid transformation matrix of each visual angle local three-dimensional point cloud to the same world coordinate system, uniformly transforming the local three-dimensional point cloud in the point cloud sequence to the same world coordinate system, and splicing and reconstructing the three-dimensional point cloud.The application can greatly expand the application range of three-dimensional vision technology, and provides high-quality three-dimensional data basis for subsequent precision measurement, detection and reverse engineering.

Description

Technical Field

[0001] This invention relates to the field of point cloud stitching and reconstruction technology, and in particular to a three-dimensional point cloud stitching and reconstruction method and apparatus based on a rotary table. Background Technology

[0002] In the field of 3D vision measurement and reconstruction, point cloud stitching is a crucial step in achieving complete 3D shape reconstruction of an object. Traditional point cloud stitching methods typically rely on extracting feature points (such as corner points and edges) from the object's surface for matching. Specifically, feature points are first extracted from point cloud data acquired from multiple perspectives. Then, algorithms such as RANSAC (Random Sample Consensus) are used to perform coarse matching between point clouds from different perspectives based on these feature points. Finally, iterative nearest point (ICP) algorithms are used for fine registration to complete the stitching.

[0003] However, this feature-point matching method has a fundamental limitation: its effectiveness and stability heavily depend on whether the target object's surface possesses sufficient, stable, and distinguishable feature information. In practical industrial applications, many objects to be measured (such as smooth ceramic parts, simple metal parts, and textureless plastic shells) lack obvious texture or geometric features. In such cases, traditional methods struggle to extract a sufficient number or quality of feature points, causing coarse matching algorithms like RANSAC to fail due to a lack of reliable data support. This manifests as a high matching error rate, a sharp decline in stitching accuracy, and even complete failure to achieve initial alignment. This not only severely impacts the starting point quality for subsequent fine registration but also significantly limits the application of 3D vision technology in a wider range of featureless or weakly feature-rich object detection scenarios.

[0004] Therefore, how to achieve a point cloud stitching method that does not rely on the object's own feature information and has higher versatility and robustness has become a core technical problem that urgently needs to be solved in this field. Summary of the Invention

[0005] This invention provides a method for three-dimensional point cloud stitching and reconstruction based on a rotary table, comprising: Step S1: Place the object to be tested in the center of the rotating platform and fix the 3D camera in a position that can cover the object on the rotating platform; Step S2: Control the turntable to rotate at fixed angular intervals. At each rotation angle, trigger the 3D camera to acquire local 3D point clouds of the object under test, and obtain a set of point cloud sequences and angle sequences from multiple perspectives. Step S3: Based on the point cloud sequence and angle sequence from multiple perspectives, solve the rotation axis parameters of the rotary table in the 3D camera coordinate system through a joint iterative calibration algorithm. Step S4: Based on the rotation axis parameters and angle sequence, calculate the rigid transformation matrix from the local 3D point cloud of each viewpoint to the same world coordinate system; Step S5: Based on each rigid transformation matrix, transform the local 3D point clouds in the point cloud sequence to the same world coordinate system and then stitch and reconstruct the 3D point clouds.

[0006] The above-described method for 3D point cloud stitching and reconstruction based on a rotary table, wherein the rotary table is controlled to rotate at fixed angular intervals, and at each rotation angle, a 3D camera is triggered to acquire a local 3D point cloud of the object under test, obtaining a set of point cloud sequences and angle sequences from multiple perspectives, includes the following sub-steps: Step S21: Based on the shape of the object being measured, set the fixed rotation angle interval and the number of rotations of the rotary table; Step S22: Drive the rotary table to rotate at fixed rotation angle intervals. After rotating to a target angle, trigger the 3D camera to acquire a frame of local 3D point cloud of the object under test and record the current target angle. Step S23: When the number of rotations of the turntable reaches the set number of rotations, a point cloud sequence composed of local 3D point clouds of each frame and an angle sequence composed of each target angle are obtained.

[0007] The above-described method for 3D point cloud stitching and reconstruction based on a rotary table includes the following sub-steps: Based on point cloud sequences and angle sequences from multiple viewpoints, a joint iterative calibration algorithm is used to solve for the rotation axis parameters of the rotary table in the 3D camera coordinate system. Step S31: Based on the positional relationship between the 3D camera and the rotary table and the point cloud sequence, initialize and set the rotation axis parameters of the rotary table; Step S32: Based on the rotation axis parameters of the rotary table and the angle sequence under multiple views, theoretically pre-align the point clouds in the point cloud sequence to generate a pre-aligned point cloud set; Step S33: Perform fine registration on adjacent point cloud pairs in the pre-aligned point cloud set to generate a fine registration transformation matrix set and a registration error set; Step S34: Based on the fine registration transformation matrix set, update the rotation axis parameters of the rotary table and obtain the updated rotation axis parameters; Step S35: Based on the rotation axis parameters before and after the update and the registration error set, calculate the changes in parameters and errors, and determine whether convergence has occurred. If convergence has not occurred, jump to step S32 and start the next iteration. If convergence has occurred, output the final rotation axis parameters of the rotary table.

