An aircraft integral tank model reconstruction method and device based on a progressive registration strategy

Abstract

The application relates to an aircraft integral tank model reconstruction method based on a progressive registration strategy, which comprises the following steps: calibrating internal and external parameters of a three-dimensional scanning module, fixing a target tool to an aircraft tank wall plate, collecting three-dimensional point cloud data by using the three-dimensional scanning module, obtaining point cloud data frames and spatial poses, performing initial registration and accurate registration between adjacent point cloud data frames, constructing an optimization objective function and a point cloud frame pose graph, optimizing the point cloud data frames, and finally generating a high-quality aircraft integral tank measured point cloud model, so as to provide high-quality model data for subsequent measurement and detection work, and improve the measurement and detection limitations of traditional methods in a closed environment of an aircraft integral tank. The application also provides a device, which also has the beneficial effects.

Description

Technical Field

[0001] This invention relates to the field of aircraft integral fuel tank data acquisition technology, and in particular to a method and apparatus for reconstructing an aircraft integral fuel tank model based on a progressive registration strategy. Background Technology

[0002] The integral fuel tank is a crucial structural component connecting the fuselage and wing box of a large aircraft, and it also serves as the aircraft's fuel storage unit. Maintaining the structural integrity of the integral fuel tank during manufacturing is essential for ensuring flight safety, and it is also crucial for improving production efficiency and reliability. Currently, the measurement and inspection of key indicators for integral fuel tank sealant coating mainly rely on manual labor and measuring tools. This method has low measurement accuracy, and quality assessment is highly dependent on the subjective judgment of the inspectors. Furthermore, traditional 3D scanning methods require manually attaching target points to the internal wall panels of the fuel tank to achieve spatial positioning of the 3D scanning equipment. However, the standardization of manual target attachment is low, which may affect the reliability of the data, a significant reason limiting the accuracy of 3D reconstruction of the fuel tank. Additionally, measuring large components requires deploying numerous targets on their surfaces, which is time-consuming and labor-intensive. Moreover, adhesive targets can easily be left inside the integral fuel tank due to human error after scanning, becoming waste material that may clog fuel lines and affect aircraft safety. Summary of the Invention

[0003] To address the shortcomings of existing technologies, this invention provides a method for reconstructing an aircraft integral fuel tank model based on a progressive registration strategy, solving the problem that traditional methods are highly dependent on manual labor and measuring tools. Another objective of this invention is to provide a device for reconstructing an aircraft integral fuel tank model based on a progressive registration strategy, solving the problem that traditional methods are highly dependent on manual labor and measuring tools.

[0004] To solve the above-mentioned technical problems, the present invention provides the following technical solution: a method for reconstructing an aircraft overall fuel tank model based on a progressive registration strategy, comprising the following steps:

[0005] S1. Calibrate the internal and external parameters of the 3D scanning module, paste the QR code on the reserved position on the target fixture, then fix the target fixture to the positioning hole designed on the aircraft fuel tank wall panel, and adjust the position of the 3D scanning module to ensure that the QR code is within the field of view of the 3D scanning module.

[0006] S2. Initialize the visual odometry of the 3D scanning module, and then start using the 3D scanning module to collect 3D point cloud data until the scanning ends when the entire glued area on the aircraft fuel tank wall is traversed, obtaining point cloud data frames and spatial poses, and aligning the point cloud data frames based on the spatial poses.

[0007] S3. Based on point-pair feature descriptors, initial registration between adjacent point cloud data frames is achieved to obtain an initial transformation matrix, which includes an initial rotation matrix. and the initial translation matrix ;

[0008] S4. Design a precise registration method based on a symmetric objective function, construct a registration objective function for adjacent point cloud data frames of aircraft fuel tank panels, optimize the initial transformation matrix, and obtain the optimal transformation matrix;

[0009] S5. Construct the optimization objective function and point cloud frame pose graph. Calculate the priority of loops in the point cloud frame pose graph based on the overlapping area of ​​adjacent point cloud data frames. Optimize the point cloud data frames by constructing a set of loop sequences with different priorities to generate the final measured point cloud model of the aircraft's overall fuel tank.

