3D image stitching method and system of 3D image stitching system

Abstract

The embodiment of the application discloses a 3D image splicing method and system of a 3D image splicing system, and belongs to the field of image data processing.The method uses a polyhedron calibration block to replace a traditional calibration block to improve the instability of feature extraction, and the polyhedron calibration block can identify the entire plane feature, thereby reducing the error caused by the instability of feature extraction.At the same time, through multi-step image feature extraction and splicing of the scanned image of the polyhedron calibration block, the problem of different position shrinkage is overcome, so that more accurate image splicing is realized in the 3D visual space.The present application fills the technical gap of the traditional image splicing method, provides an innovative solution for realizing high-precision image splicing in the 3D visual space, has a wide application prospect in the fields of industrial manufacturing, medical imaging and the like, and is expected to bring significant technical progress to high-precision image processing in related fields.

Description

Technical Field

[0001] This application relates to the field of image data processing, and in particular to a 3D image stitching method and system. Background Technology

[0002] With the widespread application of 3D vision technology in industries, medicine, science, and other fields, the demand for high-precision image stitching is increasing. In existing image stitching methods, especially in product manufacturing, the common approach is to attach calibration blocks to the product surface. However, this method has several challenging technical problems.

[0003] However, due to the instability of feature extraction from the product surface, conventional methods are prone to errors in feature extraction, which in turn affects the accuracy of image stitching. Secondly, calibration blocks made using 3D printing will shrink to varying degrees at different locations, introducing errors when matching and stitching them with calibration blocks from the 3D file. In addition, factors such as camera accuracy, robot accuracy, and calibration errors also have a certain impact on the accuracy of image combination and stitching. Summary of the Invention

[0004] This application provides a 3D image stitching method and system, which can solve the technical problems of unstable feature extraction, errors caused by shrinkage of 3D printing calibration blocks, and insufficient image stitching accuracy in traditional image processing stitching methods. The technical solution is as follows:

[0005] On one hand, a 3D image stitching method for a 3D image stitching system is provided. The system includes a control terminal and a laser 3D camera. The method is applicable to the control terminal. The 3D image stitching method for the 3D image stitching system includes:

[0006] Obtain scanned images of the polyhedral calibration block at different angles to obtain a set of scanned images;

[0007] Obtain the different feature planes on each scanned image;

[0008] In each scanned image, three adjacent feature planes are selected based on the target location and the intersecting feature lines are extracted. This step is repeated until all feature lines of the scanned image set are extracted.

[0009] Select two adjacent feature lines based on the target location and extract the intersection points of the intersecting lines. Repeat this step until all line intersection points of the scanned image set are extracted.

[0010] Select two adjacent scanned images that have overlapping feature planes, and select at least p intersection points of straight lines at the same position on each of the two scanned images to form two point sequences, where p is a positive integer greater than 2;

[0011] The two combined point sequences are matched in position to obtain the image stitching transformation matrix;

[0012] The second scanned image is transformed into the space of the first scanned image through the image stitching transformation matrix and then combined and stitched together until all scanned images are stitched together.

[0013] Optionally, scanned images of the polyhedral calibration block at different angles are obtained to obtain a scanned image set, including:

[0014] The robot is manipulated to hold the polyhedral calibration block into the scanning space of the laser 3D camera;

[0015] The laser 3D camera is manipulated to scan the polyhedral calibration block according to a preset angle requirement to obtain the scanned image set. The preset angle requirement is used to indicate that the number of polyhedral calibration block planes in each scanned image in the scanned image set is greater than m, and there are more than n identical polyhedral calibration block planes in two adjacent scanned images, where m and n are positive integers greater than 2.

[0016] Optionally, maneuvering the robot to hold the polyhedral calibration block into the scanning space of the laser 3D camera includes:

[0017] The robot is manipulated to assemble the polyhedral calibration block onto the robot actuator;

[0018] The robot is manipulated to move within the scanning area of ​​the laser 3D camera and cover the entire scanning space.

[0019] Optionally, the laser 3D camera is manipulated to scan the polyhedral calibration block according to a preset angle requirement to obtain the scanned image set, including:

[0020] The robot's motion trajectory is set according to the preset angle requirements;

[0021] The laser 3D camera acquires initial scan images from different angles.

