Collaborative scanning method, device, medium and product for an aeroengine

Abstract

The application discloses an aero-engine cooperative scanning method and device, medium and product. The method comprises the following steps: controlling a continuous detection mechanical arm to perform rough scanning on an aero-engine to obtain a rough point cloud model; forming a plurality of candidate viewpoints according to the rough point cloud model; grouping the candidate viewpoints to obtain a plurality of viewpoint groups and forming a joint frustum visibility matrix; obtaining a target viewpoint group sequence according to the joint frustum visibility matrix; adjusting a bending traction rope and / or a rotating traction rope of the continuous detection mechanical arm to control each industrial camera of the mechanical arm to be positioned at each target viewpoint of each target viewpoint group to perform high-resolution shooting, and obtaining a local fine point cloud; and obtaining a high-precision three-dimensional entity model according to the rough point cloud model and the local fine point cloud. The embodiment of the application fundamentally avoids mechanical interference collision damage, realizes full coverage of a complex curved surface space without dead angle, and improves the three-dimensional reconstruction precision and detection efficiency of the aero-engine.

Description

Technical Field

[0001] This invention relates to the fields of aerospace intelligent manufacturing, computer vision, robot coverage path planning and 3D reconstruction technology, and in particular to a collaborative scanning method, equipment, medium and product for aero-engines. Background Technology

[0002] As the core power component of aircraft, aero-engines are characterized by highly complex and unstructured external morphology. Obtaining a complete and high-precision three-dimensional digital model of the external morphology is of vital importance for aero-engine fault detection (such as lightning strikes and foreign object damage), maintenance simulation, assembly verification, and full life cycle management.

[0003] In related technologies, the three-dimensional reconstruction of the external shape of aero engines is mainly achieved through manual handheld 3D scanning, or through automated scanning methods using fixed scanning equipment or traditional industrial rigid robotic arms equipped with a single camera.

[0004] Manual scanning relies heavily on operator experience, resulting in high labor intensity and low efficiency. Furthermore, it is prone to human error and missed areas in confined spaces, making it difficult to guarantee the overall integrity of the scan data. Automated scanning methods, lacking intelligent obstacle avoidance and path planning capabilities for the complex curved surface features of aero-engines, are highly susceptible to mechanical collisions between the robotic arm and the engine body or surrounding wing structures during operation. Additionally, limited by the rigid structure's freedom of motion, single-viewpoint scanning suffers from severe occlusion, failing to acquire multi-angle detailed information, ultimately leading to holes or insufficient accuracy in the reconstructed 3D model. Summary of the Invention

[0005] This invention provides a collaborative scanning method, device, medium, and product for aero-engines, which automatically plans a collision-free scanning path in the specific scenario of aero-engine modeling, realizes collaborative multi-viewpoint acquisition, and efficiently completes high-precision point cloud fusion automated detection.

[0006] According to one aspect of the present invention, a cooperative scanning method for aero-engines is provided, the method comprising: The continuous inspection robotic arm is controlled to perform a coarse scan of the aero-engine under test at a preset safe distance, and a rough point cloud model containing the overall outline of the aero-engine is obtained based on the coarse scan results. The continuous inspection robotic arm includes a drive unit and a robotic arm segment. A preset number of inspection components are set in the axial direction of the robotic arm segment. Each inspection component is independently equipped with a high-resolution industrial camera. The inspection components are connected in series through a central flexible skeleton. Multiple curved traction ropes are set around the periphery of the central flexible skeleton. The curved traction ropes are sequentially strung on each inspection component. A rotating traction rope is wrapped around the outer edge of each inspection component. The drive unit controls the bending action of the robotic arm segment by driving the curved traction ropes, and controls the rotation action of each inspection component around the axis by driving the rotating traction ropes. Based on the geodesic distance field matched with the coarse point cloud model, a set of geodesic paths covering the surface of the aero-engine is generated, and multiple candidate viewpoints are formed based on the set of geodesic paths; each viewpoint includes a spatial position and a normal vector direction. With the constraints of coverage complementarity, spatial proximity and normal attitude consistency, multiple candidate viewpoints are clustered and grouped to obtain multiple viewpoint groups, each containing a preset number of viewpoints. The industrial cameras on each detection component of the continuous inspection robot are grouped into multiple viewpoints. Based on the shooting parameters of each industrial camera, the visibility of each triangular facet in the coarse point cloud model is calculated when the multiple viewpoint group is set on each viewpoint group, forming a joint viewpoint visibility matrix. Based on the motion parameters of the continuous detection robot arm and the joint view frustum visibility matrix, and with coverage and energy consumption path cost as constraints, a preset genetic algorithm is used to iteratively obtain the target viewpoint group sequence, wherein the target viewpoint group sequence includes multiple target viewpoint groups arranged in sequence. By adjusting the bending and / or rotating traction ropes of the continuous inspection robot arm, the industrial cameras of the continuous inspection robot arm are positioned at the target viewpoints of each target viewpoint group and high-resolution images are taken to obtain multiple sets of local fine point clouds that match the aero-engine. By using a coarse point cloud model as a fixed rigid skeleton constraint, and performing global anchoring and registration on multiple sets of local fine point clouds, a high-precision three-dimensional solid model of the aero-engine is obtained.

[0007] According to another aspect of the present invention, a cooperative scanning device for aero-engines is provided, the device comprising: The coarse point cloud model acquisition module is used to control the continuous inspection robotic arm to perform a coarse scan of the aero-engine under test at a preset safe distance, and obtain a coarse point cloud model containing the overall outline of the aero-engine based on the coarse scan results. The continuous inspection robotic arm includes a drive unit and a robotic arm segment. A preset number of inspection components are set in the axial direction of the robotic arm segment. Each inspection component is independently equipped with a high-resolution industrial camera. The inspection components are connected in series through a central flexible skeleton. Multiple curved traction ropes are set around the periphery of the central flexible skeleton. The curved traction ropes are sequentially strung on each inspection component. A rotating traction rope is wrapped around the outer edge of each inspection component. The drive unit controls the bending action of the robotic arm segment by driving the curved traction ropes, and controls the rotation action of each inspection component around the axis by driving the rotating traction ropes. The candidate viewpoint formation module is used to generate a set of geodesic paths covering the surface of the aero-engine based on the geodesic distance field matched with the coarse point cloud model, and to form multiple candidate viewpoints based on the set of geodesic paths; wherein each viewpoint includes a spatial position and a normal vector direction; The viewpoint group acquisition module is used to cluster and group multiple candidate viewpoints under the constraints of coverage complementarity, spatial proximity and normal attitude consistency, to obtain multiple viewpoint groups, each containing a preset number of viewpoints. The visibility matrix forming module is used to group the industrial cameras on each detection component of the continuous detection robot as multi-viewpoint groups. Based on the shooting parameters of each industrial camera, it calculates the visibility of each triangular facet in the coarse point cloud model when the multi-viewpoint group is set on each viewpoint group, and forms a joint frustum visibility matrix. The target viewpoint group sequence acquisition module is used to obtain the target viewpoint group sequence iteratively using a preset genetic algorithm based on the motion parameters of the continuous detection robot arm and the joint viewpoint cone visibility matrix, with coverage and energy consumption path cost as constraints. The target viewpoint group sequence includes multiple target viewpoint groups arranged in sequence. The local fine point cloud imaging module is used to control each industrial camera of the continuous inspection robot arm to be positioned at each target viewpoint of each target viewpoint group by adjusting the bending traction rope and / or rotating traction rope of the continuous inspection robot arm, and to take high-resolution pictures to obtain multiple sets of local fine point clouds that match the aero-engine. The high-precision model generation module is used to take the coarse point cloud model as a fixed rigid skeleton constraint and perform global anchoring and registration on multiple sets of local fine point clouds to obtain a high-precision three-dimensional solid model of the aero-engine.

[0008] According to another aspect of the present invention, an electronic device is provided, the electronic device comprising: At least one processor; and A memory communicatively connected to the at least one processor; wherein, The memory stores a computer program that can be executed by the at least one processor, which enables the at least one processor to perform the cooperative scanning method for aero-engines as described in any embodiment of the present invention.

[0009] According to another aspect of the present invention, a computer-readable storage medium is provided, the computer-readable storage medium storing computer instructions for causing a processor to execute and implement the cooperative scanning method for aero-engines as described in any embodiment of the present invention.

[0010] According to another aspect of the present invention, a computer program product is also provided, including a computer program that, when executed by a processor, implements the steps of the cooperative scanning method for aero-engines as described in any embodiment of the present invention.

[0011] The technical solution of this invention applies a continuous inspection robotic arm to the field of 3D modeling of aero-engines. By employing a fully automated two-stage scanning strategy, an obstacle avoidance envelope for collision detection is pre-constructed based on a coarse point cloud model, fundamentally and effectively avoiding mechanical collision interference between the robotic arm and the aero-engine or surrounding structures. This significantly shortens the planning and scanning cycle and ensures the operational safety of high-value aero-engine equipment. An innovative viewpoint planning algorithm based on geodesics and a joint view frustum visibility matrix is ​​proposed. Combining the adaptive bending capability of the continuous inspection robotic arm with the active rotation mechanism of multiple industrial cameras around their axes, and controlling the field of view overlap rate through a precise geometric model, the scanning coverage of the complex external topography of the aero-engine is maximized, effectively eliminating blind spots and obstructions. A point cloud fusion strategy of "coarse scanning global anchor points as the skeleton + fine scanning local precise registration" effectively eliminates the cumulative positioning errors caused by the long-distance movement of the multi-view robotic arm, ensuring the high fidelity and global geometric consistency of the final generated 3D solid model.