[0010] Furthermore, in step S3, the point pair feature descriptor mainly consists of point pairs from adjacent point cloud data frames. and the corresponding normal vector Its composition, and its mathematical description, are as follows:

[0011] ;

[0012] in, Represents point pairs The midpoint, i.e. The normal vector, This represents the angle between two vectors, with values ​​ranging from 1 to 2. between, and They represent The normal vector and by The angle formed by the midpoint normal vectors, express The angle between the normal vectors.

[0013] Furthermore, in step S4, the specific process includes the following steps:

[0014] S41. Based on the point-to-plane ICP method, a symmetric objective function is constructed using the initial transformation matrix as the optimization objective:

[0015] ;

[0016] in, and These are the corresponding points of adjacent point cloud data frames. and The normal vector;

[0017] S42, the initial rotation matrix in the above equation Split into two components , These correspond to the source point cloud data frame and the target point cloud data frame, respectively. Simplifying the above equation, we obtain the registration objective function for adjacent point cloud data frames of the aircraft fuel tank wall:

[0018] ;

[0019] S43. Use the LM algorithm to solve the registration objective function of adjacent point cloud data frames of aircraft fuel tank wall panels and obtain the optimal transformation matrix.

[0020] Furthermore, in step S5, the specific process includes the following steps:

[0021] S51. Minimize the Euclidean distance between point cloud data frames to construct an optimization objective function to further optimize the optimal transformation matrix obtained in step S4:

[0022] ;

[0023] ;

[0024] in, Represents a set of point cloud data frames The optimal transformation matrix corresponding to the i-th point cloud frame in the data. This represents the set of corresponding point pairs between adjacent point cloud data frames, which can be obtained through nearest neighbor search of overlapping regions; This represents the initial transformation matrix between adjacent point cloud data frames, which can be obtained through conventional point cloud registration methods.

[0025] S52. Construct a corresponding point cloud frame pose graph based on the point cloud data frame, and optimize the loop closure of the point cloud data frame according to the priority of the loop in the point cloud frame pose graph.

[0026] S53. Based on the optimization results of step S52, the point cloud data frame of the point cloud frame pose diagram formed after the loop ends is optimized again for global error to further eliminate the cumulative error and generate the final measured point cloud model of the aircraft's overall fuel tank.

[0027] Furthermore, in step S52, the specific process includes the following steps:

[0028] S521. Using point cloud data frames as nodes and overlapping areas between adjacent point cloud data frames as edges, construct a point cloud frame pose graph. The larger the overlapping area, the greater the weight assigned to its corresponding edge in the point cloud frame pose graph.

[0029] S522. A maximum spanning tree-based method is used to detect the shortest loops formed by edges in the point cloud frame pose graph. First, the maximum spanning tree is extracted based on maximizing the sum of edge weights. The remaining edges are placed in a candidate sequence. Then, the edge with the largest weight is selected from the candidate sequence and added to the maximum spanning tree. The shortest loop in the current point cloud frame pose graph is recalculated, denoted as... Next, edges from the candidate sequence are added to the maximum spanning tree in order of their weights, and the shortest cycle is updated, ultimately resulting in a set of cycle sequences. ;

[0030] S523, Calculate the set of cyclic sequences The average registration error between point cloud data frames within a single loop is used as the loop closure sequence set. Priority evaluation indicators:

[0031] ;

[0032] in and These are the shortest loops. The set of adjacent point cloud data frames in the data. This represents the number of point cloud data frames in the set.

[0033] S524. Evaluate the set of loop sequences according to priority evaluation indicators. Priority determination is performed on the point cloud data frames in the data, and the following is selected: Highest priority point cloud data frame Using the optimization objective function constructed by S51, closure loop optimization is performed. The optimized point cloud data frames are then merged into a single point cloud data set, denoted as... ;

[0034] S525, will As a single node, it reconstructs the point cloud frame pose graph with the remaining nodes. Then, using the traditional adjacent point cloud registration algorithm, it calculates and updates the correspondence between adjacent point cloud data frames in the current point cloud frame pose graph.

[0035] S526. Repeat steps S522 to S525 until all loops formed under all priorities are traversed, thus completing the optimization.

[0036] The present invention also provides an aircraft integral fuel tank model reconstruction device based on a progressive registration strategy, including a target tooling and a three-dimensional scanning module;

[0037] The internal and external parameters of the 3D scanning module are calibrated. The QR code is pasted onto the reserved position on the target fixture. Then, the target fixture is fixed to the positioning hole designed on the aircraft fuel tank wall panel. The position of the 3D scanning module is adjusted to ensure that the QR code is within the field of view of the 3D scanning module.