[0022] Each initial scan image is processed to obtain the initial scan image set that has more than m polyhedral calibration block planes and n or more identical polyhedral calibration block planes in two adjacent scan images.

[0023] Optionally, in each scanned image, the control terminal selects three adjacent feature planes based on the target location and extracts intersecting feature lines, including:

[0024] Three adjacent feature planes are selected based on a specific algorithm;

[0025] The selected adjacent planes are subjected to a line feature extraction algorithm to extract the intersecting feature lines.

[0026] Optionally, selecting two adjacent feature lines based on the target location and extracting the intersection point of the intersecting lines includes:

[0027] Select two adjacent feature lines based on the target location;

[0028] The point feature extraction algorithm is used to extract the intersection points of the intersecting lines by applying a point feature extraction algorithm to the selected adjacent feature lines.

[0029] Optionally, select two adjacent scanned images that have overlapping feature planes, and select at least p intersection points of straight lines at the same position on each of the two scanned images to form two point sequences, including:

[0030] Using an image matching algorithm, select two adjacent scanned images with overlapping feature planes from the scanned image set;

[0031] Select at least p intersection points of straight lines at the same position on each of the two selected scan images;

[0032] Combine the intersection points of the selected lines into two point sequences.

[0033] Optionally, the second scanned image is transformed into the space of the first scanned image using the image stitching transformation matrix and then combined and stitched, including:

[0034] The image stitching transformation matrix is ​​obtained by matching the positions of the two point columns using an image registration algorithm.

[0035] The second scanned image is combined and stitched together in the space of the first scanned image using the image stitching transformation matrix.

[0036] On the other hand, a 3D image stitching system is provided, the system including a control terminal, a robot and a laser 3D camera as described above.

[0037] Optionally, the control terminal, the robot, and the laser 3D camera can communicate via wired or wireless means.

[0038] The 3D image stitching method of the 3D image stitching system provided in this application embodiment includes at least the following technical effects.

[0039] This application provides a 3D image stitching method for a 3D image stitching system. It employs a polyhedral calibration block instead of a traditional calibration block to improve the instability of feature extraction. The polyhedral calibration block can identify features across the entire plane, thereby reducing errors caused by feature extraction instability. Simultaneously, by performing multi-step image feature extraction and stitching on the scanned image of the polyhedral calibration block, the problem of shrinkage at different locations is overcome, thus achieving more accurate image stitching in 3D visual space. Furthermore, this invention fills a technological gap in traditional image stitching methods, providing an innovative solution for achieving high-precision image stitching in 3D visual space. It has broad application prospects in fields such as industrial manufacturing and medical imaging, and is expected to bring significant technological advancements to high-precision image processing in related fields. In this application embodiment, by introducing a polyhedral calibration block and using scanned images from different angles using the polyhedral calibration block, the image stitching error problem caused by unstable feature extraction in traditional methods is solved. The geometry of the polyhedral calibration block can improve the instability of feature extraction and increase the accuracy of image stitching.

[0040] In other embodiments, by selecting the condition that there are repeating polyhedral calibration block planes on two adjacent scanned images, and using the feature points of the polyhedral calibration blocks for matching during the image stitching process, the stitching error caused by calibration error, camera accuracy and robot accuracy is reduced, and the stability of the overall system is improved.

[0041] In other embodiments, a polyhedral calibration block is used to stabilize the feature extraction process by extracting the feature lines and intersections of adjacent planes, thus avoiding errors caused by different shrinkage amounts of different planes.

[0042] In the remaining embodiments, the introduction of polyhedral calibration blocks, laser 3D cameras, collaborative operation of robots, and the calculation and application of image stitching transformation matrices collectively achieve high-precision image stitching in 3D visual space. This has significant practical implications for applications with high precision requirements, such as process execution.

[0043] In the remaining embodiments, through reasonable robot movement and scanning image acquisition strategies, the scheme ensures that the polyhedral calibration block is fully covered in the scanning space of the laser 3D camera, thereby improving the stability of the entire system. Simultaneously, the adoption of preset angle requirements effectively avoids the problem of insufficient geometric structure in the scanned images, further improving the reliability of the system.