[0012] It should be understood that the description in this section is not intended to identify key or essential features of the embodiments of the present invention, nor is it intended to limit the scope of the invention. Other features of the invention will become readily apparent from the following description. Attached Figure Description

[0013] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0014] Figure 1This is a flowchart of a cooperative scanning method for aero-engines provided according to an embodiment of the present invention; Figure 2 This is a schematic diagram of the structure of a continuous detection robotic arm applicable to an embodiment of the present invention; Figure 3 This is a principle block diagram of a multi-set local fine point cloud to global coarse point cloud model registration and fusion applicable to the embodiments of the present invention; Figure 4 This is a schematic diagram illustrating the construction process of a coarse scanning and safe obstacle avoidance envelope surface applicable to an embodiment of the present invention; Figure 5 This is a schematic diagram of a coarse point cloud model applicable to an embodiment of the present invention; Figure 6 This is a schematic diagram of a high-precision three-dimensional solid model applicable to embodiments of the present invention; Figure 7 This is a structural diagram of a cooperative scanning device for aero-engines according to an embodiment of the present invention; Figure 8 This is a schematic diagram of the structure of an electronic device that implements the cooperative scanning method for aero-engines according to an embodiment of the present invention. Detailed Implementation

[0015] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0016] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0017] Figure 1This is a flowchart illustrating a collaborative scanning method for aero-engines provided in an embodiment of the present invention. This embodiment is applicable to situations where aero-engines are safely, accurately, and efficiently modeled in 3D using a continuous inspection robotic arm. The method can be executed by a collaborative scanning device for aero-engines, which can be implemented in hardware and / or software and is generally configured in a computer device with data processing capabilities. Figure 1 As shown, the method includes: S110: Control the continuous inspection robotic arm to perform a coarse scan of the aero-engine under test at a preset safe distance, and obtain a rough point cloud model containing the overall outline of the aero-engine based on the coarse scan results.

[0018] The continuous inspection robotic arm includes a drive unit and a robotic arm segment. A preset number of inspection components are arranged in the axial direction of the robotic arm segment. Each inspection component is independently equipped with a high-resolution industrial camera. The inspection components are connected in series through a central flexible skeleton. Multiple curved traction ropes are arranged around the periphery of the central flexible skeleton. The curved traction ropes are sequentially strung on each inspection component. A rotating traction rope is wrapped around the outer edge of each inspection component. The drive unit controls the bending motion of the robotic arm segment by driving the curved traction ropes, and controls the rotation motion of each inspection component around the axis by driving the rotating traction ropes.

[0019] Furthermore, in Figure 2 The diagram shows a structural schematic of a continuous detection robotic arm applicable to an embodiment of the present invention. Figure 2 Sub-figure (1) shows the overall structure of the robotic arm segment. Figure 2 Subfigure (2) shows a magnified view of the detection component, and subfigure (3) shows a schematic diagram of the field of view (or frustum) model when the image is acquired by the high-resolution industrial camera in the detection component.

[0020] By using a drive unit to flexibly control the bending and rotating traction ropes in the robotic arm segments, the continuous inspection robotic arm can adapt flexibly to various confined inspection spaces and large-area curved surface inspections. In fact, this invention utilizes the characteristics of existing continuous inspection robotic arms—their flexible movement and multiple industrial cameras—to creatively propose a novel multi-viewpoint collaborative scanning modeling scheme for the external morphology of aero-engines.

[0021] In this embodiment, the continuous inspection robotic arm is first moved to a position absolutely safe from the aircraft engine under test, for example, a safe envelope distance of 1.5 meters. The industrial cameras mounted on the robotic arm segments, typically RGB-D depth cameras, are then controlled to rotate rapidly around the aircraft engine axis at a preset angular velocity (e.g., 0.5 rad / s). The industrial cameras acquire panoramic depth images at a preset image acquisition rate. The acquired depth map sequence is converted into a 3D point cloud in real time. After filtering through a voxel mesh of a set voxel size, an initial coarse point cloud model containing the overall outline of the aircraft engine is obtained.

[0022] Based on the above embodiments, after obtaining the initial coarse point cloud model, the initial coarse point cloud model can first be filtered and denoised. For example, based on a preset number of neighborhood points (e.g., 50) and a standard deviation factor (e.g., 1.0), statistical filtering can be performed on the initial coarse point cloud model to remove outlier noise points. Subsequently, voxel downsampling based on a set voxel size is performed to compress the data volume and improve subsequent computational efficiency. Accordingly, the initial coarse point cloud model after the above filtering and downsampling processing can be determined as the coarse point cloud model.

[0023] To facilitate subsequent calculations, we can first calculate the normal vector and curvature of each point in the coarse point cloud model. Specifically, principal component analysis can be used to calculate the curvature and normal vector of each point. For example, suppose there are k preset point cloud points in the local neighborhood of each point, then the mean value of the center point is: ,in, Let be the i-th point in a set of k point cloud points. Its covariance matrix C is defined as: Perform eigenvalue decomposition on C. The eigenvalues ​​can be obtained. and its corresponding eigenvectors . , , Among them, the smallest eigenvalue Corresponding feature vector This is the normal vector of the point in the point cloud. The surface curvature of this point in the point cloud... It can be approximated as: .

[0024] Furthermore, a safe space can be constructed for the aero-engine. Specifically, the characteristics of interference obstacles such as protruding hanging structures on the exterior of the aero-engine can be used as a constraint base map, and a three-dimensional morphological expansion operation can be used to construct a safe operating space. Specifically, a large safe movement space can be pre-defined based on the geographical location of the aero-engine. Then, each point in the coarse point cloud model is extended outward along the normal vector direction by a predetermined distance (e.g., 1 meter) as a safe obstacle avoidance envelope to ensure that the robotic arm never makes physical contact with the aero-engine surface during operation. For example, let the set of point cloud surfaces of the coarse point cloud model be... If the safety expansion radius is r, then the safety obstacle avoidance envelope is... The distance field can be expressed as: , where p is One point in the equation, q is One point in the middle, It represents three-dimensional Euclidean space.

[0025] After obtaining the safe obstacle avoidance envelope, the difference between the safe movement space and the safe obstacle avoidance envelope can be calculated as the safe motion corridor, which is the safe operating space of the continuous detection robot arm. After obtaining this safe operating space, a three-dimensional safe motion corridor constraint vector can be generated. This three-dimensional safe motion corridor constraint vector can be specifically a row vector. Each column in this row vector represents a point in the safe movement space, and the element value of each column in the three-dimensional safe motion corridor constraint vector represents whether each point in the safe movement space is passable; for example, passable is 1, and impassable is 0. The above-mentioned three-dimensional safe motion corridor constraint vector can be used to further screen the safety of the finally calculated target viewpoint group sequence.

[0026] S120. Based on the geodesic distance field matched with the coarse point cloud model, generate a set of geodesic paths covering the surface of the aero-engine, and form multiple candidate viewpoints based on the set of geodesic paths; wherein, each viewpoint includes spatial position and normal vector direction.

[0027] After obtaining the coarse point cloud model, the geodesic distance field on the surface of the coarse point cloud model can be calculated based on a preset algorithm (e.g., the Fast Marching algorithm). The geodesic distance field refers to the three-dimensional scalar field in which the shortest path distance (i.e., geodesic distance) from each point cloud point on the coarse point cloud model to the seed point is calculated, starting from a preset seed point, and this distance value is used as a scalar.

[0028] After obtaining the geodesic distance field, two core parameters can be set according to the size of the aero-engine and the scanning accuracy requirements: path interval and path coverage. The path interval describes the surface distance between two adjacent geodesic paths, while the path coverage describes the maximum geodesic distance from the seed point to the end of the aero-engine, ensuring that the path covers the entire surface. Then, starting from the seed point, a series of distance thresholds can be generated according to the path interval, such as single, double, and triple path intervals. Next, all point cloud points with distance values ​​equal to a distance threshold can be selected from the geodesic distance field. By sorting and fitting the selected point cloud points, a continuous closed or semi-closed curve can be obtained, which is a geodesic path. Then, geodesic paths corresponding to all distance thresholds are generated to obtain a set of geodesic paths covering the entire engine surface.

[0029] In an optional embodiment of this invention, the interval and direction of the geodesics can be automatically adjusted based on changes in the curvature of the aero-engine surface (such as differences between the arc segments or straight segments of the casing), so that the viewpoint layout matches the surface shape in real time. Simultaneously, the geodesic parameters are dynamically corrected by combining real-time curvature data of the aero-engine surface, avoiding scanning blind spots caused by abrupt changes in the surface and solving the problem that traditional fixed-path geodesic generation schemes cannot adapt to complex surfaces.

[0030] After obtaining the aforementioned geodesic path set, it can be optimized. For example, abnormal paths near engine edges and holes can be removed. Furthermore, for high-curvature areas (such as the casing transition section), the path spacing can be appropriately reduced to generate denser geodesic paths, ensuring no blind spots in subsequent viewpoint layout. Additionally, the geodesic paths can be smoothed to eliminate path jitter caused by point cloud discrepancies, ensuring path continuity.

[0031] After optimizing the geodesic path set, each point in the geodesic path set can be extended outward along the direction of its normal vector (e.g., by 1-1.5 meters) to form multiple discrete candidate viewpoints. Each candidate viewpoint carries specific three-dimensional coordinates (spatial position) and orientation information (i.e., the direction of the three-dimensional normal vector, or simply the normal vector).

