A laser scanning viewpoint adaptive planning method for surface data collection of a speed reducer

Abstract

This invention relates to adaptive planning, and more particularly to an adaptive planning method for laser scanning viewpoints in data acquisition of speed reducer surfaces. The method includes: S1, constructing an acquisition system comprising a six-axis robot, a laser scanning camera, and an industrial control computer, and loading the speed reducer contour information and camera hardware parameters; S2, based on the speed reducer contour information, constructing an adaptively stretched ellipsoidal envelope model as the geometric reference for viewpoint generation; S3, establishing a dual-constraint model including hard and soft constraints, and determining the effective working range of the camera on the ellipsoidal envelope model; S4, constructing a viewpoint comprehensive evaluation function, and iteratively selecting the next optimal acquisition viewpoint within the effective working range based on the viewpoint comprehensive evaluation function; S5, sorting and smoothing the generated viewpoint set, and outputting a scan path without redundancy. This method effectively improves the efficiency and accuracy of data acquisition from complex curved surfaces of speed reducers, and has good engineering practicality and promotional value.

Description

Technical Field

[0001] This invention relates to adaptive planning, and more particularly to an adaptive planning method for laser scanning viewpoints for data acquisition on the surface of a speed reducer. Background Technology

[0002] Accurate 3D data acquisition of the surface geometry of a speed reducer is a fundamental prerequisite for digital modeling, quality inspection, fault diagnosis, and digital twin construction. This is especially true for non-standard speed reducers, whose housings often exhibit complex geometric structures such as elliptical eccentric cavities, irregular curved surfaces, and free-form surface transitions, posing a significant challenge to laser-scan-based 3D reconstruction data acquisition.

[0003] To overcome the aforementioned bottlenecks, laser scanning and robotic collaborative sampling have become the mainstream approach for industrial 3D reconstruction data acquisition. By integrating laser scanning sensors at the end of a robotic arm, the scanning viewpoint can be flexibly adjusted, enabling multi-view, high-coverage acquisition of complex surfaces of speed reducers. However, how to scientifically and automatically plan the next optimal viewpoint for each laser scan, and how to construct a complete and accurate 3D model of the speed reducer surface under conditions of no CAD model, no prior structure, and unknown freeform surface, remains a challenge in viewpoint planning research in the field of laser scanning 3D reconstruction.

[0004] In existing research, the best perspective methods can be roughly divided into two categories: surface model-based and voxel space modeling-based. These methods perform well in the reconstruction of regular structures or standard models, but in the surface scanning scenario of speed reducers, especially non-standard speed reducers, they are prone to scanning registration failure, model missing or redundant acquisition due to over-reliance on the continuity assumption or spatial consistency constraint, and the low fit between the spherical enclosing strategy and the long shaft structure of the speed reducer. They are difficult to meet comprehensive requirements. Summary of the Invention

[0005] This invention addresses the shortcomings of existing technologies by providing a laser scanning viewpoint adaptive planning method for data acquisition on the surface of a speed reducer.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: an adaptive planning method for laser scanning viewpoint for data acquisition on the surface of a speed reducer, comprising the following steps:

[0007] S1. Construct a data acquisition system including a six-axis robot, a laser scanning camera, and an industrial control computer, and load the reducer contour information and camera hardware parameters.

[0008] S2. Based on the reducer contour information, construct an adaptively stretched ellipsoidal envelope model as the geometric reference for viewpoint generation;

[0009] S3. Based on the camera hardware parameters, establish a dual constraint model that includes hard constraints and soft constraints, and determine the effective working range of the camera on the ellipsoidal envelope model.

[0010] S4. Based on the current point cloud coverage status, construct a viewpoint comprehensive evaluation function, and iteratively select the next optimal acquisition viewpoint within the effective working range according to the viewpoint comprehensive evaluation function until the preset coverage threshold is met.

[0011] S5. Sort and smooth the generated viewpoint set to output a scan path without redundancy.