[0044] In summary, the 3D image stitching method of the 3D image stitching system provided in the various embodiments of this application effectively solves the common error and instability problems in 3D image stitching through meticulous robot control, feature extraction and image stitching process, and provides reliable technical support for high-precision 3D vision applications. Attached Figure Description

[0045] Figure 1 This invention illustrates a schematic diagram of the structure of a 3D image stitching system provided in an exemplary embodiment of this application;

[0046] Figure 2 A flowchart illustrating a 3D image stitching method of a 3D image stitching system provided in an exemplary embodiment of this application is shown.

[0047] Figures 3 to 6 For the corresponding Figure 2 A schematic diagram illustrating the process of generating a scanned image set by scanning a polyhedral calibration block. Detailed Implementation

[0048] To make the objectives, technical solutions, and advantages of this application clearer, the embodiments of this application will be described in further detail below with reference to the accompanying drawings.

[0049] In this article, "multiple" refers to two or more. "And / or" describes the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A alone, A and B simultaneously, or B alone. The character " / " generally indicates that the preceding and following related objects have an "or" relationship.

[0050] like Figure 1 The diagram illustrates a 3D image stitching system according to an exemplary embodiment of this application. The system includes a control terminal 10, a robot 20, and a laser 3D camera 30.

[0051] In this system, the control terminal, as the main controller of the 3D image stitching method, plays a crucial role in system control and is connected to both the robot and the laser 3D camera. The control terminal can be an industrial console, an embedded computer, such as an industrial PC or an embedded industrial control system; there are no restrictions on its type.

[0052] In various embodiments of this application, the control terminal is responsible for communicating with the robot and the laser 3D camera, and executing the control logic for image stitching, including angle scanning of polyhedral calibration blocks, execution of feature extraction algorithms, point matching, and generation of image stitching transformation matrices.

[0053] In this system, the robot, as the executor of the 3D image stitching method, plays a key operational role, primarily executing tasks under the instructions of the control terminal. For example... Figure 1 As shown, in order to ensure that the polyhedral calibration block remains within the scanning area during movement, the position information of the polyhedral calibration block can be sent to the control terminal in real time to ensure the accuracy of movement.

[0054] In the various embodiments of this application, the robot's main task is to hold the polyhedral calibration block and move it to the scanning area of ​​the laser 3D camera. The robot has high-precision motion control to ensure that the calibration block covers the entire space within the scanning area and completes image scanning according to preset angle requirements.

[0055] The laser 3D camera is responsible for scanning the polyhedral calibration block at different angles, generating a scanned image set, and acquiring high-precision 3D point cloud data, providing a foundation for subsequent feature extraction and image stitching. Furthermore, the laser 3D camera performs the scanning task according to preset angle requirements and generates a scanned image set that meets the requirements.

[0056] In addition, the system's control terminal, robot, and laser 3D camera have good compatibility and interactivity, ensuring that the system can operate stably and reliably in practical applications.

[0057] Please refer to Figure 2 This document illustrates a flowchart of a 3D image stitching method for a 3D image stitching system according to an exemplary embodiment of this application. This method is applicable to… Figure 1 The 3D image stitching system shown includes the following method:

[0058] Step 201: The control terminal acquires scanned images of the polyhedral calibration block at different angles to obtain a set of scanned images.

[0059] In order to obtain scanned images from different angles, it is necessary to ensure that the polyhedral calibration block is fully covered in the scanning space of the laser 3D camera. Therefore, the control terminal manipulates the robot to carry the polyhedral calibration block into the scanning space of the laser 3D camera.

[0060] In one example, the control terminal manipulates the robot to assemble the polyhedral calibration block onto the robot actuator; the control terminal also manipulates the robot to move within the scanning area of ​​the laser 3D camera and cover the entire scanning space.

[0061] First, the control terminal needs to determine the scanning area of ​​the laser 3D camera, either by pre-setting scanning area parameters or by measuring the environment in real time, where the scanning area meets the requirements for image stitching.