[0032] After obtaining the above multiple candidate viewpoints, each candidate viewpoint can be matched with the three-dimensional safe motion corridor constraint vector to filter out candidate viewpoints that are not located within the three-dimensional safe motion corridor constraint vector, so as to fundamentally avoid mechanical collision interference between the robotic arm and the aircraft engine or surrounding structures.

[0033] S130. Using coverage complementarity, spatial proximity and normal attitude consistency as constraints, multiple candidate viewpoints are clustered and grouped to obtain multiple viewpoint groups, each containing a preset number of viewpoints.

[0034] In an optional implementation of this embodiment, multiple candidate viewpoints are clustered and grouped based on constraints such as coverage complementarity, spatial proximity, and normal pose consistency to obtain multiple viewpoint groups, which may include: S1301. Calculate the single-point coverage score corresponding to each candidate viewpoint, and sort the candidate viewpoints in descending order of single-point coverage score to obtain the candidate viewpoint sequence.

[0035] The single-point coverage score for a candidate viewpoint can be calculated by considering the number of point cloud points it covers in the coarse point cloud model. In other words, the more point cloud points it covers, the higher the single-point coverage score. Specifically, the single-point coverage score for each candidate viewpoint can be calculated by combining the frustum model of a single industrial camera, the position coordinates of each point cloud point in the coarse point cloud model, and information on occlusions around the coarse point cloud model.

[0036] S1302. Sequentially select one candidate viewpoint from the candidate viewpoint sequence as the seed viewpoint, and then select multiple related viewpoints from the remaining candidate viewpoints that simultaneously satisfy the conditions of complementary coverage, spatial distance, and angle between normal vectors.

[0037] In an optional implementation of this embodiment, the coverage complementarity condition can be... ;in, Candidate viewpoints With candidate viewpoints The coverage complementarity index between them The preset coverage complementarity threshold, , Indicate candidate viewpoints The number of point clouds in the coarse point cloud model that can be covered. Indicate candidate viewpoints The number of point clouds in the coarse point cloud model that can be covered. .

[0038] The spatial distance condition is: ;in, Candidate viewpoints With candidate viewpoints The distance between them The lower limit of the preset distance threshold. This is the preset upper limit of the distance threshold.

[0039] The condition for the included angle of the normal vectors is: ;in, Candidate viewpoints With candidate viewpoints The angle between the normal vectors, This is a preset threshold for the angle between the normal vectors.

[0040] Understandably, the more complementary the coverage of the point cloud points in the coarse point cloud model by two candidate viewpoints, the closer their distance, and the more consistent their orientation, the easier it is for the two candidate viewpoints to be classified into the same viewpoint group.

[0041] It is understandable that when calculating the above-mentioned complementary coverage condition, spatial distance condition, and normal vector angle condition, one candidate viewpoint is fixedly selected as the seed viewpoint, and the other candidate viewpoint is obtained by traversing the above-mentioned candidate viewpoint sequence in turn.

[0042] S1303. Based on the condition of uniformity of distance distribution, multiple associated viewpoints are filtered to obtain a preset number minus one target associated viewpoint, and each target associated viewpoint and seed viewpoint are combined into a viewpoint group.

[0043] In this embodiment, the hardware characteristics of the continuous inspection robot were comprehensively considered when dividing the viewpoint groups, ensuring that the number of viewpoints in each viewpoint group is consistent with the number of industrial cameras in the continuous inspection robot. This allows the continuous inspection robot to acquire high-precision images of a single viewpoint group in a single operation.

[0044] To achieve the aforementioned technical effects, the arrangement of viewpoints in a viewpoint group needs to be consistent with the arrangement of industrial cameras in a continuous inspection robot arm. For example, assuming that in the continuous inspection robot arm, the industrial cameras are evenly arranged on the axis of the robot arm, and the distance between them is adjustable between 2 and 4 decimeters. Furthermore, after obtaining all associated viewpoints corresponding to the seed viewpoint, the associated viewpoints can be further filtered according to the uniformity of distance distribution, taking into account the hardware layout characteristics of the industrial cameras, to obtain a preset number - 1 target associated viewpoints. These target associated viewpoints, along with the seed viewpoint, are packaged into a viewpoint group containing the preset number of viewpoints. Simultaneously, the spatial arrangement of the viewpoints in this viewpoint group should be similar to the spatial arrangement of the industrial cameras on the axis of the continuous inspection robot arm.

[0045] S1304. After filtering the generated viewpoint groups from the candidate viewpoint sequence, determine whether the processing of all candidate viewpoints in the candidate viewpoint sequence is complete. If yes, determine that multiple viewpoint groups have been obtained; otherwise, return to execute S1302.

[0046] Through the above iterative processing, multiple viewpoint groups can be obtained based on all candidate viewpoints. It is understandable that the viewpoint groups generated in later iterations will have lower single-point coverage scores for each viewpoint. In order to ensure the efficiency of subsequent local fine point cloud generation, the viewpoint groups obtained in the initialization can be further filtered.

[0047] Accordingly, based on the above embodiments, after obtaining multiple viewpoint groups, the method may further include: Based on the coverage complementarity index, distance, and normal vector angle between adjacent viewpoints in each viewpoint group, calculate the comprehensive quality score corresponding to each viewpoint group, and filter out invalid viewpoint groups whose comprehensive quality scores are less than a preset score threshold.

[0048] In this embodiment, the mean coverage complementarity, mean distance, and mean normal angle of each viewpoint group can be calculated based on the coverage complementarity index, distance, and normal angle between adjacent viewpoints in each viewpoint group. Then, by weighting and summing these three means based on preset weighting coefficients, a comprehensive quality score corresponding to each viewpoint group is calculated. Finally, viewpoint groups with comprehensive quality scores below a preset threshold are considered invalid viewpoint groups and filtered out from the multiple viewpoint groups.

[0049] By setting the above parameters, it can be ensured that the viewpoint groups that are ultimately retained are all viewpoint groups that can effectively perform three-dimensional measurements of the aero-engine.

[0050] S140. The industrial cameras on each detection component of the continuous detection robot arm are grouped into a multi-viewpoint group. Based on the shooting parameters of each industrial camera, the visibility of each triangular facet in the coarse point cloud model is calculated when the multi-viewpoint group is set on each viewpoint group, forming a joint viewpoint visibility matrix.

[0051] To obtain the target viewpoint group sequence required for fine scanning, a joint frustum visibility matrix must first be formed. Each column of the joint frustum visibility matrix represents a triangular facet in the coarse point cloud model, and each row represents a viewpoint group. Furthermore, the elements in the joint frustum visibility matrix, defined by specified rows and columns, represent the visibility of a specified viewpoint group to a specified triangular facet; for example, they can be considered as 1 and cannot be considered as 0.

[0052] In an optional implementation of this embodiment, the industrial cameras on each detection component of the continuous inspection robot are grouped into a multi-viewpoint group. Based on the shooting parameters of each industrial camera, the visibility of each triangular facet in the coarse point cloud model is calculated when the multi-viewpoint group is set on each viewpoint group, forming a joint frustum visibility matrix, which may include: S1401. Divide the coarse point cloud model into facets to obtain triangular facets that match the coarse point cloud model.

[0053] Optionally, an implicit Poisson surface equation can be constructed based on the curvature and normal vector of each point in the coarse point cloud model. Solving this implicit Poisson surface equation yields an implicit surface, which is then transformed into an explicit triangular mesh using a pre-defined algorithm, such as the MarchingCubes algorithm. Finally, the reconstructed triangular mesh is smoothed to eliminate mesh jitter caused by point cloud discretization. The mesh is then re-meshed, and invalid faces near engine edges and holes are marked and removed, ultimately resulting in triangular faces that match the coarse point cloud model.

[0054] S1402. Calculate the field of view volume of a single industrial camera based on the horizontal and vertical field of view angles of the industrial camera.

[0055] In this embodiment, it is assumed that all industrial cameras in the continuous inspection robotic arm have the same camera parameters. Therefore, based on the horizontal field of view H and vertical field of view V of the industrial cameras, the following formula can be used: The field of view volume VO of a single industrial camera is calculated. Where D is the distance to the far plane of the industrial camera, and d is the distance to the near plane of the industrial camera.

[0056] S1403. Sequentially acquire a current viewpoint group, and calculate the joint field of view volume matching the multi-viewpoint group when each industrial camera in the multi-viewpoint group is deployed to each viewpoint in the current viewpoint group according to the field of view volume of a single industrial camera and the shooting parameters of the industrial camera.

[0057] In an optional implementation of this embodiment, calculating the joint field of view volume matching the multi-view group when each industrial camera in the multi-view group is deployed onto each viewpoint in the current viewpoint group, based on the field of view volume of a single industrial camera and the imaging parameters of the industrial camera, may include: When each industrial camera in the multi-viewpoint group is deployed to each viewpoint in the current viewpoint group, the bending diameter R and arc length l between two adjacent industrial cameras are obtained, and the deflection angle θ between two adjacent industrial cameras is calculated according to the formula θ=l / R. According to the formula: The field-of-view overlap rate α between two adjacent industrial cameras was calculated.

[0058] The combined field of view volume of a single industrial camera and the field of view overlap rate between two adjacent industrial cameras are calculated based on the field of view volume of the multi-view group for the current view group.