[0012] Furthermore, step S2 specifically includes the following steps:

[0013] S21. Based on the reducer contour information, construct the equation of an ellipsoid with coordinate axes parallel to the principal axis; wherein, the center coordinates and semi-axis lengths of the ellipsoid equation correspond to the center and major semi-axis, middle semi-axis, and minor semi-axis of the reducer, respectively.

[0014] S22. Obtain the bounding box point cloud of the reducer's position contour and construct an optimization objective function; the optimization objective function is to minimize the sum of the distances from any point in the bounding box point cloud to the ellipsoid; use the particle swarm optimization algorithm to solve the optimization objective function, obtain the center coordinates and semi-axis length parameters of the ellipsoid, and generate the ellipsoid envelope model;

[0015] S23. Calculate the ratio of the volume of the ellipsoidal envelope model to the actual volume of the reducer, and use it as the envelope redundancy rate; if the envelope redundancy rate is within a preset threshold range, it is determined that the ellipsoidal envelope model fits the shape of the reducer, and the model construction is valid.

[0016] Furthermore, step S3 includes the following steps:

[0017] S31. Based on the intrinsic parameters of the laser camera and the acquisition distance D, construct a mathematical model of the camera's field of view, and establish the spatial mapping relationship between the acquisition distance D and the maximum field of view size in a single scan, as a mathematical benchmark for calculating the hard constraint interval and the soft constraint interval.

[0018] S32. Using the aforementioned spatial mapping relationship, and combining the sensor physical size of the laser camera and the minimum resolvable feature size corresponding to the lens resolution, calculate in reverse the minimum effective viewing distance D that satisfies imaging integrity. min With the maximum effective viewing distance D that satisfies the feature recognition resolution max Thus, a hard constraint interval [D] is established for the acquisition distance D. min D max ];

[0019] S33. For the geometric dimensions of the local area of ​​the reducer, the distance adaptation range that satisfies the complete coverage is calculated using the spatial mapping relationship, and a soft constraint interval for the acquisition distance D is established by combining the minimum pixel density threshold of the local details.

[0020] S34. Perform an intersection operation on the hard constraint interval and the soft constraint interval to determine the final effective working interval of the camera acquisition distance.

[0021] Furthermore, in S31, the camera field of view mathematical model is constructed using the pinhole imaging model, and the maximum horizontal field of view size and the maximum vertical field of view size of a single scan are calculated based on the horizontal and vertical field of view angles of the laser camera at the current acquisition distance D.

[0022] In step S32, the minimum effective line-of-sight distance D is calculated. min At that time, the established physical constraint is that the horizontal dimension of the camera sensor must cover the width of the scanning area at the corresponding acquisition distance; the maximum effective viewing distance D is calculated. max At that time, the established physical constraint is that the number of imaging pixels of the minimum feature size of the reducer surface on the camera sensor is not less than the preset feature recognition pixel threshold.

[0023] In step S33, the soft constraint interval is established in the following way:

[0024] Based on the geometric constraint that the camera's field of view is not less than the maximum size of the local area currently being scanned by the reducer, the first distance adaptation range is determined.

[0025] The second distance adaptation range is determined based on the density constraint that the total number of pixels in the local area of ​​the camera is not less than the total number of pixels required for detail recognition in that area.

[0026] Furthermore, S33 specifically includes the following steps:

[0027] S331. Using the aforementioned spatial mapping relationship, establish the maximum field of view w for a single camera scan. h (That is, the maximum field of view w) h (and the maximum size L of the local area currently being scanned by the reducer) max Geometric constraint relationship: w h Greater than or equal to L max Based on this geometric constraint, the first distance adaptation range that satisfies the requirement of covering the entire local region is derived.

[0028] S332. Combining the total number of pixels of the laser camera's sensor and the aforementioned spatial mapping relationship, establish the total number of pixels N in the local area image and the pixel threshold N required for detail recognition in that area. target Density constraint relationship: N is greater than or equal to N targetBased on this density constraint, a second distance adaptation range that satisfies the requirement of clear detail recognition is derived.

[0029] S333. Find the intersection of the first distance adaptation range and the second distance adaptation range to generate a soft constraint interval for the acquisition distance D.