[0062] Furthermore, the control terminal determines the robot's movement path based on the scanning area of ​​the laser 3D camera, and sends motion commands through the communication interface with the robot system to control the robot to move to the predetermined scanning area.

[0063] During robot movement, the control terminal may need to monitor the robot's position in real time to ensure it accurately covers the entire scanning area. If errors exist or adjustments are needed, the control terminal can send corresponding correction commands. Once the robot has moved to the scanning area and covered the entire space, the control terminal notifies the robot to complete the task. The robot can pause at a certain position to wait for the laser 3D camera to scan, or perform other actions as needed.

[0064] Figures 3 to 6 For the corresponding Figure 2 This diagram illustrates the process of generating a scanned image set by scanning polyhedral calibration blocks. The control terminal manipulates a laser 3D camera to scan the polyhedral calibration blocks according to preset angle requirements, resulting in a scanned image set. The preset angle requirements indicate that each scanned image in the image set contains more than m polyhedral calibration block planes, and that there are more than n identical polyhedral calibration block planes between two adjacent scanned images, where m and n are positive integers greater than 2.

[0065] In one example, the control terminal sets the robot's motion trajectory according to preset angle requirements and acquires initial scan images at different angles using a laser 3D camera. The control terminal then processes each initial scan image, acquiring those with more than m polyhedral calibration block planes and n or more identical polyhedral calibration block planes between adjacent scan images, thus obtaining a scan image set. For example, m is 5 and n is 5.

[0066] In this step, the selection of the number of polyhedral calibration block planes (m) and the number of identical polyhedral calibration block planes (n) on two adjacent scanned images is related to the acquisition accuracy.

[0067] Selection of the number of polyhedral calibration block planes (m): When selecting m, the geometry and surface complexity of the scanned object need to be considered. Choosing a suitable value for m ensures that each scanned image contains a sufficient number of polyhedral calibration block planes, thereby improving the accuracy of subsequent image stitching.

[0068] Selection of the number (n) of identical polyhedral calibration block planes on two adjacent scanned images: The rationality of choosing n is related to the image matching algorithm. By ensuring that there are a sufficient number of identical polyhedral calibration block planes on two adjacent images, the reliability of the matching points can be improved, which helps to reduce errors in the subsequent image stitching process. In practice, based on the performance of the image processing algorithm, repeated testing and adjustments can be made to achieve the optimal value of n.

[0069] Selecting details related to acquisition accuracy: Factors related to acquisition accuracy include the resolution of the laser 3D camera, the accuracy of robot motion, and the precision of calibration block manufacturing. Ensuring sufficient accuracy in these aspects is to reduce errors during image acquisition. For example, for laser 3D cameras, choosing cameras with higher resolution can improve acquisition accuracy, while precise robot motion and high-precision manufacturing of calibration blocks are also crucial to ensuring system accuracy.

[0070] In one possible implementation, relevant parameters are adjusted according to the specific application scenario and system performance requirements to achieve high-quality image stitching.

[0071] In one example, to ensure stitching accuracy, a sufficient number of identical polyhedral calibration planes are required between adjacent images. Simultaneously, to guarantee that each image provides enough information for stitching, a minimum number of polyhedral calibration planes is set. A minimum number of planes is set, for example, m = 5. This means that each scanned image contains at least 5 polyhedral calibration planes. An appropriate value for n is chosen, for example, n = 4. This indicates that there are at least 4 identical polyhedral calibration planes between adjacent images. The choice of this value typically depends on the requirements and performance of the stitching algorithm, as well as the layout and density of calibration blocks in the scene.

[0072] In the example above, the choices of m and n are based on an understanding of the algorithm and the scenario, and can be optimized through experimentation and adjustment. It should be noted that the optimal values ​​in practical applications may vary depending on the scenario, device, and algorithm. Through repeated testing and adjustments in real-world applications, the most suitable parameter values ​​for a specific situation can be found.

[0073] Step 202: Obtain different feature planes on each scanned image.