[0038] The 3D scanning module includes a visual odometer, a surface structured light 3D scanner, an adapter, and a flange. The visual odometer includes a binocular camera, and the surface structured light 3D scanner includes a structured light projection device and a monocular grayscale camera.

[0039] The binocular camera and the surface structured light 3D scanner are fixed together by an E-shaped adapter. The adapter is equipped with a flange at the bottom, which can be fixed to different actuators or the end of a tripod.

[0040] The intrinsic parameters of the binocular camera are calibrated, and the extrinsic parameters of the binocular camera and the surface structured light 3D scanner are calibrated simultaneously using a dual-camera calibration method.

[0041] Furthermore, the target tooling is affixed with square coded target pieces and circular coded target dots, and has an embedded RFID tag to enable tracking of the target tooling deployed inside the aircraft fuel tank.

[0042] Furthermore, when fixing the target fixture to the aircraft fuel tank wall panel, the deployment density of the target fixture can be adjusted according to the different field of view volume range of the 3D scanning module, that is, to ensure that the 3D scanning module can cover 2-3 target fixtures within the effective field of view, so as to ensure the positioning accuracy of the 3D scanning module.

[0043] By employing the above technical solution, the present invention provides a method for reconstructing an aircraft overall fuel tank model based on a progressive registration strategy, which has at least the following beneficial effects:

[0044] (1) This invention deploys target tooling on the aircraft fuel tank wall panel by combining a three-dimensional scanning module, thereby obtaining target tooling with small tooling volume, simple structure, strong adaptability, good maintainability and low cost, and improving the standardization of deployment and improving the efficiency of data acquisition.

[0045] (2) The present invention uses standardized target tooling to deploy target pieces and target points, which overcomes the problems of low target deployment efficiency and target omission in traditional three-dimensional laser measurement operations, while avoiding mechanical damage to the surface of the oil tank or contamination such as the introduction of colloidal deposits.

[0046] (3) This invention enables the tracking of a large number of target tools deployed inside the aircraft fuel tank by embedding RFID electronic tags in the target tooling, ensuring that no targets are left behind after the three-dimensional laser measurement operation is completed;

[0047] (4) By designing a progressive registration method, this invention greatly reduces the configuration error introduced by point cloud data during the registration process, thereby generating a high-quality measured point cloud model of the aircraft's overall fuel tank.

[0048] The present invention also provides an aircraft overall fuel tank model reconstruction device based on a progressive registration strategy, which also has the above-mentioned beneficial effects, and will not be described in detail here. Attached Figure Description

[0049] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments of this application and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings:

[0050] Figure 1 This is a flowchart of a method for reconstructing an aircraft integral fuel tank model based on a progressive registration strategy, according to the present invention.

[0051] Figure 2 This is a schematic diagram of the three-dimensional scanning module structure of an aircraft integral fuel tank model reconstruction device based on a progressive registration strategy according to the present invention.

[0052] Figure 3 This is a flowchart of the optimization process for loop closure of point cloud data frames based on pose graphs, as described in this invention. Detailed Implementation

[0053] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. This will allow for a full understanding of how the present application uses technical means to solve technical problems and achieve technical effects, and to facilitate its implementation.

[0054] Those skilled in the art will understand that all or part of the steps in the methods of the above embodiments can be implemented by a program instructing related hardware. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Moreover, this application can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0055] Please refer to Figures 1-3This illustrates a specific implementation of the present embodiment. By constructing and deploying a 3D scanning module and target fixture, acquiring 3D point cloud data, progressively registering point cloud data frames, and performing cyclic closure optimization based on priority, a high-quality measured point cloud model of the aircraft's overall fuel tank is finally generated. This provides high-quality model data for subsequent measurement and inspection work, improving the limitations of traditional methods in measurement and inspection within the enclosed environment of the aircraft's overall fuel tank.

[0056] Please refer to Figure 1 , Figure 1 This is a flowchart of a method for reconstructing an aircraft overall fuel tank model based on a progressive registration strategy proposed in this embodiment. The method includes the following steps:

[0057] S1. Calibrate the internal and external parameters of the 3D scanning module, paste the QR code to the reserved position on the target fixture, and then fix the target fixture to the positioning hole designed on the aircraft fuel tank wall panel. Adjust the position of the 3D scanning module to ensure that the QR code is within the field of view of the 3D scanning module. The 3D scanning module includes a visual odometry composed of binocular cameras and a surface structured light 3D scanner.