[0059] Specifically, the bending radius R is the radius of curvature of the arc corresponding to the line connecting the centers of two adjacent industrial cameras after the continuous inspection robot arm bends. It is determined by the flexible skeleton of the robot arm and the tension of the traction rope. The arc length l is the distance between the centers of two adjacent industrial cameras along the curved axis of the flexible robot arm when the continuous inspection robot arm is bent. The deflection angle θ can be understood as the angle between the observation directions of two adjacent industrial cameras.

[0060] It is understandable that when the industrial cameras in the multi-viewpoint group are arranged to viewpoints in different viewpoint groups, the curvature of the continuous detection robot arm is different, and consequently, the calculated field-of-view overlap rate α of any two industrial cameras is also different.

[0061] The field-of-view overlap rate α can be understood as the proportion of the overlapping area between two adjacent industrial cameras to the field of view of a single industrial camera. A larger α indicates greater overlap and redundant coverage between the two cameras; a smaller α indicates less overlap and stronger complementarity, but may introduce blind spots. In an optional implementation of this embodiment, a reasonable range of values ​​for α can be preset. If one or more α values ​​calculated for a certain viewpoint group do not fall within this reasonable range, the viewpoint group can be directly filtered out to further improve the efficiency of subsequent fine-tuning scanning.

[0062] Furthermore, the combined field-of-view volume can be understood as the total effective coverage volume of multiple industrial cameras combined, after deducting the overlapping portion. For example, if the single field-of-view volume of two industrial cameras is W, and the field-of-view overlap rate is α, then the combined field-of-view volume ≈ W + W - α × W = W × (2 - α). The calculation method is similar when there are multiple industrial cameras, and will not be elaborated here.

[0063] S1404. Match and map the joint field of view volume with each triangular facet of the coarse point cloud model to obtain the visibility of the current viewpoint group to each triangular facet, forming the visibility vector of the current viewpoint group.

[0064] The joint field of view volume characterizes the effective observation coverage area of ​​the multi-viewpoint group in three-dimensional space. This continuous, closed effective observation coverage area is spatially mapped and matched with all triangular facets after the rough point cloud model of the aero-engine is meshed. Specifically, each triangular facet is traversed sequentially, and through spatial enclosure detection, view frustum inclusion determination, and occlusion removal verification, it is determined whether a single triangular facet is within the effective observation coverage area and free from structural occlusion. Then, a unified binary visibility identifier is output for each triangular facet, arranged sequentially according to the globally fixed numbering order of the triangular facets, and combined to form a one-dimensional sequence vector, which is the visibility vector corresponding to the current viewpoint group.

[0065] S1405. Determine whether the processing of all viewpoint groups is complete: if yes, execute S1406; otherwise, return to execute S1403.

[0066] S1406. Combine all visibility vectors to obtain the joint view frustum visibility matrix.

[0067] S150. Based on the motion parameters of the continuous detection robot arm and the joint view frustum visibility matrix, and with coverage and energy consumption path cost as constraints, a preset genetic algorithm is used to iteratively obtain the target viewpoint group sequence, wherein the target viewpoint group sequence includes multiple target viewpoint groups arranged in sequence.

[0068] In an optional implementation of this embodiment, based on the motion parameters of the continuous detection robot arm and the joint view frustum visibility matrix, and constrained by coverage and energy path cost, a preset genetic algorithm is used to iteratively obtain the target viewpoint group sequence, which may include: S1501. Calculate the group geometric center and main direction corresponding to each viewpoint group.

[0069] Optionally, let a certain viewpoint group be... , where k is the number of candidate viewpoints within the viewpoint group, i.e., the aforementioned preset number. Furthermore, the weighted average method can be used to calculate the group geometric center of viewpoint group G. The formula is as follows: , , ; where the i-th candidate viewpoint The corresponding spatial coordinates are , For the i-th candidate viewpoint The weight can be determined by the candidate viewpoint. The single-point coverage score determines the weight; for example, the higher the single-point coverage score, the greater the weight, to ensure that the group geometric center is more biased towards candidate viewpoints with strong coverage capabilities, which aligns with the core requirements of aero-engine scanning.

[0070] Of course, the arithmetic mean method can also be used directly, that is, setting all =1, this setting simplifies calculations and does not affect the overall logic, and can be selected according to the actual scanning accuracy requirements.

[0071] The principal direction of a viewpoint group is the "dominant direction" of all normal vectors within that viewpoint group, used to characterize the overall observation attitude of the entire viewpoint group. Continuing with the viewpoint group as... For example, the specific calculation steps for the main direction of the group are as follows: Extract the normal vectors of all candidate viewpoints within the viewpoint group. Construct the normal vector matrix; perform principal component analysis on the normal vector matrix to calculate the covariance matrix of the normal vector matrix; solve for the eigenvalues ​​and eigenvectors of the covariance matrix, and select the eigenvector with the largest eigenvalue as the principal direction of the viewpoint group.

[0072] S1502. Construct a fitness function; wherein, the fitness function includes a coverage calculation term and an energy consumption path cost calculation term. The coverage calculation term is calculated based on the joint view frustum visibility matrix. The energy consumption path cost calculation term includes the movement distance when moving between adjacent viewpoint groups and the joint rotation angle when moving between adjacent viewpoint groups. The movement distance is calculated based on the group geometric center of the viewpoint group, and the joint rotation angle is calculated based on the group principal direction of the viewpoint group and the motion parameters of the continuous detection robot arm.

[0073] In a specific example, it can be constructed as follows: The fitness function F(X) is given, where X is an individual in the genetic algorithm, i.e., a sequence of viewpoints. This is a coverage calculation item; the larger the value, the higher the coverage scan rate. This is the energy consumption path cost calculation item. The smaller this value, the lower the energy consumption of the continuous inspection robot arm. and These are preset weighting coefficients. Specifically, these two weighting coefficients can be adjusted according to actual needs; for example, when scanning the core region of an aero-engine, increase... Prioritize ensuring coverage; when scanning large, flat areas of aero-engines, increase... Prioritize reducing energy consumption.

[0074] Furthermore, based on the joint cone visibility matrix, the ratio of the total number of triangular faces covered by individual X to the total number of triangular faces contained in the aero-engine can be calculated as... .

[0075] Furthermore, the energy consumption path cost is calculated from the travel distance cost and the joint angle cost. The travel distance cost is obtained by weighted summation of the Euclidean distances between the geometric centers of adjacent viewpoint groups in individual X, while the joint angle cost is determined by the angle between the principal directions of adjacent viewpoint groups in individual X and the angle energy consumption coefficient determined by the motion parameters of the continuous detection robot arm.

[0076] S1503. Based on the fitness function, with the goal of maximizing coverage and minimizing energy consumption path cost, a preset genetic algorithm is used to iteratively obtain the target viewpoint group sequence.

[0077] In this embodiment, after setting the key parameters in the genetic algorithm, such as population size, number of iterations, crossover probability, mutation probability and termination condition, the optimal viewpoint group sequence can be iteratively obtained through heuristic combinatorial optimization as the target viewpoint group sequence.

[0078] Based on the above embodiments, after obtaining the target viewpoint group sequence, the discrete pose points in the sequence can be smoothly fitted by B-spline curves to generate a continuous movement trajectory of the robot arm's central axis, so as to control the continuous detection robot arm to move smoothly along the B-spline trajectory.

[0079] S160. By adjusting the bending and / or rotating traction ropes of the continuous inspection robot arm, the industrial cameras of the continuous inspection robot arm are positioned at the target viewpoints of each target viewpoint group and take high-resolution pictures to obtain multiple sets of local fine point clouds that match the aero-engine.

[0080] In an optional implementation of this embodiment, by adjusting the bending and / or rotating traction ropes of the continuous inspection robot arm, each industrial camera of the continuous inspection robot arm is controlled to be positioned at a target viewpoint of each target viewpoint group and to perform high-resolution imaging. This may include: S1601. In the target viewpoint group sequence, obtain a current target viewpoint group in sequence, and according to the spatial position and curvature information of each target viewpoint in the current target viewpoint group, fit the ideal shooting position curve that matches the current target viewpoint group.

[0081] S1602. Obtain the current position trajectory of the central axis of the continuous detection robot arm, and calculate the first type of tension of each curved traction rope in the continuous detection robot arm when controlling the current position trajectory to move towards the ideal shooting position curve based on the current position trajectory and the ideal shooting position curve. Then, perform first type of bending control on the continuous detection robot arm based on the first type of tension.

[0082] S1603. When the central axis of the continuous inspection robot arm is determined to be closely aligned with the ideal shooting position curve, the camera positions of each industrial camera in the continuous inspection robot arm are obtained.

[0083] S1604. Based on the camera positions of each industrial camera and the spatial positions of each target viewpoint in the current target viewpoint group, calculate the second type of tension of each curved traction rope in the continuous detection robot arm when controlling each industrial camera to move toward each target viewpoint, and perform second type of curvature control on the continuous detection robot arm based on the second type of tension.

[0084] S1605. When it is determined that each industrial camera in the continuous inspection robot arm is close to the matching target viewpoint, the rotation amount of each rotating traction rope in the continuous inspection robot arm is calculated according to the normal vector direction of each target viewpoint. When the camera optical axis of each industrial camera moves towards the normal vector direction of each target viewpoint, the rotation control of each inspection component in the continuous inspection robot arm is performed according to the calculated rotation amount.