[0008] The above-described method for 3D point cloud stitching and reconstruction based on a rotary table includes the following sub-steps for calculating the rigid transformation matrix from the local 3D point cloud of each viewpoint to the same world coordinate system, based on the rotation axis parameters and angle sequence: Step S41: Based on the rotation axis parameters and angle sequence, calculate the rigid transformation matrix of the point cloud for each viewpoint; Step S42: Verify the rigid transformation matrix of the point cloud for each viewpoint.

[0009] The method for stitching and reconstructing 3D point clouds based on a rotary table, as described above, involves transforming local 3D point clouds in a point cloud sequence to the same world coordinate system based on various rigid transformation matrices, and then stitching and reconstructing the 3D point clouds. This process includes the following sub-steps: Step S51: Based on each rigid transformation matrix, perform a unified coordinate transformation on all local 3D point clouds in the point cloud sequence, and perform preliminary fusion to generate an overall point cloud set. Step S52: Optimize the overall point cloud set and perform 3D point cloud stitching and reconstruction.

[0010] The present invention also provides a three-dimensional point cloud stitching and reconstruction device based on a rotary table, comprising: The system layout module places the object to be measured in the center of the rotating platform and fixes the 3D camera in a position that can cover the object on the rotating platform; The data acquisition module controls the turntable to rotate at fixed angular intervals. At each rotation angle, it triggers the 3D camera to acquire local 3D point clouds of the object under test, obtaining a set of point cloud sequences and angle sequences from multiple perspectives. The rotation axis parameter acquisition module, based on point cloud sequences and angle sequences from multiple perspectives, solves the rotation axis parameters of the rotary table in the 3D camera coordinate system through a joint iterative calibration algorithm. The rigid transformation matrix generation module calculates the rigid transformation matrix from the local 3D point cloud of each viewpoint to the same world coordinate system based on the rotation axis parameters and angle sequence. The point cloud stitching and reconstruction module, based on various rigid transformation matrices, transforms the local 3D point clouds in the point cloud sequence to the same world coordinate system, and then stitches and reconstructs the 3D point clouds.

[0011] As described above, a 3D point cloud stitching and reconstruction device based on a rotary table includes a data acquisition module, specifically comprising: The parameter setting submodule sets the fixed rotation angle interval and number of rotations of the turntable based on the shape of the object being measured. The point cloud acquisition submodule drives the rotary table to rotate step by step at fixed rotation angle intervals. After rotating to a target angle, it triggers the 3D camera to acquire a frame of local 3D point cloud of the object under test and records the current target angle. The point cloud and angle sequence construction submodule obtains a point cloud sequence composed of local 3D point clouds of each frame and an angle sequence composed of each target angle after the number of rotations of the turntable reaches the set number of rotations.

[0012] As described above, a 3D point cloud stitching and reconstruction device based on a rotary table includes a rotation axis parameter acquisition module, specifically comprising: The rotation axis parameter initialization submodule initializes and sets the rotation axis parameters of the rotary table based on the positional relationship between the 3D camera and the rotary table and the point cloud sequence. The pre-aligned point cloud generation submodule, based on the rotation axis parameters of the rotary table and the angle sequence under multiple views, theoretically pre-aligns the point clouds in the point cloud sequence to generate a pre-aligned point cloud. The point cloud fine registration submodule performs fine registration on adjacent point cloud pairs in the pre-aligned point cloud set, generating a fine registration transformation matrix set and a registration error set; The rotation axis parameter update submodule updates the rotation axis parameters of the rotary table based on the fine registration transformation matrix set and obtains the updated rotation axis parameters. The convergence judgment submodule calculates the changes in parameters and errors based on the rotation axis parameters and registration error set before and after the update, and determines whether convergence has occurred. If convergence has not occurred, it jumps to the pre-alignment point cloud generation submodule to start the next round of iteration. If convergence has occurred, it outputs the final rotation axis parameters of the rotary table.

[0013] As described above, a 3D point cloud stitching and reconstruction device based on a rotary table includes, in particular, a rigid transformation matrix calculation module comprising: The rigid transformation matrix calculation submodule calculates the rigid transformation matrix of the point cloud for each viewpoint based on the rotation axis parameters and angle sequence. The verification submodule verifies the rigid transformation matrix of the point cloud for each viewpoint.