[0074] In step 201, the laser 3D camera is controlled via the control terminal to scan the polyhedral calibration blocks, obtaining a set of scanned images. This step ensures that the system obtains scanned images containing multiple polyhedral calibration blocks. Further feature plane extraction is performed on each scanned image, such as... Figure 3 As shown, the control terminal selects three adjacent feature planes based on the target location and extracts the intersecting feature lines (step 203). Figure 3 The image shows a scanned image, with three adjacent feature planes indicated from left to right. This step uses a specific algorithm to extract features from the selected adjacent planes to obtain different feature planes in the image.

[0075] In each scanned image, after selecting three adjacent feature planes, as follows: Figure 4As shown, the intersection points of three adjacent feature planes are obtained by finding the intersection points of two intersecting feature lines, i.e., the intersection points of intersecting lines are extracted. Specifically, a line feature extraction algorithm is used to extract the intersecting feature lines (step 203). These feature lines will be used for point combination and position matching in subsequent steps.

[0076] Through the above steps, the system can acquire different feature planes on each scanned image and extract the corresponding feature lines, laying the foundation for subsequent image stitching and 3D reconstruction. This process ensures detailed feature extraction of scanned images of the polyhedral calibration block at different angles, supporting subsequent steps such as image stitching and 3D model generation.

[0077] Step 203: In each scanned image, the control terminal selects three adjacent feature planes based on the target location and extracts the intersecting feature lines.

[0078] Repeat this step until all feature lines in the scanned image set are extracted.

[0079] To obtain the feature lines, the control terminal selects three adjacent feature planes based on the target location and uses a line feature extraction algorithm on the selected adjacent planes to extract the intersecting feature lines. The target location refers to the identical feature planes on two adjacent scanned images.

[0080] Step 204: The control terminal selects two adjacent feature lines based on the target location and extracts the intersection point of the intersecting lines.

[0081] Repeat this step until all line intersections of the scanned image set are extracted.

[0082] Furthermore, after obtaining the feature lines, the intersection points of the lines are obtained. The control terminal selects two adjacent feature lines according to the target position, and uses a point feature extraction algorithm on the selected adjacent feature lines to extract the intersection points of the intersecting lines.

[0083] Step 205: The control terminal selects two adjacent scanned images with overlapping feature planes and selects at least p intersection points of straight lines at the same position on each of the two scanned images to form two point sequences.

[0084] Where p is a positive integer greater than 2. For example, p = 3. In this step, two adjacent scanned images with overlapping feature planes are selected, and at least p intersection points of straight lines at the same position are selected on each of these two images to form two point sequences. The following is a specific supplement and example of this step.

[0085] The selection of p is related to the stitching accuracy: the specific value of parameter p needs to consider the accuracy requirements of image stitching. A larger p value may increase the stability of matching, but may also lead to point-to-point matching failure. Conversely, a smaller p value may increase the flexibility of stitching, but may also be susceptible to noise interference. Therefore, depending on the specific application scenario and system requirements, an appropriate p value can be selected through experimentation and adjustment.

[0086] In one example, assume that the same feature plane exists in two scanned images, where the p-value is set to 3. The control terminal selects two adjacent scanned images with overlapping feature planes using an image matching algorithm. On each image, at least three intersection points of straight lines at the same location are selected, forming two point sequences.

[0087] The coordinates of the intersection points of the lines on image A are: (x1, y1), (x2, y2), (x3, y3).

[0088] The coordinates of the intersection points of the corresponding lines on image B are: (x1', y1'), (x2', y2'), (x3', y3').

[0089] Thus, the two point sequences are: [(x1,y1),(x2,y2),(x3,y3)] and [(x1',y1'),(x2',y2'),(x3',y3')].

[0090] By selecting images with repeating feature planes and choosing the intersection points of straight lines at the same locations on them, the system obtains two corresponding point sequences, providing crucial data for the next step of location matching.

[0091] In one possible implementation, the control terminal uses an image matching algorithm to select two adjacent scanned images with overlapping feature planes in the scanned image set. The control terminal then selects at least p intersection points of straight lines at the same position on the two selected scanned images and combines the selected intersection points into two point sequences.

[0092] Step 206: The control terminal performs position matching on the two combined point columns to obtain the image stitching transformation matrix.

[0093] In step 206, the control terminal performs position matching on the two combined point columns to finally obtain the image stitching transformation matrix.