[0085] Specifically, the spatial curvature of the ideal shooting position curve can be converted into the linear tension parameter of the bending traction rope inside the continuous detection robot arm, controlling the robot arm to achieve flexible adaptive bending. Simultaneously, the end-effector pose of the continuous detection robot arm is acquired in real time, and the surface normal vector of the corresponding region is indexed from the coarse point cloud model. The angle of rotation of multiple industrial cameras around the axis is adaptively matched with the surface normal vector, so that the angle between the optical axis of each industrial camera and the normal of the surface to be measured is kept within a preset threshold. Finally, the normal vector deviation angle at each target viewpoint is converted into the number of motor rotation steps of the rotating traction rope, pulling the detection component to rotate around the axis, and simultaneously triggering the camera array to acquire images and local point clouds.

[0086] In this embodiment, by converting the trajectory curvature command into the tension of the internal traction rope of the continuous inspection robot arm, the flexible joints of the continuous inspection robot arm automatically adjust their curvature, achieving a high-fidelity fit between the arm's contour and the curvature of the aircraft engine's outer surface. When the continuous inspection robot arm reaches the preset viewpoint group position, a set number of industrial cameras synchronously perform fine-tuning rotation around their respective mounting axes based on real-time feedback of local surface normal vectors. The rotation compensation angle is obtained by comparing the surface normal vector of the current local coarse point cloud in real time. The closed-loop control keeps the angle between the camera's optical axis and the normal of the surface under test strictly within ±10°, thereby eliminating visual blind spots caused by narrow gaps from a physical perspective. After reaching the position, each industrial camera acquires high-resolution images at a rate of 10 frames per second, generating a high-density local fine point cloud with a point spacing of approximately 0.2 mm.

[0087] In an optional embodiment of this invention, three types of sensors can be embedded in each joint of the continuous inspection robot arm, the inspection component (camera mounting location), and the traction rope connection point: 1. A high-precision pressure sensor, used to collect the traction rope tension and the contact pressure between the continuous inspection robot arm and the surface of the aero-engine in real time; 2. A laser distance sensor, used to collect the vertical distance between the end of the continuous inspection robot arm and the surface of the aero-engine in real time; 3. An attitude sensor, used to collect the joint rotation angle of the continuous inspection robot arm and the camera mounting attitude in real time, and to provide real-time feedback of the offset. Afterwards, all sensor data needs to be fused to remove abnormal data (such as sensor false alarms or mechanical vibration interference), generating a real-time state matrix of the continuous inspection robot arm to provide accurate data support for subsequent traction control.

[0088] In another optional embodiment of this example, the curvature gradient of the triangular facets on the surface of the aero-engine can also be extracted. For high curvature areas (such as blade edges or arc transition sections of the casing), the tension of the corresponding traction rope is automatically increased (for example, by 10%-20%) to ensure that the continuous detection robotic arm fits the curved surface and avoids camera offset due to abrupt changes in the curved surface. For low curvature areas (such as straight sections of the engine), the tension of the traction rope is appropriately reduced to reduce energy consumption.

[0089] Furthermore, a closed-loop adjustment of the traction rope tension can be introduced. This involves real-time collection of tension data from each traction rope and comparison with preset thresholds. If a tension deviation occurs (e.g., a traction rope becomes loose or has insufficient tension), the speed of the corresponding traction rope's drive motor is automatically adjusted to ensure balanced tension across all ropes. This prevents the robotic arm from shifting or the camera from positioning due to excessive tension in a single rope. Additionally, preset traction rope tension thresholds can be set for different parts of the aero-engine (e.g., high-temperature or easily worn areas). In high-temperature areas, the tension can be appropriately reduced to prevent tension attenuation caused by rope softening at high temperatures, thus improving scanning stability.

[0090] Furthermore, a robotic arm attitude compensation model can be constructed by combining the real-time curvature changes of the aero-engine surface to solve the problem of traditional robotic arms having fixed attitudes and being unable to adapt to abrupt changes in the surface. Specifically, the rotation angle data of each joint of the continuous detection robotic arm can be collected in real time, and combined with the curvature gradient of the aero-engine surface, the motion trajectory of the continuous detection robotic arm can be predicted, and the joint angles can be adjusted in advance to avoid collisions between the robotic arm and the engine surface. This is especially important for the concave and convex parts of the complex curved surface of the aero-engine, where the robotic arm attitude can be adjusted in advance. In addition, an attitude error compensation algorithm can be introduced. When the attitude deviation occurs during the movement of the continuous detection robotic arm (such as the deviation between the camera optical axis and the normal vector of the aero-engine surface exceeding 5°), attitude correction is automatically initiated. By adjusting the tension of the traction rope (with a single-sided tension ≤ 0.5mm), the precise matching between the robotic arm attitude and the surface normal vector is achieved, ensuring that the camera shooting angle is always optimal.

[0091] Based on the above embodiments, three traction control modes can be set according to different scanning requirements of aero-engines: Mode 1 (high-precision mode): for the core area of ​​the aero-engine (such as the blades and the corresponding surface of the combustion chamber), the movement speed of the traction rope is reduced (≤5mm / s), the damping coefficient is increased, the attitude stability is improved, the camera shooting accuracy is ensured, and the scanning requirements of complex curved areas are adapted; Mode 2 (high-efficiency mode): for the large flat area of ​​the aero-engine, the movement speed of the traction rope is increased (≤10mm / s), the damping coefficient is reduced, the scanning efficiency is improved, and the coverage integrity is ensured at the same time; Mode 3 (emergency mode): when the distance between the continuous detection robotic arm and the surface of the aero-engine is less than the safety threshold (3mm), emergency braking is automatically triggered, the damping is increased and the tension of the traction rope is reduced to avoid collision, and the backup traction scheme is activated to ensure that the scanning is not interrupted.

[0092] Furthermore, to address the issues of decreased accuracy caused by traction rope wear during long-term scanning and joint fatigue in continuous inspection robotic arms, real-time detection of traction rope wear can be considered. For example, by detecting parameters such as changes in tension and vibration frequency of the continuous inspection robotic arm, when the wear exceeds a preset threshold (e.g., wear ≥ 1 mm), the tension distribution of the traction rope can be automatically adjusted, the stress point shifted, and the service life of the traction rope extended. Simultaneously, by combining the fatigue level of the continuous inspection robotic arm joints, the movement rhythm of the traction rope can be dynamically adjusted to avoid excessive joint wear and improve the long-term stability of the overall scanning process.

[0093] S1606. After the rotation control ends, control each industrial camera to take high-resolution pictures to obtain a local fine point cloud that matches the current target viewpoint group.

[0094] S170. Using the coarse point cloud model as a fixed rigid skeleton constraint, global anchoring and registration are performed on multiple sets of local fine point clouds to obtain a high-precision three-dimensional solid model of the aero-engine.

[0095] Because long-distance flexible motion of a robotic arm inevitably results in accumulated errors in body localization, simply relying on forward kinematics solutions to stitch together point clouds will lead to model ghosting. Therefore, this invention adopts a "global anchoring registration" strategy, namely: Scale-Invariant Feature Transform (SIFT) key points of the local fine point cloud are extracted, and initial coarse registration is performed using Fast Point Feature Histogram (FPFH) to obtain the initial transformation matrix. Using the initial coarse point cloud model as a fixed skeleton constraint, an improved Iterative Closest Point (ICP) algorithm based on point-to-surface distance and incorporating the Huber loss function is adopted to unify the local fine point clouds into the global coordinate system, thereby eliminating the cumulative positioning error caused by the long-distance movement of the robotic arm.

[0096] Specifically, in Figure 3This is a block diagram illustrating the principle of registering and fusing multiple sets of local fine point clouds to a global coarse point cloud model, applicable to embodiments of the present invention. For example... Figure 3 As shown, firstly, multiple sets of local fine point clouds (i.e., Figure 3 The system first generates local fine point clouds 1, 2, ..., N. Then, SIFT keypoints are extracted from each group of local fine point clouds, and a 33-dimensional fast point feature histogram descriptor is calculated. Simultaneously, using the fixed coarse point cloud model from the first stage (i.e., the global coarse point cloud) as a fixed rigid skeleton constraint, the RANSAC algorithm is used for coarse registration and feature matching to obtain the initial approximate transformation matrix from each group of local fine point clouds to the global coordinate system.

[0097] Furthermore, using the coarse registration result as the initial value for iteration, an improved iterative nearest-point algorithm is employed for fine registration. Specifically, let the i-th point in the local fine point cloud be... Its corresponding point in the global skeleton point cloud (coarse point cloud model) is The normal vector of the corresponding point is To improve the registration slip resistance of the smooth metal surface region of the engine and suppress isolated noise points, an error objective functional based on point-to-plane distance metric combined with the Huber robust loss function is adopted: Where R and t are the optimal rotation matrix and translation vector to be solved; Here is the Huber loss function, where: ,in, This is a preset threshold parameter.

[0098] The global coarse point cloud obtained after coarse scanning is always used as a fixed rigid skeleton constraint to limit the scale drift and non-rigid deformation of the local fine point cloud during the optimization process. After multiple threshold iterations, the massive amount of local fine point cloud from multiple perspectives is precisely registered to a unified global coordinate system, and finally a complete high-precision 3D solid model of the aero-engine is output. After calibration and verification by a laser tracker, the overall surface reconstruction error of the model is strictly controlled below 0.05mm.