[0014] As described above, a 3D point cloud stitching and reconstruction device based on a rotary table, wherein the point cloud stitching and reconstruction module specifically includes: The overall point cloud set generation submodule, based on each rigid transformation matrix, performs a unified coordinate transformation on all local 3D point clouds in the point cloud sequence, performs preliminary fusion, and generates an overall point cloud set. The 3D point cloud stitching and reconstruction submodule optimizes the overall point cloud set and performs 3D point cloud stitching and reconstruction.

[0015] The beneficial effects achieved by this invention are as follows: This invention can achieve high-precision point cloud stitching on textureless, smooth, or simple-structured objects, greatly expanding the application scope of 3D vision technology; it can also adaptively and accurately solve the spatial geometric parameters of the rotation axis using only the multi-view point cloud and rotation angle of the object being measured, without any external calibration objects, making the calibration process simple and highly repeatable; it can also achieve complete and seamless 3D reconstruction of complex curved surface objects, providing a high-quality 3D data foundation for subsequent precision measurement, detection, and reverse engineering. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0017] Figure 1 This is a flowchart of a three-dimensional point cloud stitching and reconstruction method based on a rotary table, provided in Embodiment 1 of this application; Figure 2 This is a schematic diagram of a three-dimensional point cloud stitching and reconstruction device based on a rotary table, provided in Embodiment 2 of this application. Detailed Implementation

[0018] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0019] Example 1 like Figure 1 As shown, Embodiment 1 of this application provides a method for 3D point cloud stitching and reconstruction based on a rotary table. The method includes the following steps: Step S1: Place the object to be tested in the center of the rotating platform and fix the 3D camera in a position that can cover the object on the rotating platform; Specifically, the object to be measured is placed stably at the geometric center of the rotary table. The 3D camera is fixed to the side or above the rotary table using an industrial bracket, tripod, or other device (the specific fixing position of the 3D camera is selected according to the shape of the object to be measured). The height, focal length, and shooting distance of the 3D camera are adjusted to ensure that the field of view of the 3D camera can completely cover all surfaces of the object to be measured during its 360° rotation, with no blind spots. After the 3D camera is fixed, its intrinsic parameters are calibrated using its matching calibration plate. The rotary table, 3D camera, and main control terminal are connected. The main control terminal is used to send commands to the rotary table and 3D camera, and to receive and process data.

[0020] Step S2: Control the turntable to rotate at fixed angular intervals. At each rotation angle, trigger the 3D camera to acquire local 3D point clouds of the object under test, and obtain a set of point cloud sequences and angle sequences from multiple perspectives. Furthermore, the rotary table is controlled to rotate at fixed angular intervals. At each rotation angle, the 3D camera is triggered to acquire a local 3D point cloud of the object under test, obtaining a set of point cloud sequences and angle sequences from multiple perspectives, including the following sub-steps: Step S21: Based on the shape of the object being measured, set the fixed rotation angle interval and the number of rotations of the rotary table; Specifically, based on the three-dimensional shape characteristics of the object being measured (such as size, symmetry, and surface undulation complexity), as well as the point cloud density and integrity requirements for the final reconstruction, a fixed rotation angle interval is set. Based on the 360° full-circle rotation coverage requirement and the fixed rotation angle interval, the number of rotations is calculated. The final set fixed rotation angle interval and the number of rotations are used as acquisition parameters and saved to the main control terminal.

[0021] Step S22: Drive the rotary table to rotate at fixed rotation angle intervals. After rotating to a target angle, trigger the 3D camera to acquire a frame of local 3D point cloud of the object under test and record the current target angle. Specifically, the main control unit sends initialization commands to the rotary table and 3D camera simultaneously at fixed rotation angle intervals. The rotary table returns to its mechanical zero point, and the 3D camera enters the acquisition state. A single-step rotation command is sent to the rotary table, which includes the target rotation angle (initial angle is 0°, and subsequent rotation angles are accumulated at fixed intervals). The rotation is driven to rotate around the central axis at a preset angular velocity. When the rotary table rotates to the target rotation angle, it is determined that the rotation is in place, and a rotation stop command is immediately sent, triggering a stabilization delay. After the stabilization delay ends, the main control unit sends a point cloud acquisition trigger signal to the 3D camera. After receiving the signal, the 3D camera immediately acquires a frame of local 3D point cloud of the measured object. After acquisition, the original point cloud is initially denoised, and the local 3D point cloud of the corresponding angle is sent back to the main control unit. Upon receiving the local 3D point cloud, the main control unit extracts the current target rotation angle, establishes a real-time mapping table of point cloud frames and target angles, names the acquired point cloud frames according to the acquisition sequence, and binds them one by one with the current target angle.