[0058] In this embodiment, the two-dimensional image data acquired by the binocular camera and the three-dimensional point cloud data acquired by the surface structured light three-dimensional scanner can be fused and complemented by algorithms, thereby further improving the three-dimensional spatial positioning accuracy and positioning stability of the device.

[0059] S2. Initialize the visual odometry of the 3D scanning module, and then start using the 3D scanning module to collect 3D point cloud data until the entire adhesive area on the aircraft fuel tank wall panel is traversed, and the scanning ends. Obtain point cloud data frames and spatial poses, and align the point cloud data frames based on the spatial poses. The initialization of the visual odometry must be performed while ensuring that the QR code is within the field of view of the 3D scanning module. Hold the scanner and perform translational movements in the left-right or up-down directions. Based on this, the collected image data can be processed to obtain the spatial pose based on the deployed target fixture. After the system initialization is completed, start collecting 3D point cloud data of the aircraft fuel tank wall panel until the entire adhesive area on the aircraft fuel tank wall panel is traversed, and then the scanning ends.

[0060] S3. Based on point-pair feature descriptors, initial registration between adjacent point cloud data frames is achieved to obtain an initial transformation matrix, which includes an initial rotation matrix. and the initial translation matrix ;

[0061] As a preferred embodiment of step S3, it specifically includes: the point pair feature descriptor is composed of point pairs from adjacent point cloud data frames. and the corresponding normal vector Its composition, and its mathematical description, are as follows:

[0062] ;

[0063] in, Represents point pairs The midpoint, i.e. The normal vector, This represents the angle between two vectors, with values ​​ranging from 1 to 2. between, and They represent The normal vector and by The angle formed by the midpoint normal vectors, express The angle between the normal vectors.

[0064] S4. Design a precise registration method based on a symmetric objective function, construct a registration objective function for adjacent point cloud data frames of aircraft fuel tank panels, optimize the initial transformation matrix, and obtain the optimal transformation matrix;

[0065] As a preferred embodiment of step S4, the specific process includes the following steps:

[0066] S41, Two corresponding points on a two-dimensional plane and its corresponding direction vector For example, a symmetric target can be constructed:

[0067] ;

[0068] The result is 0 when the vectors formed by two corresponding points are perpendicular to each other, i.e., the corresponding points... All points will lie on the same second-order surface. For all corresponding point sets, they may form multiple second-order surfaces with different radii on the symmetric target based on different geometric correspondences. In the process of minimizing the point spacing using the point-to-plane ICP method, all corresponding points in the set should lie on the same plane. This problem is extended to 3D point cloud registration, where the initial transformation matrix is ​​used as the optimization objective, and a symmetric objective function is further constructed.

[0069] ;

[0070] in, and These are the corresponding points of adjacent point cloud data frames. and The normal vector;

[0071] S42, the initial rotation matrix in the above equation Split into two components , These correspond to the source point cloud data frame and the target point cloud data frame, respectively, to reduce the rotation angle of a single rotation component during rotation transformation, thereby reducing the error of linearizing the symmetric objective function. Simultaneously, rotation transformation is performed on both the source and target point clouds. The specific mathematical description is as follows:

[0072] ;

[0073] It should be noted that, using the initial positions of the point-pair descriptors found in adjacent point cloud data frames as constraints, the approximate... Normal vectors corresponding to two points and Considering them as coplanar, the components in the symmetric objective function are then... and Equivalently, it can be further simplified to obtain the registration objective function for adjacent point cloud data frames of aircraft fuel tank panels. :

[0074] ;

[0075] S43, Minimize The essence of this problem is a nonlinear least squares optimization problem. Therefore, the LM (Levenberg-Marquardt) algorithm is used to solve the registration objective function of adjacent point cloud data frames of the aircraft fuel tank wall to obtain the optimal transformation matrix.

[0076] In this embodiment, accurate registration will provide a highly accurate initial position for the subsequent pose graph optimization process, and on this basis, the registration error between all point cloud data frames will be further optimized.