[0094] like Figure 5 As shown, in one example, at least three intersection points are extracted at the same location between two adjacent images and combined into a point column. The image stitching transformation matrix of the two images is obtained by matching. The following is a detailed description of the principle.

[0095] Location matching: The location matching process typically uses image registration algorithms. A common approach is feature point matching, where points in two point sequences are matched to find corresponding points using a certain algorithm. This can include feature descriptor-based methods, such as SIFT (Scale Invariant Feature Transform) or SURF (Speed-Up Robust Feature Transform). The matching result will give the correspondence between points in the two point sequences.

[0096] Image stitching transformation matrix: Once position matching is complete, the image stitching transformation matrix can be calculated using these corresponding points. This is typically achieved using a homography matrix, especially in the case of perspective transformations. Image processing libraries such as OpenCV provide functions for calculating this transformation.

[0097] For a sequence of points on two two-dimensional planes, the homography matrix H can be calculated using algorithms such as DLT (Direct Linear Transform) or RANSAC (Random Sample Consensus).

[0098] Once the homography matrix H is obtained, it can be used to transform the space of one image into that of another, thereby achieving image stitching.

[0099] Through this process, the control terminal successfully obtained the stitching transformation matrix between the two images, providing accurate spatial transformation information for subsequent image stitching.

[0100] Step 207: The control terminal converts the second scanned image into the space of the first scanned image through the image stitching transformation matrix and then combines and stitches them together until all scanned images are stitched together.

[0101] The control terminal uses an image registration algorithm to match the positions of two point sequences to obtain an image stitching transformation matrix. The control terminal then uses the image stitching transformation matrix to combine and stitch the second scanned image in the space of the first scanned image.

[0102] In one example, such as Figure 6 As shown, the second scanned image is transformed into the space of the first scanned image through a stitching transformation matrix, and the images are then stitched together until all scanned images are stitched together.

[0103] In this embodiment, by introducing a polyhedral calibration block and using scanned images from different angles of the polyhedral calibration block, the image stitching error problem caused by unstable feature extraction in traditional methods is solved. The geometry of the polyhedral calibration block can improve the instability of feature extraction and increase the accuracy of image stitching.

[0104] In this embodiment, by selecting the condition that there are repeating polyhedral calibration block planes on two adjacent scanned images, and using the feature points of the polyhedral calibration blocks for matching during the image stitching process, the stitching error caused by calibration error, camera accuracy and robot accuracy is reduced, and the overall system stability is improved.

[0105] In this embodiment, a polyhedral calibration block is used to extract the feature lines and intersections of adjacent planes, thereby stabilizing the feature extraction process and avoiding errors caused by different shrinkage amounts of different planes.

[0106] In the embodiments of this application, the introduction of polyhedral calibration blocks, laser 3D cameras, collaborative operation of robots, and the calculation and application of image stitching transformation matrices collectively achieve high-precision image stitching in 3D visual space. This has significant practical implications for applications with high precision requirements, such as process execution.

[0107] In this embodiment, by employing a reasonable robot movement and scanning image acquisition strategy, the scheme ensures that the polyhedral calibration block is fully covered within the scanning space of the laser 3D camera, thereby improving the stability of the entire system. Simultaneously, the use of preset angle requirements effectively avoids the problem of insufficient geometric structure in the scanned images, further enhancing the system's reliability.

[0108] In summary, the 3D image stitching method of the 3D image stitching system provided in this application effectively solves the common error and instability problems in 3D image stitching through meticulous robot control, feature extraction and image stitching process, and provides reliable technical support for high-precision 3D vision applications.

[0109] This application also provides a computer-readable storage medium located within a control terminal. The storage medium stores at least one instruction, at least one program, a code set, or an instruction set. The at least one instruction, the at least one program, the code set, or the instruction set is loaded and executed by the processor to implement the 3D image stitching method of the 3D image stitching system provided in the above embodiments.

[0110] Optionally, the computer-readable storage medium may include: read-only memory (ROM), random access memory (RAM), solid-state drives (SSDs), or optical discs, etc. The random access memory may include resistive random access memory (ReRAM) and dynamic random access memory (DRAM).