[0022] Step S23: When the number of rotations of the turntable reaches the set number of rotations, a point cloud sequence composed of local 3D point clouds of each frame and an angle sequence composed of each target angle are obtained. Specifically, the main control terminal counts the number of rotation acquisitions completed in real time and compares them with the set number of rotations. When the two values ​​match, an acquisition end command is sent, and standby signals are sent to the rotary table and the 3D camera respectively. The rotary table returns to the mechanical zero point, and the 3D camera exits the acquisition state. The main control terminal reorders all bound point cloud frame data and corresponding target angle data according to the actual acquisition time sequence (the step rotation sequence of the rotary table). All sorted local 3D point cloud frames are integrated into an ordered point cloud sequence. Each point cloud frame in the sequence retains its original name and complete point cloud data. The frame order corresponds completely to the step angle sequence of the rotary table. Simultaneously, all sorted target angles are integrated into an ordered angle sequence in ascending order of value. The angle sequence is stored in the form of a one-dimensional array, and the index position of each angle in the array completely matches the index position of the corresponding point cloud frame in the point cloud sequence.

[0023] Step S3: Based on the point cloud sequence and angle sequence from multiple perspectives, solve the rotation axis parameters of the rotary table in the 3D camera coordinate system through a joint iterative calibration algorithm. Furthermore, based on point cloud sequences and angle sequences from multiple viewpoints, the rotation axis parameters of the rotary table in the 3D camera coordinate system are solved using a joint iterative calibration algorithm, including the following sub-steps: Step S31: Based on the positional relationship between the 3D camera and the rotary table and the point cloud sequence, initialize and set the rotation axis parameters of the rotary table; Specifically, based on the physical positional relationship between the 3D camera and the rotating stage, the initial direction of the rotating stage's rotation axis is determined. Set as the unit vector of the Z-axis in the 3D camera coordinate system, i.e. Calculate the three-dimensional centroid of all points in the point cloud sequence as the initial rotation center. Set iteration count Simultaneously, convergence thresholds are set based on calibration requirements, including directional change thresholds. Central change threshold Registration error change threshold , initialize the rotation axis parameters As iteration variables initial value , For the number of iterations, For the first The direction of the rotation axis in the next iteration. For the first The rotation center coordinates for the next iteration.

[0024] Step S32: Based on the rotation axis parameters of the rotary table and the angle sequence under multiple views, theoretically pre-align the point clouds in the point cloud sequence to generate a pre-aligned point cloud set; Specifically, based on the current rotation axis parameters Extract the unit direction vector of the rotation axis. Three-dimensional components Construct the cross product matrix element by element according to the definition of the cross product matrix. For angle sequences Each rotation angle Corresponding perspective Through the rotation formula Element-by-element calculation around the axis of rotation Rotation The theoretical transformation matrix, where, For the first Perspective during the next iteration The rotation transformation matrix, It is the identity matrix. From the perspective The reverse rotation angle, It is the cross product matrix; based on the theoretical transformation matrix. pass Calculate the translation vector of the center of rotation, where, For the first Perspective during the next iteration Rotation Center The translation vector, It is the identity matrix. For the first Perspective during the next iteration The rotation transformation matrix, For the first Rotation center at the next iteration; based on the rotation transformation matrix With translation vector Construct the first Perspective during the next iteration homogeneous transformation matrix ,in, ;pass Local 3D point clouds from various perspectives in a point cloud sequence Each point in Transform to the same reference coordinate system (the coordinate system of the first viewpoint), where, For the first The homogeneous coordinates after transformation in the next iteration. For the first Perspective during the next iteration The homogeneous transformation matrix, For local 3D point cloud The Middle Points The homogeneous coordinates before transformation, representing all local 3D point clouds Homogeneous coordinates after cloud transformation Perform set operation to obtain the first... Perspective during the next iteration Pre-aligned point cloud , For the first Point index in a local 3D point cloud of a frame. The range of values ​​is , For the first The number of points in the local 3D point cloud of the frame will be the first Perspective during the next iteration Pre-aligned point cloud Perform set operation to generate the first... Pre-aligned point cloud at the next iteration , For local 3D point cloud indexing, The range of values ​​is , This represents the number of points in the local 3D point cloud.

[0025] Step S33: Perform fine registration on adjacent point cloud pairs in the pre-aligned point cloud set to generate a fine registration transformation matrix set and a registration error set; Specifically, regarding the first Pre-aligned point cloud at the next iteration Each pair of adjacent point clouds ,in, For reference point cloud, To obtain the registration point cloud, fine registration is performed using the iterative nearest point algorithm to obtain adjacent point cloud pairs. fine registration transformation matrix Simultaneously calculate the root mean square error of this registration. All adjacent point cloud pairs fine registration transformation matrix and root mean square error Perform set operation to generate the first... The set of fine registration transformation matrices at the next iteration and registration error set .