[0077] S5. Construct the optimization objective function and point cloud frame pose graph. Calculate the priority of loops in the point cloud frame pose graph based on the overlapping area of ​​adjacent point cloud data frames. Optimize the point cloud data frames by constructing a set of loop sequences with different priorities to generate the final measured point cloud model of the aircraft's overall fuel tank.

[0078] As a preferred embodiment of step S5, the specific process includes the following steps:

[0079] S51. Minimize the Euclidean distance between point cloud data frames to construct an optimization objective function to further optimize the optimal transformation matrix obtained in step S4:

[0080] ;

[0081] ;

[0082] in, Represents a set of point cloud data frames The optimal transformation matrix corresponding to the i-th point cloud frame in the data can transform the point cloud data frame to the world coordinate system established with the first frame point cloud as the reference. This represents the set of corresponding point pairs between adjacent point cloud data frames, which can be obtained through nearest neighbor search of overlapping regions; This represents the initial transformation matrix between adjacent point cloud data frames, which can be obtained through conventional point cloud registration methods, such as Iterative Closest Point.

[0083] S52. Construct a corresponding point cloud frame pose graph based on the point cloud data frame. According to the priority of loop closures in the point cloud frame pose graph, perform loop closure optimization on the point cloud data frame. The process is as follows: Figure 3 As shown;

[0084] As a preferred embodiment of step S52, the specific process includes the following steps:

[0085] S521. The overlap relationship between adjacent point cloud data frames is obtained by using the nearest neighbor search method. Point cloud data frames are used as nodes and the overlapping areas between adjacent point cloud data frames are used as edges to construct a point cloud frame pose graph. The size of the overlapping area between point cloud data frames directly determines the priority of the registration order. The larger the overlapping area, the greater the weight of its corresponding edge in the point cloud frame pose graph, and vice versa.

[0086] S522. Loop detection needs to consider three factors: First, prioritize loops with the largest overlapping area, i.e., loops with the largest edge weights; second, the edges forming loops should ensure the shortest path; third, redundant loops should be minimized. Based on these considerations, a maximum spanning tree (MPL) method is used to detect the shortest loops formed by edges in the point cloud frame pose graph. First, the MPL is extracted based on maximizing the sum of edge weights. The remaining edges are placed in a candidate sequence. Then, the edge with the largest weight is selected from the candidate sequence and added to the MPL. The shortest loop in the current point cloud frame pose graph is recalculated, denoted as... Next, edges from the candidate sequence are added to the maximum spanning tree in order of their weights, and the shortest cycle is updated, ultimately resulting in a set of cycle sequences. ;

[0087] S523, Calculate the set of cyclic sequences The average registration error between point cloud data frames within a single loop is used as the loop closure sequence set. Priority evaluation indicators:

[0088] ;

[0089] in and They are loops The set of adjacent point cloud data frames in the data. The number of point cloud data frames in the set indicates that the smaller the average registration error between point cloud frames, the higher the priority of the loop.

[0090] S524. Evaluate the set of loop sequences according to priority evaluation indicators. Priority determination is performed on the point cloud data frames in the data, and the following is selected: Highest priority point cloud data frame Using the optimization objective function constructed by S51, closure loop optimization is performed. The optimized point cloud data frames are then merged into a single point cloud data set, denoted as... ;

[0091] S525, will As a single node, it reconstructs the point cloud frame pose graph with the remaining nodes. Then, using traditional point cloud registration algorithms (e.g., ICP, point-to-plane ICP), it calculates and updates the correspondence between adjacent point cloud data frames in the current point cloud frame pose graph.

[0092] S526. Repeat steps S522 to S525 until all loops formed under all priorities are traversed, and the optimization is complete.

[0093] S53. Based on the optimization results of step S52, the point cloud data frame of the point cloud frame pose diagram formed after the loop ends is optimized again for global error to further eliminate the cumulative error and generate the final measured point cloud model of the aircraft's overall fuel tank.

[0094] In this embodiment, under the premise that the scanning viewpoint fully ensures sufficient overlap between adjacent point cloud data frames, a loop closure detection method based on maximum spanning tree is used on the pose map of the point cloud frames composed of point cloud data frames. This effectively utilizes the overlapping areas between adjacent point cloud data frames to optimize registration errors. Simultaneously, by determining the optimization order of loop closures, the optimal point cloud data frame registration order is ensured as input, thereby obtaining high-precision optimization results. Furthermore, a hierarchical optimization strategy is used to optimize local loop closures formed by the loop closure sequence, effectively reducing the possibility of the registration results of adjacent point cloud data frames getting trapped in local optima.