[0030] Further, S4 includes the following steps:

[0031] S41. Determine the acquisition distance within the final effective working range, construct a virtual camera observation surface that maintains the acquisition distance with the surface of the ellipsoidal envelope model, and use it as a candidate viewpoint set space;

[0032] S42, Based on the currently uncovered area U k Based on the distribution of the virtual camera observation surface, a comprehensive viewpoint evaluation function F(V) is constructed. Points that maximize F(V) are then selected as the next optimal acquisition viewpoint V. k+1 Wherein, the comprehensive evaluation function F(V) is the weighted fusion result of each evaluation index, and the evaluation index includes:

[0033] The weighting coefficients for each evaluation indicator;

[0034] Viewpoint to uncovered area U k The shortest distance, and the closer the distance, the larger the value of this indicator;

[0035] The percentage of the area where the viewpoint's field of view intersects with the uncovered area;

[0036] Spatial uniformity of the front viewpoint and the already generated set of viewpoints;

[0037] S43. Calculate in real time the ratio of the area of ​​the currently covered ellipsoid surface to the total area of ​​the ellipsoid envelope to obtain the model coverage rate Cov; if Cov reaches the preset coverage rate threshold, terminate viewpoint generation, otherwise return to S42 to continue iteration.

[0038] Further, S5 includes:

[0039] S51. Calculate the distance score based on the normalized Euclidean distance between the candidate viewpoint and the current viewpoint, and calculate the direction score based on the change of the azimuth angle of the candidate viewpoint relative to the current viewpoint in the ellipsoidal central coordinate system. Then, weight and fuse the distance score and the direction score to construct a comprehensive viewpoint score function.

[0040] S52. Starting from the initial viewpoint, iteratively select the viewpoint that gives the highest comprehensive score function value from the candidate viewpoints as the next acquisition viewpoint, until all viewpoints are sorted and an acquisition sequence is generated.

[0041] S53. Based on the generated acquisition sequence, output a smooth scanning path.

[0042] Compared with the prior art, the present invention has the following advantages.

[0043] This invention utilizes an ellipsoidal envelope model that conforms to the shape of a speed reducer, combined with camera constraints and point cloud quality feedback, to achieve iterative generation and path optimization of the optimal viewpoint. The generated scanning path features low redundancy and high coverage, effectively improving the efficiency and accuracy of data acquisition from the complex curved surfaces of the speed reducer, and possesses good engineering practicality and promotional value. Attached Figure Description

[0044] The present invention will be further described below with reference to the accompanying drawings and specific embodiments. The scope of protection of the present invention is not limited to the following description.

[0045] Figure 1 This is a schematic diagram of the overall process of the present invention;

[0046] Figure 2 This is a schematic diagram of the camera field of view imaging according to the present invention;

[0047] Figure 3 This is a schematic diagram illustrating the camera viewpoint selection of the present invention. Detailed Implementation

[0048] The technical solutions of the embodiments of this disclosure will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this disclosure, and not all embodiments. Based on the embodiments of this disclosure, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this disclosure.

[0049] The terminology used in the embodiments of this disclosure is for the purpose of describing particular embodiments only and is not intended to be limiting of this disclosure. The singular forms “a,” “the,” and “the” as used in the embodiments of this disclosure and the appended claims are also intended to include the plural forms unless the context clearly indicates otherwise.

[0050] Depending on the context, words such as “if” or “suppose” used here can be interpreted as “when”, “in response to determination”, or “in response to detection”.

[0051] For ease of understanding, the embodiments of this disclosure will be described in detail first.

[0052] like Figure 1-3 As shown in the specific embodiment, the laser scanning viewpoint adaptive planning method for data acquisition on the surface of a speed reducer includes the following logical steps:

[0053] S1. Construct a data acquisition system including a six-axis robot, a laser scanning camera, and an industrial control computer, and load the reducer contour information and camera hardware parameters.

[0054] S2. Based on the reducer contour information, construct an adaptively stretched ellipsoidal envelope model as the geometric reference for viewpoint generation;

[0055] S3. Based on the camera hardware parameters, establish a dual constraint model that includes hard constraints and soft constraints, and determine the effective working range of the camera on the ellipsoidal envelope model.