[0026] Step S34: Based on the fine registration transformation matrix set, update the rotation axis parameters of the rotary table and obtain the updated rotation axis parameters; Specifically, the angle difference between two adjacent viewpoints is calculated based on the angle sequence. (The angle difference theory refers to the fixed rotation angle interval of the rotary table.) From the perspective rotation angle, From the perspective The rotation angle will be the first The set of fine registration transformation matrices at the next iteration Each fine registration transformation matrix Decompose into rotation matrices Translation vector Construct an optimization function that minimizes the difference between the theoretical and actual transformations. ,in, It is about the axis Rotation angle difference The theoretical transformation matrix, , For a unit vector Rotation angle difference The rotation matrix; for each rotation matrix Convert to axis angle representation ,in, The unit vector of the rotation axis. For the rotation angle, to ensure all Pointing in roughly the same direction, unifying their signs, if Then let Update formula via rotation axis Update rotation axis direction vector ,in, For the first The rotation axis direction after the next iteration. Let the true rotation axis direction of the rotary table be determined. Given the unit vector of the rotation axis, normalize the updated rotation axis direction vector; update the formula through the rotation center. Solving for the center of rotation ,in, It is the identity matrix. For the axis Rotation The matrix, For the first The rotation center after the next iteration update It is a translation vector. This represents the number of points in the local 3D point cloud.

[0027] Step S35: Based on the rotation axis parameters before and after the update and the registration error set, calculate the changes in parameters and errors, and determine whether convergence has occurred. If convergence has not occurred, jump to step S32 and start the next iteration. If convergence has occurred, output the final rotation axis parameters of the rotary table. Specifically, based on the rotation axis parameters before the update and updated rotation axis parameters With registration error set Through parameter change formula Calculate the changes in direction, center, and error, where, The change in direction. For the central change, The amount of error change, , The number of local 3D point clouds, For the first Perspective during the next iteration Corresponding point cloud and view For the registration error of the corresponding point cloud, the three variables are compared with the convergence threshold. When the convergence condition is met... When the algorithm converges, it is determined that the algorithm has converged. , Let these be the rotation axis parameters of the final rotary table, where... The threshold for direction change. The threshold for change at the center, The registration error variation threshold is used; when the convergence condition is not met, the iteration count is updated to... ,Will Proceed to step S32 to continue iterating.

[0028] Step S4: Based on the rotation axis parameters and angle sequence, calculate the rigid transformation matrix from the local 3D point cloud of each viewpoint to the same world coordinate system; Furthermore, based on the rotation axis parameters and angle sequence, calculating the rigid transformation matrix from the local 3D point cloud to the same world coordinate system for each viewpoint includes the following sub-steps: Step S41: Based on the rotation axis parameters and angle sequence, calculate the rigid transformation matrix of the point cloud for each viewpoint; Specifically, for each rotation angle in the angle sequence Corresponding perspective Based on rigid body kinematics, construct a rotation axis. Rotation A rotational transformation of an angle, which transforms a point at an angle... The local 3D point cloud of the measured object is rotated back to its zero-degree position. Based on this rotation transformation matrix and the rotation center, the corresponding translation transformation is solved. The rotation and translation transformations are combined into a homogeneous transformation matrix. This matrix represents the viewpoint. The local 3D point cloud is accurately mapped to the same world coordinate system (defined as the first-view coordinate system).

[0029] Step S42: Verify the rigid transformation matrix of the point cloud for each viewpoint; Specifically, select the rigid transformation matrix of adjacent viewpoints, calculate their relative transformation, compare it with the consistent physical rotation angle difference, and check whether their rotation axis and rotation angle are consistent with the calibration structure.

[0030] Step S5: Based on each rigid transformation matrix, transform the local 3D point clouds in the point cloud sequence to the same world coordinate system and then stitch and reconstruct the 3D point clouds. Furthermore, based on each rigid transformation matrix, the local 3D point clouds in the point cloud sequence are uniformly transformed to the same world coordinate system, and the 3D point cloud stitching and reconstruction includes the following sub-steps: Step S51: Based on each rigid transformation matrix, perform a unified coordinate transformation on all local 3D point clouds in the point cloud sequence, and perform preliminary fusion to generate an overall point cloud set. Specifically, the local 3D point cloud in each viewpoint of the point cloud sequence is traversed sequentially. The coordinates of each 3D point in the local 3D point cloud in each viewpoint are transformed by its rigid transformation matrix. The new coordinates in the world coordinate system are calculated. All the transformed point cloud data are merged into a preliminary overall point cloud set containing a large number of overlapping areas.