[0095] The following describes an aircraft integral fuel tank model reconstruction device based on a progressive registration strategy provided by an embodiment of the present invention. The aircraft integral fuel tank model reconstruction device based on a progressive registration strategy described below can be referred to in correspondence with the aircraft integral fuel tank model reconstruction device based on a progressive registration strategy described above.

[0096] In this embodiment of the invention, the aircraft overall fuel tank model reconstruction device based on the progressive registration strategy includes a target tooling and a three-dimensional scanning module.

[0097] The internal and external parameters of the 3D scanning module are calibrated. The QR code is pasted onto the reserved position on the target fixture. Then, the target fixture is fixed to the positioning hole designed on the aircraft fuel tank wall panel. The position of the 3D scanning module is adjusted to ensure that the QR code is within the field of view of the 3D scanning module.

[0098] like Figure 2 As shown, the 3D scanning module includes a visual odometer, a surface structured light 3D scanner, an adapter, and a flange. The visual odometer includes a binocular camera, and the surface structured light 3D scanner includes a structured light projection device and a monocular grayscale camera.

[0099] The binocular camera and the surface structured light 3D scanner are fixed together by an E-shaped adapter. The adapter is equipped with a flange at the bottom, which can be fixed to different actuators or the end of a tripod.

[0100] The intrinsic parameters of the binocular camera are calibrated, and the extrinsic parameters of the binocular camera and the surface structured light 3D scanner are calibrated simultaneously using a dual-camera calibration method.

[0101] The target fixture structure includes a square target platform, a square magnetic suction plate, bolts, a single-degree-of-freedom rotating shaft, a rectangular frame, hinges, clamping plates, a rubber suction cup, a Z-shaped spring, and a rubber magnetic strip. A single-degree-of-freedom rotating shaft is installed in the middle of the rectangular frame, and the two sides of the middle section of the rectangular frame are removed to ensure that the single-degree-of-freedom rotating shaft can rotate 180 degrees along the axis of the rectangular frame. This allows for adjustment of the square target platform at different clamping and suction positions, ensuring that the square coded target piece is as close as possible to the field of view of the 3D scanning equipment. The rectangular frame of the device is physically connected by multiple hinges and clamping plates on both sides. A Z-shaped spring is fixed along the inner edge of the clamping plate, and a rubber magnetic strip is fixed to the other end of the Z-shaped spring. The square coded target piece is pasted onto the surface of the square target platform, and circular coded target points are pasted onto the surfaces of the rectangular frame and clamping plates. When pasting the circular coded target points, it is important to ensure that no three circular coded target points are on the same straight line, i.e., every three circular coded target points can independently form a triangle.

[0102] The target fixture has square coded target pieces and circular coded target points affixed to it, and an RFID tag is built into it to enable tracking of the target fixture deployed inside the aircraft fuel tank. When affixing the circular coded target points, try to ensure that no three circular coded target points are on the same straight line, that is, every three circular coded target points can independently form a triangle. As shown, the RFID tag is affixed to the bottom surface of the square magnetic piece, and the square magnetic piece is fixed to the bottom surface of the square target platform by bolts to enable tracking of a single target fixture.

[0103] The target fixture is fixed to the surface of different types of reinforcing rib structures inside the aircraft fuel tank by clamping. The reinforcing rib structure of the aircraft fuel tank is generally in the range of 2.0~3.5 mm. Therefore, the target fixture can be clamped on the reinforcing rib structure by mutual attraction of magnetic strips fixed on both sides of the clamping plate. The friction generated by the rubber magnetic strips and the surface of the reinforcing rib structure can ensure the stability of the target fixture. At the same time, the clamping plate can be unfolded horizontally, and the target fixture can be attached to the side wall or top wall of the aircraft fuel tank by using the rubber suction cups fixed to the bottom of the rectangular frame.

[0104] When fixing the target fixture to the aircraft fuel tank wall, the deployment density of the target fixture can be adjusted according to the different field of view volume range of the 3D scanning module. This ensures that the 3D scanning module can cover 2-3 target fixtures within the effective field of view, thereby ensuring the accuracy of positioning the 3D scanning module.