[0111] The sequence numbers of the embodiments in this application are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments. The above descriptions are merely optional embodiments of this application and are not intended to limit this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

Claims

1. A 3D image stitching method for a 3D image stitching system, characterized in that, The system includes a control terminal and a laser 3D camera. The method is applicable to the control terminal. The 3D image stitching method of the 3D image stitching system includes: The scanned images of the polyhedral calibration block at different angles are obtained to obtain a set of scanned images, wherein the scanned images are three-dimensional images acquired by the laser 3D camera; Obtain the different feature planes on each scanned image; In each scanned image, three adjacent feature planes are selected based on the target location and the intersecting feature lines are extracted. This step is repeated until all feature lines of the scanned image set are extracted. Select two adjacent feature lines based on the target location and extract the intersection points of the intersecting lines. Repeat this step until all line intersection points of the scanned image set are extracted. Select two adjacent scanned images that have overlapping feature planes, and select at least p intersection points of straight lines at the same position on each of the two scanned images to form two point sequences, where p is a positive integer greater than 2; The two combined point sequences are matched in position to obtain the image stitching transformation matrix; The second scanned image is transformed into the space of the first scanned image through the image stitching transformation matrix and then combined and stitched together until all scanned images are stitched together. The step of selecting three adjacent feature planes and extracting intersecting feature lines in each scanned image based on the target location includes: Three adjacent feature planes are selected in the scanned image according to a specific algorithm; The intersecting feature lines are extracted by using a line feature extraction algorithm on the selected adjacent planes. Based on the target location, two adjacent feature lines are selected and their intersection points are extracted, including: Select two adjacent feature lines based on the target location; The point feature extraction algorithm is used to extract the intersection points of the intersecting lines by selecting adjacent feature lines; Select two adjacent scanned images that have overlapping feature planes, and select at least p intersection points of straight lines at the same position on each of the two scanned images to form two point sequences, including: Using an image matching algorithm, select two adjacent scanned images with overlapping feature planes from the set of scanned images; Select at least p intersection points of straight lines at the same position on each of the two selected scan images; Combine the intersection points of the selected lines into two point sequences; The second scanned image is transformed into the spatial representation of the first scanned image using the image stitching transformation matrix and then combined and stitched, including: The image stitching transformation matrix is ​​obtained by matching the positions of the two point columns using an image registration algorithm. The second scanned image is combined and stitched together in the space of the first scanned image using the image stitching transformation matrix.

2. The 3D image stitching method of the 3D image stitching system according to claim 1, characterized in that, Obtain scanned images of the polyhedral calibration block at different angles to obtain a scanned image set, including: The robot is manipulated to hold the polyhedral calibration block into the scanning space of the laser 3D camera; The laser 3D camera is manipulated to scan the polyhedral calibration block according to a preset angle requirement to obtain the scanned image set. The preset angle requirement is used to indicate that the number of polyhedral calibration block planes in each scanned image in the scanned image set is greater than m, and there are more than n identical polyhedral calibration block planes in two adjacent scanned images, where m and n are positive integers greater than 2.

3. The 3D image stitching method of the 3D image stitching system according to claim 2, characterized in that, Manipulating the robot to hold the polyhedral calibration block into the scanning space of the laser 3D camera includes: The robot is manipulated to assemble the polyhedral calibration block onto the robot actuator; The robot is manipulated to move within the scanning area of ​​the laser 3D camera and cover the entire scanning space.

4. The 3D image stitching method of the 3D image stitching system according to claim 2, characterized in that, The laser 3D camera is manipulated to scan the polyhedral calibration block according to a preset angle requirement to obtain the scanned image set, including: The robot's motion trajectory is set according to the preset angle requirements; The laser 3D camera acquires initial scan images from different angles. Each initial scan image is processed to obtain the initial scan image set that has more than m polyhedral calibration block planes and n or more identical polyhedral calibration block planes in two adjacent scan images.

5. A 3D image stitching system, characterized in that, The system includes a control terminal, a robot, and a laser 3D camera as described in any one of claims 1 to 4.

6. The 3D image stitching system according to claim 5, characterized in that, The control terminal, the robot, and the laser 3D camera communicate via wired or wireless means.

Images