[0031] Step S52: Optimize the overall point cloud set and perform 3D point cloud stitching and reconstruction; Specifically, point cloud dilution is performed on densely overlapping areas of the overall point cloud set to remove redundant points caused by multi-view observations, while retaining the key point set representing the object surface to form a uniform point cloud; based on the optimized point cloud, a continuous and smooth three-dimensional mesh surface model is generated through a surface reconstruction algorithm.

[0032] Example 2 like Figure 2 As shown, Embodiment 2 of this application provides a three-dimensional point cloud stitching and reconstruction device based on a rotary table, comprising: The system layout module 21 places the object to be measured at the center of the rotating platform and fixes the 3D camera at a position that can cover the object on the rotating platform. The data acquisition module 22 controls the rotary table to rotate at fixed angular intervals. At each rotation angle, it triggers the 3D camera to acquire the local three-dimensional point cloud of the object under test, and obtains a set of point cloud sequences and angle sequences from multiple perspectives. Furthermore, the data acquisition module 22 includes the following sub-modules: The parameter setting submodule sets the fixed rotation angle interval and number of rotations of the turntable based on the shape of the object being measured. The point cloud acquisition submodule drives the rotary table to rotate step by step at fixed rotation angle intervals. After rotating to a target angle, it triggers the 3D camera to acquire a frame of local 3D point cloud of the object under test and records the current target angle. The point cloud and angle sequence construction submodule obtains a point cloud sequence composed of local 3D point clouds of each frame and an angle sequence composed of each target angle after the number of rotations of the turntable reaches the set number of rotations. The rotation axis parameter acquisition module 23, based on the point cloud sequence and angle sequence under multiple views, solves the rotation axis parameters of the rotary table in the 3D camera coordinate system through a joint iterative calibration algorithm. Furthermore, the rotation axis parameter acquisition module 23 includes the following sub-modules: The rotation axis parameter initialization submodule initializes and sets the rotation axis parameters of the rotary table based on the positional relationship between the 3D camera and the rotary table and the point cloud sequence. The pre-aligned point cloud generation submodule, based on the rotation axis parameters of the rotary table and the angle sequence under multiple views, theoretically pre-aligns the point clouds in the point cloud sequence to generate a pre-aligned point cloud. The point cloud fine registration submodule performs fine registration on adjacent point cloud pairs in the pre-aligned point cloud set, generating a fine registration transformation matrix set and a registration error set; The rotation axis parameter update submodule updates the rotation axis parameters of the rotary table based on the fine registration transformation matrix set and obtains the updated rotation axis parameters. The convergence judgment submodule calculates the changes in parameters and errors based on the rotation axis parameters and registration error set before and after the update, and determines whether convergence has occurred. If convergence has not occurred, it jumps to the pre-alignment point cloud generation submodule to start the next iteration. If convergence has occurred, it outputs the final rotation axis parameters of the rotary table. The rigid transformation matrix generation module 24 calculates the rigid transformation matrix from the local 3D point cloud of each viewpoint to the same world coordinate system based on the rotation axis parameters and angle sequence. Furthermore, the rigid transformation matrix calculation module 24 includes the following sub-modules: The rigid transformation matrix calculation submodule calculates the rigid transformation matrix of the point cloud for each viewpoint based on the rotation axis parameters and angle sequence. The verification submodule verifies the rigid transformation matrix of the point cloud for each viewpoint; The point cloud stitching and reconstruction module 25, based on each rigid transformation matrix, transforms the local three-dimensional point clouds in the point cloud sequence to the same world coordinate system and performs stitching and reconstruction of the three-dimensional point clouds. Furthermore, the point cloud stitching and reconstruction module 25 includes the following sub-modules: The overall point cloud set generation submodule, based on each rigid transformation matrix, performs a unified coordinate transformation on all local 3D point clouds in the point cloud sequence, performs preliminary fusion, and generates an overall point cloud set. The 3D point cloud stitching and reconstruction submodule optimizes the overall point cloud set and performs 3D point cloud stitching and reconstruction. Corresponding to the above embodiments, the present invention provides a computer storage medium, including: at least one memory and at least one processor; The memory is used to store one or more program instructions; A processor is used to run one or more program instructions to execute a 3D point cloud stitching and reconstruction method based on a rotary table.

[0033] Corresponding to the above embodiments, this embodiment of the invention provides a computer-readable storage medium containing one or more program instructions, which are used by a processor to perform a three-dimensional point cloud stitching and reconstruction method based on a rotary table.