[0105] In this embodiment, by using standardized target fixtures to deploy target sheets and target points, the problems of low target deployment efficiency and target omission in traditional 3D laser measurement operations are overcome. At the same time, mechanical damage to the surface of the fuel tank or contamination such as the introduction of colloidal deposits are avoided. Furthermore, by embedding RFID electronic tags in the target fixtures, the large number of target fixtures deployed inside the aircraft fuel tanks can be tracked, ensuring that no targets are left behind after the 3D laser measurement operation is completed.

[0106] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of this application. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of those different embodiments or examples.

[0107] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus or device (such as a computer-based system, a processor-included system or other system that can fetch and execute instructions from, an instruction execution system, apparatus or device).

[0108] The above embodiments provide a detailed description of the present invention. Specific examples have been used to illustrate the principles and implementation methods of the present invention. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of the present invention. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of the present invention. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A method for aircraft tank model reconstruction based on a progressive registration strategy, characterized in that, The method comprises the following steps: S1, inner and outer parameter calibration is performed on the three-dimensional scanning module, a two-dimensional code is pasted to a reserved position on a target tool, the target tool is fixed to a designed positioning hole on the aircraft fuel tank panel, the position of the three-dimensional scanning module is adjusted, and it is ensured that the two-dimensional code is in the field of view volume of the three-dimensional scanning module; S2, the visual odometry of the three-dimensional scanning module is initialized, then the three-dimensional scanning module is used to collect three-dimensional point cloud data, and the scanning is ended when the entire aircraft fuel tank panel is traversed, point cloud data frames and spatial poses are obtained, and the point cloud data frames are aligned based on the spatial poses; S3、based on the point-to-feature descriptor, implement initial registration between adjacent point cloud data frames to obtain an initial transformation matrix, the initial transformation matrix including an initial rotation matrix and an initial translation matrix ; S4, an accurate registration method based on a symmetric target function is designed, a registration target function of adjacent point cloud data frames of the aircraft fuel tank panel is constructed, an initial transformation matrix is optimized, and an optimal transformation matrix is obtained; Specifically, the method comprises the following steps: S41, based on a point-to-plane ICP method, the initial transformation matrix is taken as an optimization target, and a symmetric target function is constructed: ； wherein, and are normal vectors of corresponding points of adjacent point cloud data frames, respectively; and are normal vectors of corresponding points of adjacent point cloud data frames, respectively; S42, the initial rotation matrix in the above formula Split into two components , , respectively corresponding to the source point cloud data frame and the target point cloud data frame, and simplifying the above formula, the registration objective function of the adjacent point cloud data frames of the aircraft fuel tank wall plate is obtained: ； S43, the L-M algorithm is used to solve the registration target function of the adjacent point cloud data frames of the aircraft fuel tank panel, and the optimal transformation matrix is obtained; S5, an optimization target function and a point cloud frame pose graph are constructed, the priority of a loop in the point cloud frame pose graph is calculated based on an overlapping area of adjacent point cloud data frames, the point cloud data frames are optimized through loop sequence sets of different priorities, and a final aircraft overall fuel tank measured point cloud model is generated; specifically, the method comprises the following steps: S51, the Euclidean distance between the point cloud data frames is minimized, and an optimization target function is constructed to further optimize the optimal transformation matrix obtained in step S4: ； ； wherein, represents the optimal transformation matrix corresponding to the i-th frame of point cloud frames in a set of point cloud data frames, represents a set of corresponding point pairs between adjacent point cloud data frames, which can be obtained by an overlap region nearest neighbor search; represents an initial transformation matrix between adjacent point cloud data frames, which can be obtained by a conventional point cloud registration method;​ S52, a corresponding point cloud frame pose graph is constructed based on the point cloud data frames, and the point cloud data frames are cyclically closed and optimized according to the priority of the loop in the point cloud frame pose graph; S53, based on the optimization result of step S52, the point cloud data frames of the point cloud frame pose graph formed after the cycle is ended are subjected to global error optimization again, so that the cumulative error is further eliminated, and a final aircraft overall fuel tank measured point cloud model is generated.