[0099] The technical solution of this invention applies a continuous inspection robotic arm to the field of 3D modeling of aero-engines. By employing a fully automated two-stage scanning strategy, an obstacle avoidance envelope for collision detection is pre-constructed based on a coarse point cloud model, fundamentally and effectively avoiding mechanical collision interference between the robotic arm and the engine or surrounding structures. This significantly shortens the planning and scanning cycle and ensures the operational safety of high-value aero-engine equipment. An innovative viewpoint planning algorithm based on geodesics and a joint view frustum visibility matrix is ​​proposed. Combining the adaptive bending capability of the continuous flexible robotic arm with the multi-camera active rotation mechanism around the axis, and controlling the field of view overlap rate through a precise geometric model, the scanning coverage of the complex external topography of the engine is maximized, effectively eliminating blind spots and obstructions. A point cloud fusion strategy of "coarse scanning global anchor points as the skeleton + fine scanning local precise registration" effectively eliminates the cumulative positioning errors caused by the long-distance movement of the multi-view robotic arm, ensuring the high fidelity and global geometric consistency of the final generated 3D digital twin model.

[0100] Specifically, in Figure 4 The diagram illustrates a construction process for a coarse scanning and safe obstacle avoidance envelope applicable to an embodiment of the present invention. Figure 5 The diagram shows a schematic of a coarse point cloud model applicable to an embodiment of the present invention. Figure 6 The diagram shows a high-precision three-dimensional solid model applicable to an embodiment of the present invention.

[0101] Figure 7 This is a schematic diagram of a cooperative scanning device for aero-engines provided as an embodiment of the present invention. Figure 7 As shown, the device includes: The coarse point cloud model acquisition module 710 is used to control the continuous inspection robotic arm to perform a coarse scan of the aero-engine under test at a preset safe distance, and obtain a coarse point cloud model containing the overall outline of the aero-engine based on the coarse scan results. The continuous inspection robotic arm includes a drive unit and a robotic arm segment. A preset number of inspection components are set in the axial direction of the robotic arm segment. Each inspection component is independently equipped with a high-resolution industrial camera. The inspection components are connected in series through a central flexible skeleton. Multiple curved traction ropes are set around the periphery of the central flexible skeleton. The curved traction ropes are sequentially strung on each inspection component. A rotating traction rope is wrapped around the outer edge of each inspection component. The drive unit controls the bending action of the robotic arm segment by driving the curved traction ropes, and controls the rotation action of each inspection component around the axis by driving the rotating traction ropes. The candidate viewpoint formation module 720 is used to generate a set of geodesic paths covering the surface of the aero-engine based on the geodesic distance field matched with the coarse point cloud model, and to form multiple candidate viewpoints based on the set of geodesic paths; wherein each viewpoint includes a spatial position and a normal vector direction; The viewpoint group acquisition module 730 is used to cluster and group multiple candidate viewpoints under the constraints of coverage complementarity, spatial proximity and normal attitude consistency to obtain multiple viewpoint groups, each viewpoint group containing a preset number of viewpoints. The visibility matrix forming module 740 is used to group the industrial cameras on each detection component of the continuous detection robot arm as multi-viewpoint groups. Based on the shooting parameters of each industrial camera, it calculates the visibility of each triangular facet in the coarse point cloud model when the multi-viewpoint group is set on each viewpoint group, and forms a joint viewpoint visibility matrix. The target viewpoint group sequence acquisition module 750 is used to obtain the target viewpoint group sequence iteratively using a preset genetic algorithm based on the motion parameters of the continuous detection robot arm and the joint viewpoint cone visibility matrix, with coverage and energy consumption path cost as constraints. The target viewpoint group sequence includes multiple target viewpoint groups arranged in sequence. The local fine point cloud imaging module 760 is used to control each industrial camera of the continuous inspection robot arm to be positioned at each target viewpoint of each target viewpoint group by adjusting the bending traction rope and / or rotating traction rope of the continuous inspection robot arm, and to take high-resolution pictures to obtain multiple sets of local fine point clouds that match the aero-engine. The high-precision model generation module 770 is used to take the coarse point cloud model as a fixed rigid skeleton constraint and perform global anchoring and registration on multiple sets of local fine point clouds to obtain a high-precision three-dimensional solid model of the aero-engine.

[0102] The technical solution of this invention applies a continuous inspection robotic arm to the field of 3D modeling of aero-engines. By employing a fully automated two-stage scanning strategy, an obstacle avoidance envelope for collision detection is pre-constructed based on a coarse point cloud model, fundamentally and effectively avoiding mechanical collision interference between the robotic arm and the engine or surrounding structures. This significantly shortens the planning and scanning cycle and ensures the operational safety of high-value aero-engine equipment. An innovative viewpoint planning algorithm based on geodesics and a joint view frustum visibility matrix is ​​proposed. Combining the adaptive bending capability of the continuous flexible robotic arm with the multi-camera active rotation mechanism around the axis, and controlling the field of view overlap rate through a precise geometric model, the scanning coverage of the complex external topography of the engine is maximized, effectively eliminating blind spots and obstructions. A point cloud fusion strategy of "coarse scanning global anchor points as the skeleton + fine scanning local precise registration" effectively eliminates the cumulative positioning errors caused by the long-distance movement of the multi-view robotic arm, ensuring the high fidelity and global geometric consistency of the final generated 3D digital twin model.

[0103] Based on the above embodiments, the viewpoint group acquisition module 730 can be specifically used for: Calculate the single-point coverage score corresponding to each candidate viewpoint, and sort the candidate viewpoints in descending order of single-point coverage score to obtain the candidate viewpoint sequence; From the candidate viewpoint sequence, one candidate viewpoint is sequentially selected as the seed viewpoint, and from the remaining candidate viewpoints, multiple related viewpoints that simultaneously satisfy the conditions of complementary coverage, spatial distance, and angle between normal vectors are selected. Based on the condition of uniform distance distribution, multiple associated viewpoints are filtered to obtain a preset number minus one target associated viewpoint, and each target associated viewpoint and seed viewpoint are combined into a viewpoint group. After filtering out the generated viewpoint groups from the candidate viewpoint sequence, the process returns to extracting one candidate viewpoint from the candidate viewpoint sequence as a seed viewpoint in sequence, until all candidate viewpoints in the candidate viewpoint sequence have been processed, resulting in multiple viewpoint groups.

[0104] Based on the above embodiments, the coverage complementarity condition can be: ; in, Candidate viewpoints With candidate viewpoints The coverage complementarity index between them The preset coverage complementarity threshold, , Indicate candidate viewpoints The number of point clouds in the coarse point cloud model that can be covered. Indicate candidate viewpoints The number of point clouds in the coarse point cloud model that can be covered. ; The spatial distance condition is: ; in, Candidate viewpoints With candidate viewpoints The distance between them The lower limit of the preset distance threshold. This is the preset upper limit of the distance threshold; The condition for the included angle of the normal vectors is: ; in, Candidate viewpoints With candidate viewpoints The angle between the normal vectors, This is a preset threshold for the angle between the normal vectors; Accordingly, after obtaining multiple viewpoint groups, the method further includes: Based on the coverage complementarity index, distance, and normal vector angle between adjacent viewpoints in each viewpoint group, calculate the comprehensive quality score corresponding to each viewpoint group, and filter out invalid viewpoint groups whose comprehensive quality scores are less than a preset score threshold.

[0105] Based on the above embodiments, the visibility matrix forming module 740 can be specifically used for: The coarse point cloud model is divided into facets to obtain triangular facets that match the coarse point cloud model. Calculate the field-of-view volume of a single industrial camera based on its horizontal and vertical field-of-view angles. A current viewpoint group is obtained sequentially, and based on the field of view volume of a single industrial camera and the imaging parameters of the industrial camera, the joint field of view volume matching the multi-viewpoint group is calculated when each industrial camera in the multi-viewpoint group is respectively placed on each viewpoint in the current viewpoint group. The joint field of view volume is matched and mapped to each triangular facet of the coarse point cloud model to obtain the visibility of the current viewpoint group to each triangular facet, forming the visibility vector of the current viewpoint group. Return to the previous operation to obtain the current viewpoint group one by one until the visibility vectors corresponding to all viewpoint groups are obtained; By combining all the visibility vectors, we obtain the joint view frustum visibility matrix.

[0106] Based on the above embodiments, the visibility matrix forming module 740 can be further used for: When each industrial camera in the multi-viewpoint group is deployed to each viewpoint in the current viewpoint group, the bending diameter R and arc length l between two adjacent industrial cameras are obtained, and the deflection angle θ between two adjacent industrial cameras is calculated according to the formula θ=l / R. According to the formula: The field-of-view overlap rate α between two adjacent industrial cameras is calculated. Where D is the far-plane distance of the industrial camera, d is the near-plane distance of the industrial camera, H is the horizontal field of view of the industrial camera, and V is the vertical field of view of the industrial camera. The combined field of view volume of a single industrial camera and the field of view overlap rate between two adjacent industrial cameras are calculated based on the field of view volume of the multi-view group for the current view group.

[0107] Based on the above embodiments, the target viewpoint group sequence acquisition module 750 can be further used for: Calculate the group geometric center and the main direction of the group corresponding to each viewpoint group; Construct a fitness function; the fitness function includes a coverage calculation term and an energy consumption path cost calculation term. The coverage calculation term is calculated based on the joint view frustum visibility matrix. The energy consumption path cost calculation term includes the movement distance when moving between adjacent view groups and the joint angle when moving between adjacent view groups. The movement distance is calculated based on the group geometric center of the view group, and the joint angle is calculated based on the group principal direction of the view group and the motion parameters of the continuous detection robot arm. Based on the fitness function, with the goal of maximizing coverage and minimizing energy consumption path cost, a preset genetic algorithm is used to iteratively obtain the target viewpoint group sequence.