[0056] S4. Based on the current point cloud coverage status, construct a viewpoint comprehensive evaluation function, and iteratively select the next optimal acquisition viewpoint within the effective working range according to the viewpoint comprehensive evaluation function until the preset coverage threshold is met.

[0057] S5. Sort and smooth the generated viewpoint set to output a scan path without redundancy.

[0058] 1. The hardware components for step S1 are as follows:

[0059] (1) Six-axis industrial robot: A six-axis robot is used to carry a laser scanning camera and complete spatial posture adjustment according to the planned path.

[0060] (2) Laser scanning camera: A structured light camera is used to collect point cloud data on the surface of the reducer and output high-precision three-dimensional coordinate information.

[0061] (3) Industrial computer: As the core of system control, it carries the functional modules of the software layer to realize the integrated scheduling of robot motion control, camera data acquisition and algorithm operation.

[0062] (4) Fixtures: used to fix the reducer workpiece and ensure the workpiece position is stable during the acquisition process.

[0063] 2. To achieve the goal of "constructing an adaptively stretched ellipsoidal envelope model" in step S2, this embodiment employs a bounding box point cloud optimization method. The specific construction process is as follows:

[0064] (1) Establish the equation of the standard ellipsoid:

[0065] Let the coordinates of the center of the ellipsoid in three-dimensional space be... , its in , and The semi-axis lengths in the axial direction are respectively (Long half-shaft, corresponding to the long shaft direction of the reducer) (middle half axis) (Short half-shaft, corresponding to the radial section of the reducer), and satisfying (Matching the long axis structure). Then, with... The equation of a standard ellipsoid centered at x and whose coordinate axes are parallel to the principal axes of the ellipsoid is:

[0066] (1)

[0067] Let be the spatial coordinates of any point on the ellipsoid; when When the ellipsoid degenerates into a sphere, it is a special form of the traditional enclosing model, which proves the compatibility and generalization of the ellipsoidal envelope to the traditional method.

[0068] (2) Optimization of ellipsoidal parameters based on the profile of the reducer:

[0069] The 3D shape data of a non-standard speed reducer is unknown. Obtaining its positional contour information allows for the establishment of an initial modeling foundation using the bounding box method. First, an AABB bounding box can be constructed based on this contour information, and then a point cloud can be generated on the surface of the bounding box. ( Given the number of point clouds, and using the minimum sum of distances from the point clouds to the ellipsoid as the optimization objective, we solve for the optimal parameters of the ellipsoid. .

[0070] Define point cloud any point in The distance to the ellipsoid is Its calculation must satisfy the constraints of the ellipsoid, namely:

[0071] (2)

[0072] For point To the center of the ellipsoid The vector whose unit vector is . The ellipsoid is The length of the semi-axis in the direction is ,in , and yes Direction cosine, , and Similarly, satisfying

[0073] ;

[0074] To ensure the ellipsoid fits tightly to the shape of the reducer, the sum of the distances between all points needs to be minimized. Therefore, the objective function for optimization is:

[0075] (3)

[0076] Finally, particle swarm optimization (PSO) was used to solve the objective function, and after convergence, an ellipsoidal envelope that closely fits the reducer enclosure was obtained.

[0077] (3) Verification of the validity of the ellipsoidal envelope:

[0078] To quantify the fit of the ellipsoidal envelope to the reducer structure, an envelope redundancy rate is defined. As an evaluation metric, the ratio of the ellipsoidal volume to the actual volume of the reducer is used.

[0079] (4)

[0080] For the volume of the ellipsoid, This represents the actual volume of the speed reducer obtained through point cloud reconstruction. This indicates that the ellipsoid has almost no redundant space compared to the actual shape, and the number of subsequent viewpoints can be significantly reduced.

[0081] 3. Regarding step S3, "establishing a dual-constraint model including hard and soft constraints," this embodiment derives the effective working range of the acquisition distance based on the pinhole imaging principle and camera hardware parameters. The specific modeling process is as follows:

[0082] (1) Mathematical model of laser camera field of view.