2. The method of claim 1, wherein: In step S3, the point pair feature descriptor, mainly composed of point pairs of adjacent point cloud data frames and corresponding normal vectors is mathematically described as follows: ； wherein, represents the midpoint of the pair of points , i.e. , the normal vector of represents the angle between two vectors, which takes a value in the range , and represent the normal vector of and the angle formed by the midpoint normal vector of , represents the angle of the normal vector of .

3. The method of claim 1, wherein: The specific process of step S52 comprises the following steps: S521, the point cloud data frames are taken as nodes, and the overlapping areas between adjacent point cloud data frames are taken as edges, a point cloud frame pose graph is constructed, and the greater the overlapping area, the greater the weight of the corresponding edge in the point cloud frame pose graph; S522, a method based on the maximum spanning tree is used to detect the shortest loop composed of edges in the point cloud frame pose graph. First, the maximum spanning tree is extracted according to the maximum sum of edge weights, and the remaining edges will be put into the candidate sequence. Then, the edge with the maximum weight is selected from the candidate sequence and added to the maximum spanning tree, and the shortest loop of the current point cloud frame pose graph is recalculated, denoted as . Then, in order of edge weight size, the edges in the candidate sequence are sequentially added to the maximum spanning tree, and the shortest loop is updated. Finally, the loop sequence set is obtained. S523, calculate the average registration error between the single loop-in point cloud data frames in the loop sequence set as the priority evaluation index of the loop sequence set ​ ； wherein and are the shortest loop a set of adjacent point cloud data frames in the loop, is the number of point cloud data frames in the set. S524, priority determination is performed on the point cloud data frames in the loop sequence set according to the priority evaluation index, and the point cloud data frame with the highest priority in the loop sequence set is selected The optimized target function constructed by S51 is used for loop closure optimization, and the optimized point cloud data frames are combined into an integrated point cloud data, denoted as ;​​ S525、reconstitute the point cloud frame pose graph with the remaining nodes as a single node, and then calculate and update the correspondence relationship of the adjacent point cloud data frames in the current point cloud frame pose graph by using a traditional adjacent point cloud registration algorithm; As a single node, reconstitute the point cloud frame pose graph with the remaining nodes, and then calculate and update the correspondence relationship of the adjacent point cloud data frames in the current point cloud frame pose graph by using a traditional adjacent point cloud registration algorithm; S526, the process of steps S522 to S525 is repeated until all loops formed under all priorities are optimized.

4. An apparatus for model reconstruction of a fuel tank of an aircraft based on a progressive registration strategy, for applying a method for model reconstruction of a fuel tank of an aircraft based on a progressive registration strategy according to any one of claims 1 to 3, characterized in that: The target tool and the three-dimensional scanning module are included. Inner and outer parameter calibration is performed on the three-dimensional scanning module, a two-dimensional code is pasted to a reserved position on a target tool, the target tool is fixed to a designed positioning hole on the aircraft fuel tank panel, the position of the three-dimensional scanning module is adjusted, and it is ensured that the two-dimensional code is in the field of view volume of the three-dimensional scanning module.

5. The apparatus for reconstruction of an aircraft integral tank model based on a progressive registration strategy as claimed in claim 4, wherein: The three-dimensional scanning module comprises a visual odometry, a planar structured light three-dimensional scanner, an adapter and a flange, the visual odometry comprises a binocular camera, and the planar structured light three-dimensional scanner comprises a structured light projection device and a monocular grayscale camera; The binocular camera and the surface structured light three-dimensional scanner are fixed through an E-shaped adapter, and a flange is arranged at the lower part of the adapter to be fixed to different actuators or the end of a tripod; The binocular camera is internally calibrated, and the binocular camera and the surface structured light three-dimensional scanner are externally calibrated simultaneously by using a double-camera calibration method.

6. The apparatus for aircraft integral tank model reconstruction based on progressive registration strategy according to claim 4, characterized in that: The target tool is pasted with a square code target sheet and a circular code target point, and is internally provided with an RFID label to realize tracking of the target tool arranged in the aircraft fuel tank.

7. The apparatus for reconstruction of an aircraft integral tank model based on a progressive registration strategy as claimed in claim 4, wherein: When the target tool is fixed to the wall plate of the aircraft fuel tank, the deployment density of the target tool can be adjusted according to different field-of-view volume ranges of the three-dimensional scanning module, that is, the three-dimensional scanning module can cover 2-3 target tools in the effective field-of-view to ensure the positioning accuracy of the three-dimensional scanning module.

Images