[0034] The embodiments disclosed in this invention provide a computer-readable storage medium storing computer program instructions. When the computer program instructions are executed on a computer, the computer performs the above-described method for three-dimensional point cloud stitching and reconstruction based on a rotary table.

[0035] In this embodiment of the invention, the processor can be an integrated circuit chip with signal processing capabilities. The processor can be a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components.

[0036] The various methods, steps, and logic diagrams disclosed in the embodiments of this invention can be implemented or executed. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the methods disclosed in the embodiments of this invention can be directly implemented by a hardware decoding processor, or implemented by a combination of hardware and software modules in the decoding processor. The software modules can reside in random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, registers, or other mature storage media in the art. The processor reads information from the storage medium and, in conjunction with its hardware, completes the steps of the above methods.

[0037] The storage medium can be memory, such as volatile memory or non-volatile memory, or may include both volatile and non-volatile memory.

[0038] Among them, non-volatile memory can be read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), or flash memory.

[0039] Volatile memory can be random access memory (RAM), which is used as an external cache. By way of example, but not limitation, many forms of RAM are available, such as static random access memory (SRAM), dynamic random access memory (DRAM), synchronous dynamic random access memory (SDRAM), double data rate synchronous dynamic random access memory (DDRSDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synchronous linked dynamic random access memory (Synchlink DRAM, SLDRAM), and direct memory bus RAM (DRRAM).

[0040] The storage media described in the embodiments of the present invention are intended to include, but are not limited to, these and any other suitable types of memory.

[0041] Those skilled in the art will recognize that, in one or more of the examples above, the functions described in this invention can be implemented using a combination of hardware and software. When applied as software, the corresponding functions can be stored in a computer-readable medium or transmitted as one or more instructions or code on a computer-readable medium. Computer-readable media include computer storage media and communication media, wherein communication media include any medium that facilitates the transmission of computer programs from one place to another. Storage media can be any available medium that can be accessed by a general-purpose or special-purpose computer.

[0042] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made on the basis of the technical solution of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for 3D point cloud stitching and reconstruction based on a rotary table, characterized in that, include: Step S1: Place the object to be tested in the center of the rotating platform and fix the 3D camera in a position that can cover the object on the rotating platform; Step S2: Control the turntable to rotate at fixed angular intervals. At each rotation angle, trigger the 3D camera to acquire local 3D point clouds of the object under test, and obtain a set of point cloud sequences and angle sequences from multiple perspectives. Step S3: Based on the point cloud sequence and angle sequence from multiple viewpoints, solve for the rotation axis parameters of the rotary table in the 3D camera coordinate system using a joint iterative calibration algorithm; specifically including the following sub-steps: Step S31: Based on the positional relationship between the 3D camera and the rotary table and the point cloud sequence, initialize and set the rotation axis parameters of the rotary table; Step S32: Based on the rotation axis parameters of the rotary table and the angle sequence under multiple views, theoretically pre-align the point clouds in the point cloud sequence to generate a pre-aligned point cloud set; Step S33: Perform fine registration on adjacent point cloud pairs in the pre-aligned point cloud set to generate a fine registration transformation matrix set and a registration error set; Step S34: Based on the fine registration transformation matrix set, update the rotation axis parameters of the rotary table and obtain the updated rotation axis parameters; Step S35: Based on the rotation axis parameters before and after the update and the registration error set, calculate the changes in parameters and errors, and determine whether convergence has occurred. If convergence has not occurred, jump to step S32 and start the next iteration. If convergence has occurred, output the final rotation axis parameters of the rotary table. Step S4: Based on the rotation axis parameters and angle sequence, calculate the rigid transformation matrix from the local 3D point cloud of each viewpoint to the same world coordinate system; Step S5: Based on each rigid transformation matrix, transform the local 3D point clouds in the point cloud sequence to the same world coordinate system and then stitch and reconstruct the 3D point clouds.

2. The method for three-dimensional point cloud stitching and reconstruction based on a rotary table as described in claim 1, characterized in that, The rotary table is controlled to rotate at fixed angular intervals. At each rotation angle, a 3D camera is triggered to acquire a local 3D point cloud of the object under test, obtaining a set of point cloud sequences and angle sequences from multiple perspectives. This includes the following sub-steps: Step S21: Based on the shape of the object being measured, set the fixed rotation angle interval and the number of rotations of the rotary table; Step S22: Drive the rotary table to rotate at fixed rotation angle intervals. After rotating to a target angle, trigger the 3D camera to acquire a frame of local 3D point cloud of the object under test and record the current target angle. Step S23: When the number of rotations of the turntable reaches the set number of rotations, a point cloud sequence composed of local 3D point clouds of each frame and an angle sequence composed of each target angle are obtained.