[0083] Using a pinhole imaging model, given a sampling distance With laser camera intrinsic parameters (horizontal field of view) and vertical field of view The maximum field of view size for a single scan can be obtained by:

[0084] (5)

[0085] In the formula, and These represent the current sampling distance. Below, the maximum horizontal and vertical size of the scanning area that the laser camera can cover; used for subsequent judgment of "whether the entire local area has been captured".

[0086] (2) Dual constraint modeling based on acquisition distance.

[0087] 1) Hard constraints (hardware limits).

[0088] Laser cameras have a minimum effective line of sight. With maximum effective line of sight This parameter is determined by the camera sensor size. With lens resolution (Unit: pixels / mm) determines this. When the sampling distance is less than... At times, the camera sensor cannot capture a complete image (outside the sensor's range); when the acquisition distance is greater than... At this time, the laser camera cannot capture the details of the target surface (insufficient resolution). Its mathematical constraints are:

[0089] (6)

[0090] Among them, the minimum effective sight distance Determined by the sensor's maximum size and field of view, i.e., the sensor's horizontal size. The scanned area needs to be covered. Combining equation (5), we can derive:

[0091] (7)

[0092] Maximum effective line of sight Determined by the required detail resolution: Let the minimum feature size of the reducer surface (such as gear tooth pitch) be... It is necessary to ensure that the number of pixels in the laser camera sensor that represent this feature is not less than [a certain number]. (Usually 2-3 pixels are chosen to ensure feature recognition), and based on the relationship of imaging resolution, we can deduce:

[0093] (8)

[0094] Laser camera resolution and sampling distance The relationship is (Derived from the ratio of field of view width to sensor size), substituting into equation (8) yields the maximum effective viewing distance:

[0095] (9)

[0096] 2) Soft constraints (local region adaptation)

[0097] In addition to hard line-of-sight constraints, the acquisition distance Further matching of the dimensional characteristics of the local area of ​​the speed reducer is needed to avoid "too close to capture the entire area, too far to capture the details clearly." Let the maximum size of the currently scanned local area of ​​the speed reducer be... (e.g., a segment of length along the major axis or the diameter of the radial section), to ensure that a single scan can cover the entire local area, the laser camera's field of view width... Must be greater than or equal to Combining equation (5), we get: (Equation 10-14).

[0098] (10)

[0099] Meanwhile, to ensure clarity of details, the pixel density of the scanned area must meet the requirements: let the number of details to be identified in a local area be... The total number of pixels the camera captures in that area. Must be greater than or equal to (Every detail needs to be...) (pixels), combined We derive that:

[0100] (11)

[0101] Combining equations (6), (10), and (11), the final constraint interval for the acquisition distance can be obtained:

[0102] (12)

[0103] (13)

[0104] (14)

[0105] By solving the above equations, we can satisfy the camera's viewing distance limit while ensuring that a single scan covers the entire local area and clearly identifies all details.

[0106] 4. Based on the "construction of a comprehensive viewpoint evaluation function" in step S4, this embodiment iteratively selects the optimal viewpoint on the virtual observation surface by fusing distance, area, and uniformity indices. The specific process is as follows:

[0107] (1) Geometric relationship between ellipsoid envelope and viewpoint

[0108] Let the first The spatial coordinates of the secondary acquisition viewpoint are: Furthermore, the viewpoint direction always points towards the center of the workpiece. For the workpiece point cloud under this viewpoint... any point in ( (Point cloud point count), points To viewpoint The distance satisfies the geometric relationship:

[0109] (15)

[0110] Enclosing the workpiece with an ellipsoid (the ellipsoid parameters are consistent with those in S2), then Equivalent to viewpoint The distance to the surface of the ellipsoid. Then, according to equations (7) and (9), we get... This ensures that the imaging at each viewpoint is within the hard-soft constraint range.

[0111] (2) Coverage and Iteration Termination

[0112] Surface area of ​​the ellipsoid Based on this, define the first Model coverage after the second scan for:

[0113] (16)

[0114] in The area of ​​the covered ellipsoid surface (i.e., the union area of ​​the fields of view from all viewpoints) is the total area of ​​the ellipsoid. The total area of ​​the ellipsoidal envelope:

[0115] (17)

[0116] when ( When the preset coverage threshold (usually 99%) is reached, it is determined that the surface of the reducer has been "completely covered" and the viewpoint generation is terminated; otherwise, it is necessary to continue predicting the next optimal viewpoint.