[0108] Based on the above embodiments, the local fine point cloud imaging module 760 can be further used for: In the target viewpoint group sequence, a current target viewpoint group is obtained sequentially, and based on the spatial position and curvature information of each target viewpoint in the current target viewpoint group, an ideal shooting position curve matching the current target viewpoint group is fitted. The current position trajectory of the central axis of the continuous inspection robot arm is obtained. Based on the current position trajectory and the ideal shooting position curve, the first type of tension of each curved traction rope in the continuous inspection robot arm is calculated when the current position trajectory is controlled to move towards the ideal shooting position curve. Based on the first type of tension, the first type of curvature control is performed on the continuous inspection robot arm. When the central axis of the continuous inspection robot arm is determined to be closely aligned with the ideal shooting position curve, the camera positions of each industrial camera in the continuous inspection robot arm are obtained. Based on the camera positions of each industrial camera and the spatial positions of each target viewpoint in the current target viewpoint group, the second type of tension of each curved traction rope in the continuous detection robot arm is calculated when each industrial camera moves toward each target viewpoint. Based on the second type of tension, the second type of curvature control is performed on the continuous detection robot arm. When it is determined that each industrial camera in the continuous inspection robot arm is close to the matching target viewpoint, the rotation amount of each rotating traction rope in the continuous inspection robot arm is calculated according to the normal vector direction of each target viewpoint. When the camera optical axis of each industrial camera moves towards the normal vector direction of each target viewpoint, the rotation amount of each rotating traction rope in the continuous inspection robot arm is calculated. Based on the calculated rotation amount, the rotation control of each inspection component in the continuous inspection robot arm is performed. After the rotation control is completed, each industrial camera is controlled to take high-resolution pictures to obtain a local fine point cloud that matches the current target viewpoint group.

[0109] The cooperative scanning device for aero-engines provided in this embodiment of the invention can execute the cooperative scanning method for aero-engines provided in any embodiment of the invention, and has the corresponding functional modules and beneficial effects of the method.

[0110] The collection, storage, use, processing, transmission, provision, and disclosure of user personal information involved in the technical solution disclosed herein comply with the provisions of relevant laws and regulations and do not violate public order and good morals.

[0111] Figure 8 A schematic diagram of an electronic device 10, which can be used to implement embodiments of the present invention, is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices (e.g., helmets, glasses, watches, etc.), and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the invention described and / or claimed herein.

[0112] like Figure 8 As shown, the electronic device 10 includes at least one processor 11 and a memory, such as a read-only memory (ROM) 12 or a random access memory (RAM) 13, communicatively connected to the at least one processor 11. The memory stores computer programs executable by the at least one processor. The processor 11 can perform various appropriate actions and processes based on the computer program stored in the ROM 12 or loaded from storage unit 18 into the RAM 13. The RAM 13 can also store various programs and data required for the operation of the electronic device 10. The processor 11, ROM 12, and RAM 13 are interconnected via a bus 14. An input / output (I / O) interface 15 is also connected to the bus 14.

[0113] Multiple components in electronic device 10 are connected to I / O interface 15, including: input unit 16, such as keyboard, mouse, etc.; output unit 17, such as various types of displays, speakers, etc.; storage unit 18, such as disk, optical disk, etc.; and communication unit 19, such as network card, modem, wireless transceiver, etc. Communication unit 19 allows electronic device 10 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0114] Processor 11 can be various general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of processor 11 include, but are not limited to, central processing unit (CPU), graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, digital signal processors (DSPs), and any suitable processor, controller, microcontroller, etc. Processor 11 performs the various methods and processes described above, such as performing the cooperative scanning method for aero-engines as described in any embodiment of the present invention. That is: The continuous inspection robotic arm is controlled to perform a coarse scan of the aero-engine under test at a preset safe distance, and a rough point cloud model containing the overall outline of the aero-engine is obtained based on the coarse scan results. The continuous inspection robotic arm includes a drive unit and a robotic arm segment. A preset number of inspection components are set in the axial direction of the robotic arm segment. Each inspection component is independently equipped with a high-resolution industrial camera. The inspection components are connected in series through a central flexible skeleton. Multiple curved traction ropes are set around the periphery of the central flexible skeleton. The curved traction ropes are sequentially strung on each inspection component. A rotating traction rope is wrapped around the outer edge of each inspection component. The drive unit controls the bending action of the robotic arm segment by driving the curved traction ropes, and controls the rotation action of each inspection component around the axis by driving the rotating traction ropes. Based on the geodesic distance field matched with the coarse point cloud model, a set of geodesic paths covering the surface of the aero-engine is generated, and multiple candidate viewpoints are formed based on the set of geodesic paths; each viewpoint includes a spatial position and a normal vector direction. With the constraints of coverage complementarity, spatial proximity and normal attitude consistency, multiple candidate viewpoints are clustered and grouped to obtain multiple viewpoint groups, each containing a preset number of viewpoints. The industrial cameras on each detection component of the continuous inspection robot are grouped into multiple viewpoints. Based on the shooting parameters of each industrial camera, the visibility of each triangular facet in the coarse point cloud model is calculated when the multiple viewpoint group is set on each viewpoint group, forming a joint viewpoint visibility matrix. Based on the motion parameters of the continuous detection robot arm and the joint view frustum visibility matrix, and with coverage and energy consumption path cost as constraints, a preset genetic algorithm is used to iteratively obtain the target viewpoint group sequence, wherein the target viewpoint group sequence includes multiple target viewpoint groups arranged in sequence. By adjusting the bending and / or rotating traction ropes of the continuous inspection robot arm, the industrial cameras of the continuous inspection robot arm are positioned at the target viewpoints of each target viewpoint group and high-resolution images are taken to obtain multiple sets of local fine point clouds that match the aero-engine. By using a coarse point cloud model as a fixed rigid skeleton constraint, and performing global anchoring and registration on multiple sets of local fine point clouds, a high-precision three-dimensional solid model of the aero-engine is obtained.

[0115] In some embodiments, the cooperative scanning method for aero-engines as described in any one of the embodiments of the present invention can be implemented as a computer program tangibly contained in a computer-readable storage medium, such as storage unit 18. In some embodiments, part or all of the computer program can be loaded and / or installed on electronic device 10 via ROM 12 and / or communication unit 19. When the computer program is loaded into RAM 13 and executed by processor 11, one or more steps of the cooperative scanning method for aero-engines as described above as any one of the embodiments of the present invention can be performed. Alternatively, in other embodiments, processor 11 can be configured by any other suitable means (e.g., by means of firmware) to perform the cooperative scanning method for aero-engines as described in any one of the embodiments of the present invention.

[0116] Various embodiments of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), systems-on-a-chip (SoCs), payload-programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments may include implementations in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a dedicated or general-purpose programmable processor, capable of receiving data and instructions from a storage system, at least one input device, and at least one output device, and transmitting data and instructions to the storage system, the at least one input device, and the at least one output device.

[0117] Computer programs used to implement the methods of the present invention may be written in any combination of one or more programming languages. These computer programs may be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, such that when executed by the processor, the computer programs cause the functions / operations specified in the flowcharts and / or block diagrams to be performed. The computer programs may be executed entirely on a machine, partially on a machine, or as a standalone software package, partially on a machine and partially on a remote machine, or entirely on a remote machine or server.

[0118] In the context of this invention, a computer-readable storage medium can be a tangible medium that may contain or store a computer program for use by or in conjunction with an instruction execution system, apparatus, or device. A computer-readable storage medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination thereof. Alternatively, a computer-readable storage medium may be a machine-readable signal medium. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof.

[0119] To provide interaction with a user, the systems and techniques described herein can be implemented on an electronic device having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the electronic device. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including sound input, voice input, or tactile input).

[0120] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as data servers), or middleware components (e.g., application servers), or frontend components (e.g., user computers with graphical user interfaces or web browsers through which users can interact with implementations of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., communication networks). Examples of communication networks include local area networks (LANs), wide area networks (WANs), blockchain networks, and the Internet.

[0121] A computing system can include clients and servers. Clients and servers are generally located far apart and typically interact through communication networks. The client-server relationship is created by computer programs running on the respective computers and having a client-server relationship with each other. The server can be a cloud server, also known as a cloud computing server or cloud host, which is a hosting product within the cloud computing service system to address the shortcomings of traditional physical hosts and VPS services, such as high management difficulty and weak business scalability.

[0122] It should be understood that the various forms of processes shown above can be used, with steps reordered, added, or deleted. For example, the steps described in this invention can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution of this invention can be achieved, and this is not limited herein.

[0123] The specific embodiments described above do not constitute a limitation on the scope of protection of this invention. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this invention should be included within the scope of protection of this invention.