3. The method for three-dimensional point cloud stitching and reconstruction based on a rotary table as described in claim 1, characterized in that, Based on the rotation axis parameters and angle sequence, calculating the rigid transformation matrix from the local 3D point cloud to the same world coordinate system for each viewpoint includes the following sub-steps: Step S41: Based on the rotation axis parameters and angle sequence, calculate the rigid transformation matrix of the point cloud for each viewpoint; Step S42: Verify the rigid transformation matrix of the point cloud for each viewpoint.

4. The method for three-dimensional point cloud stitching and reconstruction based on a rotary table as described in claim 1, characterized in that, Based on the rigid transformation matrices, the local 3D point clouds in the point cloud sequence are uniformly transformed to the same world coordinate system, and the 3D point cloud stitching and reconstruction includes the following sub-steps: Step S51: Based on each rigid transformation matrix, perform a unified coordinate transformation on all local 3D point clouds in the point cloud sequence, and perform preliminary fusion to generate an overall point cloud set. Step S52: Optimize the overall point cloud set and perform 3D point cloud stitching and reconstruction.

5. A three-dimensional point cloud stitching and reconstruction device based on a rotary table, characterized in that, include: The system layout module places the object to be measured in the center of the rotating platform and fixes the 3D camera in a position that can cover the object on the rotating platform; The data acquisition module controls the turntable to rotate at fixed angular intervals. At each rotation angle, it triggers the 3D camera to acquire local 3D point clouds of the object under test, obtaining a set of point cloud sequences and angle sequences from multiple perspectives. The rotation axis parameter acquisition module, based on point cloud sequences and angle sequences from multiple viewpoints, solves for the rotation axis parameters of the rotary table in the 3D camera coordinate system through a joint iterative calibration algorithm; specifically including: The rotation axis parameter initialization submodule initializes and sets the rotation axis parameters of the rotary table based on the positional relationship between the 3D camera and the rotary table and the point cloud sequence. The pre-aligned point cloud generation submodule, based on the rotation axis parameters of the rotary table and the angle sequence under multiple views, theoretically pre-aligns the point clouds in the point cloud sequence to generate a pre-aligned point cloud. The point cloud fine registration submodule performs fine registration on adjacent point cloud pairs in the pre-aligned point cloud set, generating a fine registration transformation matrix set and a registration error set; The rotation axis parameter update submodule updates the rotation axis parameters of the rotary table based on the fine registration transformation matrix set and obtains the updated rotation axis parameters. The convergence judgment submodule calculates the changes in parameters and errors based on the rotation axis parameters and registration error set before and after the update, and determines whether convergence has occurred. If convergence has not occurred, it jumps to the pre-alignment point cloud generation submodule to start the next iteration. If convergence has occurred, it outputs the final rotation axis parameters of the rotary table. The rigid transformation matrix generation module calculates the rigid transformation matrix from the local 3D point cloud of each viewpoint to the same world coordinate system based on the rotation axis parameters and angle sequence. The point cloud stitching and reconstruction module, based on various rigid transformation matrices, transforms the local 3D point clouds in the point cloud sequence to the same world coordinate system, and then stitches and reconstructs the 3D point clouds.

6. The three-dimensional point cloud stitching and reconstruction device based on a rotary table as described in claim 5, characterized in that, The data acquisition module specifically includes: The parameter setting submodule sets the fixed rotation angle interval and number of rotations of the turntable based on the shape of the object being measured. The point cloud acquisition submodule drives the rotary table to rotate step by step at fixed rotation angle intervals. After rotating to a target angle, it triggers the 3D camera to acquire a frame of local 3D point cloud of the object under test and records the current target angle. The point cloud and angle sequence construction submodule obtains a point cloud sequence composed of local 3D point clouds of each frame and an angle sequence composed of each target angle after the number of rotations of the turntable reaches the set number of rotations.

7. The three-dimensional point cloud stitching and reconstruction device based on a rotary table as described in claim 5, characterized in that, The rigid transformation matrix calculation module specifically includes: The rigid transformation matrix calculation submodule calculates the rigid transformation matrix of the point cloud for each viewpoint based on the rotation axis parameters and angle sequence. The verification submodule verifies the rigid transformation matrix of the point cloud for each viewpoint.

8. The three-dimensional point cloud stitching and reconstruction device based on a rotary table as described in claim 5, characterized in that, The point cloud stitching and reconstruction module specifically includes: The overall point cloud set generation submodule, based on each rigid transformation matrix, performs a unified coordinate transformation on all local 3D point clouds in the point cloud sequence, performs preliminary fusion, and generates an overall point cloud set. The 3D point cloud stitching and reconstruction submodule optimizes the overall point cloud set and performs 3D point cloud stitching and reconstruction.

Images