[0117] (3) Optimal acquisition viewpoint prediction

[0118] Uncovered areas based on the current local model Construct a comprehensive evaluation function :

[0119] (18)

[0120] in, From the perspective to uncovered areas The shortest distance, the score for this item increases as the distance decreases, that is, prioritize the viewpoint that is close to the uncovered area to improve coverage efficiency; From the perspective Field of view and uncovered areas The area intersection accounts for The proportion (i.e.) This directly measures the ability to cover "blank areas"; Indicate viewpoint With existing The spatial distribution uniformity of viewpoints quantifies the minimum Euclidean distance between viewpoints, avoiding redundant scanning due to excessive concentration of viewpoints; weighting coefficients , and satisfy In this process, by maximizing Select the next optimal viewpoint.

[0121] 5. To achieve the viewpoint set sorting and smoothing process in step S5, this embodiment constructs a comprehensive scoring function that includes distance and direction preferences to generate the optimal acquisition sequence. The specific path optimization process is as follows:

[0122] Because the adaptively generated acquisition viewpoints in step S4 are discrete and difficult to apply in practice, it is necessary to sort the viewpoints to optimize the generated viewpoint set. Constructing the optimal acquisition sequence This approach balances movement efficiency with directional continuity. The core of viewpoint optimization ranking is a comprehensive scoring function:

[0123] (19)

[0124] in and These are weighting coefficients, satisfying... Distance rating For the normalized Euclidean distance:

[0125] (20)

[0126] Directional rating Considering clockwise / counterclockwise direction preference:

[0127] (twenty one)

[0128] in, From the perspective Azimuth relative to the center of the ellipsoid Indicates a preference for clockwise direction (can be switched to) (To achieve counterclockwise rotation).

[0129] The overall sorting process of S5 is as follows: starting from the initial viewpoint, iteratively select the candidate viewpoint with the highest comprehensive score and add it to the sequence until all viewpoints are sorted. This method shortens the robotic arm's movement path by weighted fusion of distance and direction, balancing acquisition efficiency and path smoothness.

[0130] from Figure 1Looking at the process, the entire workflow begins with inputting the point cloud of the reducer workpiece. First, the workpiece point cloud is preprocessed. Then, the accuracy of the workpiece positioning is checked. If inaccurate, the workpiece position is recalculated. If accurate, an ellipsoidal envelope is applied to the workpiece point cloud. If the envelope is incomplete, the ellipsoidal parameters are adjusted and the envelope is recalculated. After the envelope is complete, camera constraint criteria are set to generate an initial acquisition viewpoint. If the initial viewpoint is unqualified, the constraint criteria parameters are optimized and acquisition is repeated until the initial viewpoint is qualified. Next, the quality and spatial distribution of the acquired points are calculated to determine if the overall coverage reaches 99%. If not, the next optimal viewpoint is adaptively iterated and recalculated until the coverage reaches the target. Once the target is met, all viewpoints are sorted, and obstacle avoidance requirements are checked. If not, obstacle avoidance is achieved through an optimized sorting algorithm, and the process is re-verified. If obstacle avoidance requirements are met, the shortest path is checked. If not, the shortest path is achieved through an optimized sorting algorithm, and the process is re-verified. When both obstacle avoidance requirements and the shortest path are met, an efficient and feasible scanning path is output, and the workflow ends.

[0131] Figure 2 This is a schematic diagram of the camera's field of view imaging. 1 represents the laser camera, 2 represents the laser camera's emission point, and 3 and 4 represent the horizontal field of view angles, respectively. and vertical field of view 5 represents the sampling distance. 6 and 7 represent the current sampling distance, respectively. Below, the maximum horizontal and vertical size of the scanning area that the laser camera can cover. and The laser camera emits a laser beam from its emission point, defining a conical boundary of the scanning space with horizontal and vertical field of view angles, at a set acquisition distance. The maximum horizontal and vertical dimensions of the scan area that can be covered at the current distance are calculated using geometric relationships. Laser scanning is then performed within this coverage area to collect 3D point cloud data.