Claims

1. A cooperative scanning method for aero-engines, characterized in that, The method includes: The continuous inspection robotic arm is controlled to perform a coarse scan of the aero-engine under test at a preset safe distance, and a rough point cloud model containing the overall outline of the aero-engine is obtained based on the coarse scan results. The continuous inspection robotic arm includes a drive unit and a robotic arm segment. A preset number of inspection components are set in the axial direction of the robotic arm segment. Each inspection component is independently equipped with a high-resolution industrial camera. The inspection components are connected in series through a central flexible skeleton. Multiple curved traction ropes are set around the periphery of the central flexible skeleton. The curved traction ropes are sequentially strung on each inspection component. A rotating traction rope is wrapped around the outer edge of each inspection component. The drive unit controls the bending action of the robotic arm segment by driving the curved traction ropes, and controls the rotation action of each inspection component around the axis by driving the rotating traction ropes. Based on the geodesic distance field matched with the coarse point cloud model, a set of geodesic paths covering the surface of the aero-engine is generated, and multiple candidate viewpoints are formed based on the set of geodesic paths; each viewpoint includes a spatial position and a normal vector direction. With the constraints of coverage complementarity, spatial proximity and normal attitude consistency, multiple candidate viewpoints are clustered and grouped to obtain multiple viewpoint groups, each containing a preset number of viewpoints. The industrial cameras on each detection component of the continuous inspection robot are grouped into multiple viewpoints. Based on the shooting parameters of each industrial camera, the visibility of each triangular facet in the coarse point cloud model is calculated when the multiple viewpoint group is set on each viewpoint group, forming a joint viewpoint visibility matrix. Based on the motion parameters of the continuous detection robot arm and the joint view frustum visibility matrix, and with coverage and energy consumption path cost as constraints, a preset genetic algorithm is used to iteratively obtain the target viewpoint group sequence, wherein the target viewpoint group sequence includes multiple target viewpoint groups arranged in sequence. By adjusting the bending and / or rotating traction ropes of the continuous inspection robot arm, the industrial cameras of the continuous inspection robot arm are positioned at the target viewpoints of each target viewpoint group and high-resolution images are taken to obtain multiple sets of local fine point clouds that match the aero-engine. By using a coarse point cloud model as a fixed rigid skeleton constraint, and performing global anchoring and registration on multiple sets of local fine point clouds, a high-precision three-dimensional solid model of the aero-engine is obtained.

2. The method according to claim 1, characterized in that, Constrained by coverage complementarity, spatial proximity, and normal attitude consistency, multiple candidate viewpoints are clustered and grouped to obtain multiple viewpoint groups, including: Calculate the single-point coverage score corresponding to each candidate viewpoint, and sort the candidate viewpoints in descending order of single-point coverage score to obtain the candidate viewpoint sequence; From the candidate viewpoint sequence, one candidate viewpoint is sequentially selected as the seed viewpoint, and from the remaining candidate viewpoints, multiple related viewpoints that simultaneously satisfy the conditions of complementary coverage, spatial distance, and angle between normal vectors are selected. Based on the condition of uniform distance distribution, multiple associated viewpoints are filtered to obtain a preset number minus one target associated viewpoint, and each target associated viewpoint and seed viewpoint are combined into a viewpoint group. After filtering out the generated viewpoint groups from the candidate viewpoint sequence, the process returns to extracting one candidate viewpoint from the candidate viewpoint sequence as a seed viewpoint in sequence, until all candidate viewpoints in the candidate viewpoint sequence have been processed, resulting in multiple viewpoint groups.

3. The method according to claim 2, characterized in that, The coverage complementarity condition is: ; in, Candidate viewpoints With candidate viewpoints The coverage complementarity index between them The preset coverage complementarity threshold, , Indicate candidate viewpoints The number of point clouds in the coarse point cloud model that can be covered. Indicate candidate viewpoints The number of point clouds in the coarse point cloud model that can be covered. ; The spatial distance condition is: ; in, Candidate viewpoints With candidate viewpoints The distance between them The lower limit of the preset distance threshold. This is the preset upper limit of the distance threshold; The condition for the included angle of the normal vectors is: ; in, Candidate viewpoints With candidate viewpoints The angle between the normal vectors, This is a preset threshold for the angle between the normal vectors; Accordingly, after obtaining multiple viewpoint groups, the method further includes: Based on the coverage complementarity index, distance, and normal vector angle between adjacent viewpoints in each viewpoint group, calculate the comprehensive quality score corresponding to each viewpoint group, and filter out invalid viewpoint groups whose comprehensive quality scores are less than a preset score threshold.

4. The method according to claim 1, characterized in that, The industrial cameras on each inspection component of the continuous inspection robot are grouped into multi-viewpoint groups. Based on the shooting parameters of each industrial camera, the visibility of each triangular facet in the coarse point cloud model is calculated when the multi-viewpoint group is set on each viewpoint group, forming a joint frustum visibility matrix, including: The coarse point cloud model is divided into facets to obtain triangular facets that match the coarse point cloud model. Calculate the field-of-view volume of a single industrial camera based on its horizontal and vertical field-of-view angles. Sequentially acquire a current viewpoint group, and calculate the joint field of view volume matching the multi-viewpoint group when each industrial camera in the multi-viewpoint group is deployed to each viewpoint in the current viewpoint group based on the field of view volume of a single industrial camera and the imaging parameters of the industrial camera. The joint field of view volume is matched and mapped to each triangular facet of the coarse point cloud model to obtain the visibility of the current viewpoint group to each triangular facet, forming the visibility vector of the current viewpoint group. Return to the previous operation to obtain the current viewpoint group one by one until the visibility vectors corresponding to all viewpoint groups are obtained; By combining all the visibility vectors, we obtain the joint view frustum visibility matrix.

5. The method according to claim 4, characterized in that, Based on the field of view volume of a single industrial camera and its imaging parameters, calculate the joint field of view volume matching the multi-view group when each industrial camera in the multi-view group is deployed to a viewpoint within the current viewpoint group, including: When each industrial camera in the multi-viewpoint group is deployed to each viewpoint in the current viewpoint group, the bending diameter R and arc length l between two adjacent industrial cameras are obtained, and the deflection angle θ between two adjacent industrial cameras is calculated according to the formula θ=l / R. According to the formula: The field-of-view overlap rate α between two adjacent industrial cameras is calculated. Where D is the far-plane distance of the industrial camera, d is the near-plane distance of the industrial camera, H is the horizontal field of view of the industrial camera, and V is the vertical field of view of the industrial camera. The combined field of view volume of a single industrial camera and the field of view overlap rate between two adjacent industrial cameras are calculated based on the field of view volume of the multi-view group for the current view group.

6. The method according to claim 2, characterized in that, Based on the motion parameters of the continuous detection robot arm and the joint view frustum visibility matrix, and constrained by coverage and energy consumption path cost, a pre-defined genetic algorithm is used to iteratively obtain the target viewpoint group sequence, including: Calculate the group geometric center and the main direction of the group corresponding to each viewpoint group; Construct a fitness function; the fitness function includes a coverage calculation term and an energy consumption path cost calculation term. The coverage calculation term is calculated based on the joint view frustum visibility matrix. The energy consumption path cost calculation term includes the movement distance when moving between adjacent view groups and the joint angle when moving between adjacent view groups. The movement distance is calculated based on the group geometric center of the view group, and the joint angle is calculated based on the group principal direction of the view group and the motion parameters of the continuous detection robot arm. Based on the fitness function, with the goal of maximizing coverage and minimizing energy consumption path cost, a preset genetic algorithm is used to iteratively obtain the target viewpoint group sequence.

7. The method according to any one of claims 1-6, characterized in that, By adjusting the bending and / or rotating traction ropes of the continuous inspection robot arm, the industrial cameras of the continuous inspection robot arm are controlled to be positioned at the respective target viewpoints of each target viewpoint group, and high-resolution images are captured, including: In the target viewpoint group sequence, a current target viewpoint group is obtained sequentially, and based on the spatial position and curvature information of each target viewpoint in the current target viewpoint group, an ideal shooting position curve matching the current target viewpoint group is fitted. The current position trajectory of the central axis of the continuous inspection robot arm is obtained. Based on the current position trajectory and the ideal shooting position curve, the first type of tension of each curved traction rope in the continuous inspection robot arm is calculated when the current position trajectory is controlled to move towards the ideal shooting position curve. Based on the first type of tension, the first type of curvature control is performed on the continuous inspection robot arm. When the central axis of the continuous inspection robot arm is determined to be closely aligned with the ideal shooting position curve, the camera positions of each industrial camera in the continuous inspection robot arm are obtained. Based on the camera positions of each industrial camera and the spatial positions of each target viewpoint in the current target viewpoint group, the second type of tension of each curved traction rope in the continuous detection robot arm is calculated when each industrial camera moves toward each target viewpoint. Based on the second type of tension, the second type of curvature control is performed on the continuous detection robot arm. When it is determined that each industrial camera in the continuous inspection robot arm is close to the matching target viewpoint, the rotation amount of each rotating traction rope in the continuous inspection robot arm is calculated according to the normal vector direction of each target viewpoint. When the camera optical axis of each industrial camera moves towards the normal vector direction of each target viewpoint, the rotation amount of each rotating traction rope in the continuous inspection robot arm is calculated. Based on the calculated rotation amount, the rotation control of each inspection component in the continuous inspection robot arm is performed. After the rotation control is completed, each industrial camera is controlled to take high-resolution pictures to obtain a local fine point cloud that matches the current target viewpoint group.

8. An electronic device, characterized in that, The electronic device includes: At least one processor; and A memory communicatively connected to the at least one processor; wherein, The memory stores a computer program that can be executed by the at least one processor, the computer program being executed by the at least one processor to enable the at least one processor to perform the cooperative scanning method for aero-engines according to any one of claims 1-7.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions that cause a processor to execute the cooperative scanning method for aero-engines as described in any one of claims 1-7.

10. A computer program product, characterized in that, The computer program product includes a computer program that, when executed by a processor, implements the cooperative scanning method for aero-engines according to any one of claims 1-7.

Images