[0132] Figure 3 This diagram illustrates the selection of camera viewpoints. 1 represents the point cloud of the speed reducer workpiece; 2 represents the ellipsoidal envelope based on the contour of the speed reducer workpiece point cloud; 3 and 4 represent the k-th and (k+1)-th acquisition viewpoints, respectively; 5 and 6 represent the acquisition areas of the k-th and (k+1)-th acquisition viewpoints on the speed reducer surface, respectively; and 7 represents the distance between the k-th and (k+1)-th viewpoints. Based on the contour of the point cloud of the speed reducer workpiece, the corresponding ellipsoidal envelope is generated. The distance between the k-th and (k+1)-th acquisition viewpoints is... Under the constraints, corresponding acquisition areas are formed on the surface of the reducer. By switching different acquisition viewpoints in sequence, complete three-dimensional data acquisition of the surface of the reducer workpiece is achieved.

[0133] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "illustrative embodiment," "preferred embodiment," "detailed description," or "preferred embodiment," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0134] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features therein. Therefore, these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope defined by the claims of the present invention.

Claims

1. A laser scanning viewpoint adaptive planning method for data acquisition on the surface of a speed reducer, characterized in that, Including the following steps: S1. Construct a data acquisition system including a six-axis robot, a laser scanning camera, and an industrial control computer, and load the reducer contour information and camera hardware parameters. S2. Based on the reducer contour information, construct an adaptively stretched ellipsoidal envelope model as the geometric reference for viewpoint generation; S3. Based on the camera hardware parameters, establish a dual constraint model that includes hard constraints and soft constraints, and determine the effective working range of the camera on the ellipsoidal envelope model. S4. Based on the current point cloud coverage status, construct a viewpoint comprehensive evaluation function, and iteratively select the next optimal acquisition viewpoint within the effective working range according to the viewpoint comprehensive evaluation function until the preset coverage threshold is met. S5. Sort and smooth the generated viewpoint set to output a scan path without redundancy.

2. The method according to claim 1, characterized in that: S2 specifically includes the following steps: S21. Based on the reducer contour information, construct the equation of an ellipsoid with coordinate axes parallel to the principal axis; wherein, the center coordinates and semi-axis lengths of the ellipsoid equation correspond to the center and major semi-axis, middle semi-axis, and minor semi-axis of the reducer, respectively. S22. Obtain the bounding box point cloud of the reducer's position contour and construct an optimization objective function; the optimization objective function is to minimize the sum of the distances from any point in the bounding box point cloud to the ellipsoid; use the particle swarm optimization algorithm to solve the optimization objective function, obtain the center coordinates and semi-axis length parameters of the ellipsoid, and generate the ellipsoid envelope model; S23. Calculate the ratio of the volume of the ellipsoidal envelope model to the actual volume of the reducer, and use it as the envelope redundancy rate; if the envelope redundancy rate is within a preset threshold range, it is determined that the ellipsoidal envelope model fits the shape of the reducer, and the model construction is valid.

3. The method according to claim 1, characterized in that: S3 includes the following steps: S31. Based on the intrinsic parameters of the laser camera and the acquisition distance D, construct a mathematical model of the camera's field of view, and establish the spatial mapping relationship between the acquisition distance D and the maximum field of view size in a single scan, as a mathematical benchmark for calculating the hard constraint interval and the soft constraint interval. S32. Using the aforementioned spatial mapping relationship, and combining the sensor physical size of the laser camera and the minimum resolvable feature size corresponding to the lens resolution, calculate in reverse the minimum effective viewing distance D that satisfies imaging integrity. min With the maximum effective viewing distance D that satisfies the feature recognition resolution max Thus, a hard constraint interval [D] is established for the acquisition distance D. min D max ]; S33. For the geometric dimensions of the local area of ​​the reducer, the distance adaptation range that satisfies the complete coverage is calculated using the spatial mapping relationship, and a soft constraint interval for the acquisition distance D is established by combining the minimum pixel density threshold of the local details. S34. Perform an intersection operation on the hard constraint interval and the soft constraint interval to determine the final effective working interval of the camera acquisition distance.

4. The method according to claim 3, characterized in that: In step S31, the camera field of view mathematical model is constructed using the pinhole imaging model, and the maximum horizontal field of view size and the maximum vertical field of view size of a single scan are calculated based on the horizontal and vertical field of view angles of the laser camera at the current acquisition distance D. In step S32, the minimum effective line-of-sight distance D is calculated. min At that time, the established physical constraint is that the horizontal dimension of the camera sensor must cover the width of the scanning area at the corresponding acquisition distance; the maximum effective viewing distance D is calculated. max At that time, the established physical constraint is that the number of imaging pixels of the minimum feature size of the reducer surface on the camera sensor is not less than the preset feature recognition pixel threshold. In step S33, the soft constraint interval is established in the following way: Based on the geometric constraint that the camera's field of view is not less than the maximum size of the local area currently being scanned by the reducer, the first distance adaptation range is determined. The second distance adaptation range is determined based on the density constraint that the total number of pixels in the local area of ​​the camera is not less than the total number of pixels required for detail recognition in that area.

5. The method according to claim 3 or 4, characterized in that, S33 specifically includes the following steps: S331. Using the aforementioned spatial mapping relationship, establish the maximum field of view w for a single camera scan. h (That is, the maximum field of view w) h (and the maximum size L of the local area currently being scanned by the reducer) max Geometric constraint relationship: w h Greater than or equal to L max Based on this geometric constraint, the first distance adaptation range that satisfies the requirement of covering the entire local region is derived. S332. Combining the total number of pixels of the laser camera's sensor and the aforementioned spatial mapping relationship, establish the total number of pixels N in the local area image and the pixel threshold N required for detail recognition in that area. target Density constraint relationship: N is greater than or equal to N target Based on this density constraint, a second distance adaptation range that satisfies the requirement of clear detail recognition is derived. S333. Find the intersection of the first distance adaptation range and the second distance adaptation range to generate a soft constraint interval for the acquisition distance D.

6. The method according to claim 3, characterized in that, S4 includes the following steps: S41. Determine the acquisition distance within the final effective working range, construct a virtual camera observation surface that maintains the acquisition distance with the surface of the ellipsoidal envelope model, and use it as a candidate viewpoint set space; S42, Based on the currently uncovered area U k Based on the distribution of the virtual camera observation surface, a comprehensive viewpoint evaluation function F(V) is constructed. Points that maximize F(V) are then selected as the next optimal acquisition viewpoint V. k+1 Wherein, the comprehensive evaluation function F(V) is the weighted fusion result of each evaluation index, and the evaluation index includes: The weighting coefficients for each evaluation indicator; Viewpoint to uncovered area U k The shortest distance, and the closer the distance, the larger the value of this indicator; The percentage of the area where the viewpoint's field of view intersects with the uncovered area; Spatial uniformity of the front viewpoint and the already generated set of viewpoints; S43. Calculate in real time the ratio of the area of ​​the currently covered ellipsoid surface to the total area of ​​the ellipsoid envelope to obtain the model coverage rate Cov; if Cov reaches the preset coverage rate threshold, terminate viewpoint generation, otherwise return to S42 to continue iteration.

7. The method according to claim 6, characterized in that, S5 includes: S51. Calculate the distance score based on the normalized Euclidean distance between the candidate viewpoint and the current viewpoint, and calculate the direction score based on the change of the azimuth angle of the candidate viewpoint relative to the current viewpoint in the ellipsoidal central coordinate system. Then, weight and fuse the distance score and the direction score to construct a comprehensive viewpoint score function. S52. Starting from the initial viewpoint, iteratively select the viewpoint that gives the highest comprehensive score function value from the candidate viewpoints as the next acquisition viewpoint, until all viewpoints are sorted and an acquisition sequence is generated. S53. Based on the generated acquisition sequence, output a smooth scanning path.

Images